<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>Siril | 星空の風景, 日々の風景</title>
	<atom:link href="https://globe3.ddns.net/wp/category/starry-sky/siril/feed/" rel="self" type="application/rss+xml" />
	<link>https://globe3.ddns.net/wp</link>
	<description>星空のこと. 日々に想うこと.</description>
	<lastBuildDate>Tue, 17 Nov 2026 07:42:50 +0000</lastBuildDate>
	<language>ja</language>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	<generator>https://wordpress.org/?v=7.1.4</generator>
	<item>
		<title>自分好みに合わせる, Siril Script</title>
		<link>https://globe3.ddns.net/wp/customizing-siril-script-my-liking/</link>
					<comments>https://globe3.ddns.net/wp/customizing-siril-script-my-liking/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Thu, 17 Sep 2026 07:24:57 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=6132</guid>

					<description><![CDATA[Siril , [Stacking]各設定よりもスクリプト Sirilで天体画像を処理していく時に、画面にある、[Calibration] や、[Registration]、[Registration]などのパラメータを [&#8230;]]]></description>
										<content:encoded><![CDATA[
<h2 class="wp-block-heading"> Siril , [Stacking]各設定よりもスクリプト</h2>



<p class="wp-block-paragraph">Sirilで天体画像を処理していく時に、画面にある、<br>[Calibration] や、[Registration]、[Registration]などのパラメータを調整したり、<br>[Stacking]では、<br>&#8220;Stacking methods&#8221;「積み重ねの方法」には、&#8221;Average Stacking With Rejection&#8221;<br>&#8220;Rejection methods&#8221;「除去方法」には、&#8221;MAD Clipping&#8221;<br>&#8220;weighting&#8221;「」には、&#8221;weighting&#8221;<br>などを細部に指定すると、好みの前処理が行なえます。</p>



<p class="wp-block-paragraph">しかし、手順が増える。。。<br>無精な自分は、<br>スクリプトに指定を埋め込んで、処理をします。その為のスクリプト変更についてのメモです。</p>



<h2 class="wp-block-heading">Sirilスクリプト</h2>



<p class="wp-block-paragraph">Sirilスクリプトは下記URLにあります。これを元に、変更していきます。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Siril Copyright © 2012-2023, Team free-astro <br>&lt;https://free-astro.org/index.php/Siril>,  &lt;https://www.siril.org><br><a rel="noopener" target="_blank" href="https://gitlab.com/free-astro/siril/">https://gitlab.com/free-astro/siril/</a></p>



<p class="wp-block-paragraph">Siril Scripts<br><a rel="noopener" target="_blank" href="https://gitlab.com/free-astro/siril/-/tree/master/scripts">https://gitlab.com/free-astro/siril/-/tree/master/scripts</a></p>
</blockquote>



<p class="wp-block-paragraph">Sirilスクリプト、及び、 Pythonスクリプトは以下URLにあります。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Siril 外部スクリプトリポジトリ<br>このリポジトリは、画像の操作や処理のさまざまな段階を効率化するために、皆様のようなユーザーから提供されたスクリプトを集めたものです。前処理作業の効率化や、複雑な処理操作の最適化をお考えの場合でも、ここで役立つスクリプトが見つかるはずです。<br><a rel="noopener" target="_blank" href="https://gitlab.com/free-astro/siril-scripts">https://gitlab.com/free-astro/siril-scripts</a></p>
</blockquote>



<h2 class="wp-block-heading">[Stacking]各設定</h2>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Average Stacking With Rejection<br>This method of stacking computes a mean of the pixels in a stack after having excluded deviant pixels and an optional normalisation of the images against the reference image. As for sum stacking, the improvement in SNR is proportional to √𝑁<br>. There are several ways to normalize the images and several ways to detect and replace or exclude deviant pixels, explained below.<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/preprocessing/stacking.html#stacking-methods">https://siril.readthedocs.io/en/latest/preprocessing/stacking.html#stacking-methods</a><br><br>このスタック法では、異常なピクセルを除外した後、スタック内のピクセルの平均値を算出し、必要に応じて参照画像に対して画像の正規化を行います。和スタックと同様に、SNRの改善度は√𝑁に比例します。<br>画像の正規化にはいくつかの方法があり、異常なピクセルの検出・置換または除外にもいくつかの方法があります。これらについては以下で説明します。</p>
</blockquote>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph"><strong>Percentile Clipping</strong>:パーセンタイル・クリッピング：これは、小規模なデータセット（最大6画像）に最適な1ステップの除去アルゴリズムです。</p>



<p class="wp-block-paragraph"><strong>Sigma Clipping</strong>:シグマ・クリッピング：これは反復アルゴリズムであり、中央値からの距離が、シグマ単位で指定された2つの値（𝜎 low、𝜎 high）よりも遠いピクセルを除去します。</p>



<p class="wp-block-paragraph"><strong>MAD Clipping</strong>:MADクリッピング：これはシグマクリッピングと同様に動作する反復アルゴリズムですが、使用される推定量が中央値絶対偏差（MAD）である点が異なります。これは一般的に、ノイズの多い赤外線画像の処理に使用されます。</p>



<p class="wp-block-paragraph"><strong>Median Sigma Clipping</strong>:中央値シグマクリッピング：これはシグマクリッピングと同じアルゴリズムですが、除外されたピクセルがスタックの中央値で置き換えられる点が異なります。</p>



<p class="wp-block-paragraph"><strong>Winsorized Sigma Clipping</strong>:ウィンザー化シグマクリッピング：これはシグマクリッピング法と非常に似ていますが、外れ値の検出に対してより頑健であると考えられています。Huber の研究を参照してください。</p>



<p class="wp-block-paragraph"><strong>Generalized Extreme Studentized Deviate Test</strong> :これは、近似的に正規分布に従う単変量データセットから 1 つまたは複数の外れ値を検出するために使用される、グラブス検定の一般化です。このアルゴリズムは、50枚以上の画像からなる大規模なデータセットにおいて優れた性能を発揮する。</p>



<p class="wp-block-paragraph"><strong>Linear Fit Clipping</strong> :線形フィット・クリッピング：ピクセルスタックに最適な直線（𝑦 =𝑎⁢𝑥 +𝑏）をフィッティングし、外れ値を排除する。このアルゴリズムは、大規模なスタックや、空間分布や方向が異なる空のグラデーションを含む画像において、非常に良好な性能を発揮する。<br><br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/preprocessing/stacking.html#rejection-methods">https://siril.readthedocs.io/en/latest/preprocessing/stacking.html#rejection-methods</a></p>
</blockquote>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Note that for drizzled CFA images, you may find difficulty in achieving satisfactory rejection in the Red and Blue channels with some methods. In these cases the MAD Clipping method appears to be the most effective.<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/preprocessing/stacking.html#rejection-methods">https://siril.readthedocs.io/en/latest/preprocessing/stacking.html#rejection-methods</a><br>ドリズル処理されたCFA画像の場合、一部の手法では赤チャンネルおよび青チャンネルで満足のいくノイズ除去が困難な場合があります。このような場合、MADクリッピング法が最も効果的であるようです。</p>
</blockquote>



<p class="wp-block-paragraph">この記述があったので、MAD Clippingを用いることにしました。確かにここで現れるノイズに悩まされました。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Images filtering/weighting<br>weighted FWHM：これは単純なFWHMを改良したものです。参照画像と比較して検出された星の数を用いることで、より多くの偽画像を除外することができます（星ベースのレジストレーションのみ）。<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/stable/preprocessing/stacking.html#image-rejection">https://siril.readthedocs.io/en/stable/preprocessing/stacking.html#image-rejection</a></p>
</blockquote>



<h2 class="wp-block-heading">Siril Script が意味するところ</h2>



<p class="wp-block-paragraph">Siril  1.5向けに行います。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">OSC_Preprocessing_BayerDrizzle.ssf<br><a rel="noopener" target="_blank" href="https://gitlab.com/free-astro/siril/-/blob/master/scripts/OSC_Preprocessing_BayerDrizzle.ssf?ref_type=heads">https://gitlab.com/free-astro/siril/-/blob/master/scripts/OSC_Preprocessing_BayerDrizzle.ssf?ref_type=heads</a></p>
</blockquote>



<h3 class="wp-block-heading">OSC_Preprocessing_BayerDrizzle.ssf</h3>



<p class="wp-block-paragraph">バイアス、フラット、ダークはひとまず、オリジナルのまま<br>Cosmetic Correctionを行わない<br>シグマクリップには、MADクリップを用いる<br>weighted FWHM<br>処理ファイル保存には、日付と時間をファイル名に入れる</p>



<h4 class="wp-block-heading">calibrate</h4>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">calibrate<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/Commands.html#calibrate">https://siril.readthedocs.io/en/latest/Commands.html#calibrate</a></p>
</blockquote>



<pre class="wp-block-code"><code>calibrate sequencename &#91;-bias=filename] &#91;-dark=filename] &#91;-flat=filename] &#91;-cc=dark &#91;siglo sighi] || -cc=bpm bpmfile] &#91;-cfa] &#91;-debayer] &#91;-fix_xtrans] &#91;-equalize_cfa] &#91;-opt&#91;=exp]] &#91;-all] &#91;-prefix=] &#91;-fitseq]</code></pre>



<p class="wp-block-paragraph">-cc=dark: これを行いたくないので外します。<br>マスターダーク（-dark= オプションでマスターダークを指定する必要があります）からホットピクセルとコールドピクセルを検出できます。</p>



<p class="wp-block-paragraph">-equalize_cfa: <br>キャリブレーション済み画像の色かぶりを防ぐためにマスターフラットのRGBレイヤーの平均輝度を均一化する。</p>



<h4 class="wp-block-heading">register</h4>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">register<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/Commands.html#register">https://siril.readthedocs.io/en/latest/Commands.html#register</a></p>
</blockquote>



<pre class="wp-block-code"><code>register sequencename &#91;-2pass] &#91;-selected] &#91;-prefix=] &#91;-scale=]
register sequencename ... &#91;-layer=] &#91;-transf=] &#91;-minpairs=] &#91;-maxstars=] &#91;-nostarlist] &#91;-disto=]
register sequencename ... &#91;-interp=] &#91;-noclamp] &#91;-extref=filepath]
register sequencename ... &#91;-drizzle &#91;-pixfrac=] &#91;-kernel=] &#91;-flat=]]</code></pre>



<h5 class="wp-block-heading">-kernel=square と -kernel=gaussian の違い</h5>



<p class="wp-block-paragraph">-kernel=square　：明確な四角状<br>□□<br>□□</p>



<p class="wp-block-paragraph">-kernel=gaussian　：中心が明確、周辺は弱い<br>ガウシアン状<br></p>



<h4 class="wp-block-heading">stack</h4>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">stack<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/Commands.html#stack">https://siril.readthedocs.io/en/latest/Commands.html#stack</a></p>
</blockquote>



<pre class="wp-block-code"><code>stack seqfilename
stack seqfilename { sum | min | max } &#91;-output_norm] &#91;-out=filename] &#91;-maximize] &#91;-upscale] &#91;-32b]
stack seqfilename { med | median } &#91;-nonorm, -norm=] &#91;-fastnorm] &#91;-rgb_equal] &#91;-output_norm] &#91;-out=filename] &#91;-32b]
stack seqfilename { rej | mean } &#91;rejection type] &#91;sigma_low sigma_high]  &#91;-rejmap&#91;s]] &#91;-nonorm, -norm=] &#91;-fastnorm] &#91;-overlap_norm] &#91;-weight={noise|wfwhm|nbstars|nbstack}] &#91;-feather=] &#91;-rgb_equal] &#91;-output_norm] &#91;-out=filename] &#91;-maximize] &#91;-upscale] &#91;-32b]</code></pre>



<p class="wp-block-paragraph">Stacking with rejection:</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Stacking with rejection:<br>Types rej or mean require the use of additional arguments for rejection type and values. The rejection type is one of n[one], p[ercentile], s[igma], m[edian], w[insorized], l[inear], g[eneralized], [m]a[d] for Percentile, Sigma, Median, Winsorized, Linear-Fit, Generalized Extreme Studentized Deviate Test or k-MAD clipping. If omitted, the default Winsorized is used.<br>The sigma low and sigma high parameters of rejection are mandatory unless none is selected.<br>Optionally, rejection maps can be created, showing where pixels were rejected in one (-rejmap) or two (-rejmaps, for low and high rejections) newly created images.<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/Commands.html#stack">https://siril.readthedocs.io/en/latest/Commands.html#stack</a><br><br>rej または mean タイプでは、除去の種類と値を示す追加の引数が必要です。除外タイプは、n[one]、p[ercentile]、s[igma]、m[edian]、w[insorized]、l[inear]、g[eneralized]、[m]a[d]（パーセンタイル、シグマ、中央値、ウィンソライズ、線形フィット、一般化極値スチューデント偏差検定、または k-MAD クリッピング）のいずれかです。<br>省略された場合、デフォルトの「Winsorized」が使用されます。<br>「rejection」の sigma low および sigma high パラメータは、何も選択されていない場合を除き、必須です。<br>オプションとして、1枚（-rejmap）または2枚（-rejmaps、低値および高値の除去用）の新規作成された画像に、ピクセルが除去された箇所を示す除去マップを作成することができます。</p>
</blockquote>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Other options for rejection stacking:<br>Weighting can be applied to the images of the sequences using the option -weight= followed by:<br>noise to add larger weights to frames with lower background noise.<br>nbstack to weight input images based on how many images were used to create them, useful for live stacking.<br>nbstars or wfwhm to weight input images based on number of stars or wFWHM computed during registration step.<br>-feather= option will apply a feathering mask on each image borders over the distance (in pixels) given in argument.<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/Commands.html#stack">https://siril.readthedocs.io/en/latest/Commands.html#stack</a></p>



<p class="wp-block-paragraph">-weight= オプションの後に以下の指定を行うことで、シーケンスの画像に重み付けを行うことができます。<br>noise：背景ノイズの少ないフレームに大きな重みを追加します。<br>nbstack：入力画像の作成に使用された画像数に基づいて重み付けを行います。ライブスタッキングに有用です。<br>nbstars または wfwhm：登録ステップで計算された星の数または wFWHM に基づいて、入力画像に重み付けを行います。<br>-feather= オプションを使用すると、引数で指定された距離（ピクセル単位）分、各画像の境界にフェザリングマスクが適用されます。</p>
</blockquote>



<h5 class="wp-block-heading">FWHM = Full Width at Half Maximum</h5>



<p class="wp-block-paragraph">FWHMが低いと星像が鋭く、シーイングが良い。<br>FWHMが高いと星像がボケている、ピントや追尾が悪い。</p>



<p class="wp-block-paragraph">wfwhm<br>-weight=wfwhmは、FWHMと星の数を組み合わせて、各画像に重みを付けるらしい。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Weighted FWHM weights individual frames based on wFWHM computed during registration step. This is a FWHM weighted by the number of stars in the image. For the same FWHM measurement, an image with more stars will have a better wFWHM than an image with fewer stars.<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/ja/latest/preprocessing/stacking.html?utm_source=chatgpt.com#images-filtering-weighting">https://siril.readthedocs.io/ja/latest/preprocessing/stacking.html?utm_source=chatgpt.com#images-filtering-weighting</a></p>



<p class="wp-block-paragraph">重み付けFWHMは、位置合わせステップで算出されたwFWHMに基づいて、個々のフレームに重み付けを行います。これは、画像内の星の数で重み付けされたFWHMです。同じFWHMの測定値であっても、星の多い画像ほど、星の少ない画像よりも優れたwFWHM値となります。</p>
</blockquote>



<h2 class="wp-block-heading">スクリプトを変更する</h2>



<p class="wp-block-paragraph">OSC_Preprocessing_BayerDrizzle を変更します。</p>



<h3 class="wp-block-heading">コスメティック補正（Cosmetic Correction）を行わない</h3>



<h3 class="wp-block-heading">Align lights with Drizzle</h3>



<h4 class="wp-block-heading">square オプション</h4>



<h4 class="wp-block-heading">gaussian オプション</h4>



<h3 class="wp-block-heading">Stack calibrated lights</h3>



<h4 class="wp-block-heading">MAD Clipping 3 3</h4>



<h4 class="wp-block-heading">低いnoise画像へ重み付け</h4>



<h4 class="wp-block-heading">wFWHM重み付け</h4>



<h4 class="wp-block-heading">保存名、撮影時間と時刻</h4>



<p class="wp-block-paragraph">以下は、複数の条件を載せています。後日、別にします。</p>



<p class="wp-block-paragraph">Syntax example<br>save時の構文は以下URLを参照します。<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/zh-tw/latest/Pathparsing.html?utm_source=chatgpt.com">https://siril.readthedocs.io/zh-tw/latest/Pathparsing.html?utm_source=chatgpt.com</a></p>



<p class="wp-block-paragraph">例えば、天体名を表示するならば、</p>



<pre class="wp-block-code"><code>$OBJECT:%s$_</code></pre>



<pre class="wp-block-code"><code>############################################
#
# Script for Siril 1.4
# December 2024
# (C) Cyril Richard
# OSC_Preprocessing_WBayerDrizzle v1.1
#
########### PREPROCESSING SCRIPT ###########
#
# Script for color camera preprocessing
# aligning using Bayer Drizzle
#
# Needs 4 sets of RAW images in the working
# directory, within 4 directories:
#   biases/
#   flats/
#   darks/
#   lights/
# Saves masters to ./masters/
#
############################################

requires 1.3.4

# Convert Bias Frames to .fit files
cd biases
convert bias -out=../process
cd ../process

# Stack Bias Frames to bias_stacked.fit
stack bias rej 3 3 -nonorm -out=../masters/bias_stacked
cd ..

# Convert Flat Frames to .fit files
cd flats
convert flat -out=../process
cd ../process

# Calibrate Flat Frames
calibrate flat -bias=../masters/bias_stacked

# Stack Flat Frames to pp_flat_stacked.fit
stack pp_flat rej 3 3 -norm=mul -out=../masters/pp_flat_stacked
cd ..

# Convert Dark Frames to .fit files
cd darks
convert dark -out=../process
cd ../process

# Stack Dark Frames to dark_stacked.fit
stack dark rej 3 3 -nonorm -out=../masters/dark_stacked
cd ..

# Convert Light Frames to .fit files
cd lights
convert light -out=../process
cd ../process

##### Calibrate Light Frames #####
### Calibrate Light Frames
### calibrate light -dark=../masters/dark_stacked -flat=../masters/pp_flat_stacked -cc=dark -cfa -equalize_cfa

# コスメティック補正（Cosmetic Correction）を行わない
calibrate light -dark=../masters/dark_stacked -flat=../masters/pp_flat_stacked  -cfa -equalize_cfa
####################################

##### Align lights with Drizzle Section #####
# Align lights with Drizzle, square オプション
register pp_light -drizzle -scale=1.0 -pixfrac=1.0 -kernel=square -flat=../masters/pp_flat_stacked

# Align lights with Drizzle, gaussian オプション
### register pp_light -drizzle -scale=1.0 -pixfrac=1.0 -kernel=gaussian -flat=../masters/pp_flat_stacked
#############################################

##### Stack calibrated lights Section #####

### Stack calibrated lights to result.fit
###stack r_pp_light rej 3 3 -norm=addscale -output_norm -rgb_equal -32b -out=result

### Stack calibrated lights, MAD Clipping 3 3
### stack r_pp_light rej a 3 3 -norm=addscale -output_norm  -rgb_equal -32b -out=result

### Stack calibrated lights, MAD Clipping 4 4
### stack r_pp_light rej a 4 4 -norm=addscale -output_norm  -rgb_equal -32b -out=result

# Stack calibrated lights, MAD Clipping 4 4, 低いnoise画像へ重み付け
stack r_pp_light rej a 4 4 -weight=noise -norm=addscale -output_norm  -rgb_equal -32b -out=result

### Stack calibrated lights, MAD Clipping 3 3, wFWHM重み付け
### stack r_pp_light rej a 4 4 -weight=wfwhm -norm=addscale -output_norm  -rgb_equal -32b -out=result

### flipping無しならば、outでファイル名指定
### stack r_light rej mad 3 3 -norm=addscale -weight=noise -output_norm -rgb_equal -32b -out=../final_drizzle_$LIVETIME:%d$s_$DATE:dt$
############################################

# and flipping if required
load result
mirrorx -bottomup

### 保存名、撮影時間と時刻
### 撮影時間のみ
### save ../result_drizzle_$LIVETIME:%d$s

### 撮影時間と時刻
### $LIVETIME と $DATE を組み合わせ
### save ../result_drizzle_$LIVETIME:%d$s_$DATE:dt$

# 天体名と撮影時間と時刻
### $OBJECT と $LIVETIME 、 $DATE を組み合わせ
save ../$OBJECT:%s$_drizzle_$LIVETIME:%d$s_$DATE:dt$

cd ..
close</code></pre>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/customizing-siril-script-my-liking/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Siril 画像処理の私的なメモ</title>
		<link>https://globe3.ddns.net/wp/siril-image-processing-workflow-2026-06/</link>
					<comments>https://globe3.ddns.net/wp/siril-image-processing-workflow-2026-06/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Sat, 12 Sep 2026 01:00:24 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<category><![CDATA[画像処理]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=3453</guid>

					<description><![CDATA[これは、Siril を用いた天体画像処理の一例です。 以下に、Siril 1.5や VeraLux 向けに、Siril Python Script コードを追い、その仕組みをメモしています。記載と編集を随時行っています。 [&#8230;]]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">これは、Siril を用いた天体画像処理の一例です。</p>



<p class="wp-block-paragraph">以下に、Siril 1.5や VeraLux  向けに、Siril Python Script コードを追い、その仕組みをメモしています。<br>記載と編集を随時行っています。画像処理の仕方以外の記事が多いと感じるでしょう。</p>



<h2 class="wp-block-heading">NGC2685</h2>



<figure class="wp-block-image size-full"><img fetchpriority="high" decoding="async" width="800" height="719" src="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/NGC2685_Crop-AGR-SPCC_GraX0.96-SyQonprism-_IHS-Link_GHS-BP_Wavelets_GHS_result_800.jpg" alt="" class="wp-image-6008" srcset="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/NGC2685_Crop-AGR-SPCC_GraX0.96-SyQonprism-_IHS-Link_GHS-BP_Wavelets_GHS_result_800.jpg 800w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/NGC2685_Crop-AGR-SPCC_GraX0.96-SyQonprism-_IHS-Link_GHS-BP_Wavelets_GHS_result_800-300x270.jpg 300w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/NGC2685_Crop-AGR-SPCC_GraX0.96-SyQonprism-_IHS-Link_GHS-BP_Wavelets_GHS_result_800-150x135.jpg 150w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/NGC2685_Crop-AGR-SPCC_GraX0.96-SyQonprism-_IHS-Link_GHS-BP_Wavelets_GHS_result_800-768x690.jpg 768w" sizes="(max-width: 800px) 100vw, 800px" /></figure>



<p class="wp-block-paragraph">NASAが公開する画像の様に、極リング内部を強調した処理を行いました。</p>



<p class="wp-block-paragraph">Name: NGC 2685<br>Date: 2026-03-20T19-36, 2026-03-21T20-56<br>SQM: 21.64 mag/arcsec2 (Yamanashi)<br>Optics: Takahashi TS-160, HEUIB-II<br>Camera: ASI533MC Pro, gain 100, offset 20, -20deg<br>Exposure: 180s x 118 (total 5h54m)</p>



<p class="wp-block-paragraph">NGC2685画像処理<br>勾配補正、色校正、 SyQon Starless 恒星分離, MTF Stretchを用いて分離：銀河星雲分離, IHSを用いて分離<br>Deconvolution：GraXpert<br>ノイズ除去：SyQon Prism<br>恒星と銀河星雲を再合成：VeraLux Star Composer</p>



<p class="wp-block-paragraph">SyQon を多様すると今までの作風と、ちょっと色合いが変わっていく感じがしてます。</p>



<h2 class="wp-block-heading">Arp 273</h2>



<figure class="wp-block-image size-full"><img decoding="async" width="1200" height="1200" src="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/Arp273_HEUIB-II_2026-01-11-12-16-17_9h48m_StarComposer.jpg" alt="" class="wp-image-6527" srcset="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/Arp273_HEUIB-II_2026-01-11-12-16-17_9h48m_StarComposer.jpg 1200w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/Arp273_HEUIB-II_2026-01-11-12-16-17_9h48m_StarComposer-300x300.jpg 300w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/Arp273_HEUIB-II_2026-01-11-12-16-17_9h48m_StarComposer-150x150.jpg 150w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/Arp273_HEUIB-II_2026-01-11-12-16-17_9h48m_StarComposer-768x768.jpg 768w" sizes="(max-width: 1200px) 100vw, 1200px" /></figure>



<p class="wp-block-paragraph">Name: Arp 273<br>Date: 2026-01-11T22-45, 2026-01-12T19-30, 2026-01-16T19-04, 2026-01-17T19-03<br>SQM: 21.64 mag/arcsec2 (Yamanashi)<br>Optics: Takahashi TS-160, HEUIB-II<br>Camera: ASI533MC Pro, gain 100, offset 20, -15deg<br>Exposure: 180s x 196 (total 9h48m)</p>



<p class="wp-block-paragraph">Arp273画像処理<br>勾配補正、色校正、分離：StarNet CLI tools<br>ノイズ除去：SyQon Prism<br>Deconvolution：GraXpert<br>恒星と銀河星雲を再合成：VeraLux Star Composer</p>



<p class="wp-block-paragraph">Pythonスクリプト機能が導入されてからは、Sirilから行えることが広がっています。</p>



<h2 class="wp-block-heading">Siril 画像処理の流れの例</h2>



<ul class="wp-block-list">
<li>勾配補正、色校正など
<ul class="wp-block-list">
<li>Astrometry: 天体位置</li>



<li>Crop: トリミング</li>



<li>Auto Gradient Removal : 勾配補正（又は、 VeraLux Nox (Background Extraction)、 GraXpert BE）</li>



<li>SPCC (Gaia Spectrum): 恒星スペクトルデータを用いた色校正</li>



<li>Cosmetic Correction (CFAあり): 必要に応じて</li>



<li>別名でファイル保存</li>
</ul>
</li>
</ul>



<ul class="wp-block-list">
<li>恒星処理
<ul class="wp-block-list">
<li>Desaturate Stars: 必要に応じて</li>



<li>Aberrations Remover: 必要に応じて</li>



<li>Deconvolution: Cosmic Clarity Sharpening,  Stellar Only (Stellar Amount: 0.16 、0.32など)</li>



<li>恒星除去：恒星部分を分離
<ul class="wp-block-list">
<li>StarNet CLI tools</li>



<li>SyQon Starless
<ul class="wp-block-list">
<li>Stretch Options<br>Method: MTF Stretch (MTF = Midtone Transfer Function)<br>Median Target: <br><br>例）MTF Stretch<br>　　　Linked stretch (preserve star colors) → ON<br>　　　Median Target → 0.050 ～ 0.2<br><br>経験則から、以下の記述をしています。<br>Median = 0.050, 0.075<br>弱いストレッチ：淡い構造のコントラストが強調された。恒星状になってしまう。<br><br>Median = 0.15 以上<br>強いストレッチ：ガス塊や星雲の細かく淡い構造は保護された。</li>



<li>Statistical Stretch：統計的ストレッチ</li>
</ul>
</li>
</ul>
</li>
</ul>
</li>



<li>銀河星雲処理
<ul class="wp-block-list">
<li>Deconvolution: 対象銀河星雲付近の恒星からDynamic PSF、FITSデータからCDELT1,2を、この2つの値から、PSF値を求める。
<ul class="wp-block-list">
<li>GraXpert (Strength=0.80など)</li>



<li>Cosmic Clarity Sharpening, Object</li>
</ul>
</li>



<li>ノイズ除去（VeraLux Silentium（GHS前に適応）、若しくは、DeepSNR denoising）
<ul class="wp-block-list">
<li>VeraLux Silentium：線形処理に設計された、AI非依存のノイズ除去</li>



<li>SyQon Prism (Strength=0.32, Taget Softなど)</li>



<li>DeepSNR denoising : </li>
</ul>
</li>



<li>恒星除去：銀河星雲部分を分離
<ul class="wp-block-list">
<li>StarNet CLI tools</li>



<li>SyQon Starless
<ul class="wp-block-list">
<li>IHS</li>
</ul>
</li>
</ul>
</li>



<li>ストレッチ: 
<ul class="wp-block-list">
<li>GHS (Stretch factor, Local stretch intensityを高める場合もある)
<ul class="wp-block-list">
<li>Linear stretch (BP shift)は行わない。最終に行う。</li>
</ul>
</li>



<li>VeraLux Hyper Metric Stretch (HMS)</li>
</ul>
</li>



<li>Wavelets Transformation</li>



<li>ノイズ除去：2度目のノイズ除去（VeraLux SilentiumはGHS後なので条件に合わない）
<ul class="wp-block-list">
<li>SyQon Prism (Strength=0.32, Taget Softなど)</li>



<li>GraXpert Denoising (Strength=0.16, 0.32 など）</li>
</ul>
</li>



<li>色相補正（Siril: Color saturation, threshold 1.00は未使用）
<ul class="wp-block-list">
<li>VeraLux Curves：Akima補間法を用いて、RGB、Lab、LCH、HSVのコントラストを輝度構造と色の強度を個別に調整</li>



<li>VeraLux Vectra：LCH（明度、彩度、色相）色空間での補正</li>
</ul>
</li>



<li>ストレッチ
<ul class="wp-block-list">
<li>GHS : BP Shift (Black Point Shift)</li>
</ul>
</li>
</ul>
</li>



<li>恒星、銀河星雲再合成
<ul class="wp-block-list">
<li>VeraLux Star Composer（以前は、Star recomposition）</li>
</ul>
</li>



<li>色相補間
<ul class="wp-block-list">
<li>LCh色空間: VeraLux Vectra</li>
</ul>
</li>
</ul>



<h2 class="wp-block-heading">勾配補正</h2>



<h3 class="wp-block-heading">Auto Gradient Removal</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">背景除去は、その役割上、常に大スケールのコントラストを除去します。そのため、暗い領域が平坦化されていると感じる場合は、通常、モデルが柔らかすぎて実際の信号を削り取っていることを意味します。</p>



<p class="wp-block-paragraph">その場合は、「Scale」を高くするか、「structure protection」をオンにしたままにするか、あるいは低度の「simplified model」に切り替えてください。<br>経験則として、実際に目に見える勾配をまだ除去できる範囲で、最も硬いモデルを使用することをお勧めします。<br><a rel="noopener" target="_blank" href="https://www.facebook.com/groups/841870683930320/posts/1490938619023520/">https://www.facebook.com/groups/841870683930320/posts/1490938619023520/</a><br><br>作者であるCyril Richardさんのコメントです。</p>
</blockquote>



<p class="wp-block-paragraph">Auto Gradient Removal は、数値を変更後は、繰り返して、[Apply]を押して行います。<br>画像を初めに戻す必要はありません。</p>



<p class="wp-block-paragraph">Cyril Richardさんのコメントにある『最も硬いモデル』は、Model degreeの数値を上げる事かな。</p>



<ul class="wp-block-list">
<li>Scale:  マルチスケールモデルの平滑化半径、ピクセル数。初めは、&#8221;5&#8243;です。
<ul class="wp-block-list">
<li>数値が大きほど、広範囲の大きな勾配のみ除去される。</li>



<li>数値が小さいと、小さく局所な変化を追える。</li>



<li>光害のにじみや局所なグラデーションがあるならば、数値を上げる。</li>
</ul>
</li>



<li>Smoothness:  追加のぼかし処理。
<ul class="wp-block-list">
<li>数値が大きいと滑らかで穏やかになる。</li>



<li>補正後の背景に斑点状のアーティファクトが見られるならば、数値を下げます。</li>
</ul>
</li>



<li>Simplefied model: 背景が無い画像、例えば画像いっぱいの銀河などに有効です。
<ul class="wp-block-list">
<li>Model degree: 初めは2です。</li>
</ul>
</li>



<li>Structure protection:  星雲や銀河など広範囲の明るい部分を検出し、それらを勾配補正から外す。
<ul class="wp-block-list">
<li>Protection threshold:  初めは、&#8221;0.05&#8243;です。<br>数値が低いほど感度が高くなり保護効果が高くなる。<br>数値が高いと明るい天体のみ保護される。暗い星雲は背景とされ、差し引かれるかもしれない。</li>



<li>Protection amount: 初めは、&#8221;0.50&#8243;です。保護マスクの広がり。数値が大きと広範囲。<br>補正後に明るい恒星や星雲周辺にかすかなリング状やハロー状のアーティファクトが見られるならば、　</li>
</ul>
</li>



<li>Downsample:  初めは、&#8221;4&#8243;です。
<ul class="wp-block-list">
<li>プレビューを手早く見たいときや、フルサイズの画像処理を早く行うなどの場合は&#8221;8&#8243;にします。</li>
</ul>
</li>



<li>Mode : 減算を行うか、加算を行うか。減算は光害、月の明かり、霧などの補正に通常は使います。 </li>
</ul>



<h2 class="wp-block-heading">収差補正</h2>



<h3 class="wp-block-heading">Aberrations Remover</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">このモデルは、バックグラウンド除去、及び、色補正後の線形画像を対象としています。<br>最良の結果を得るためには、デコンボリューションやノイズ低減処理を行う前に適用して下さい。<br><br>通常、これらのパラメータを調整する必要はありません。<br>strength: 収差補正の強度を調整します。<br>Dark ringing reduction: 強力な補正の副作用として明るい星の周囲に現れる可能性のあるダークリンギングアーティファクトを抑制するために、追加のフィルターを適用します。このオプションは、そのようなアーティファクトが確認できる場合にのみ有効にして下さい。<br>Protect Background: ARの適用を背景以外のピクセルに限定し、信号レベルの低い領域での過剰な処理を防ぎます。<br><br><a rel="noopener" target="_blank" href="https://github.com/riccardoalberghi/abberation_models/releases/tag/2.0.1">https://github.com/riccardoalberghi/abberation_models/releases/tag/2.0.1</a></p>
</blockquote>



<p class="wp-block-paragraph">作者は、 Riccardo Alberghiさん。 彼は、『天文学に情熱を注いでいます。深宇宙天体の撮像を専門としており、GraXpertやSirilの収差除去ツールなど、AIに基づくオープンソースの画像処理ツールの開発にも貢献しています。』</p>



<p class="wp-block-paragraph">Strenght: 強度。1のままで良いらいしです。</p>



<p class="wp-block-paragraph">Dark riging reduction: 恒星の周りに、黒い輪が見られる場合に使用します。</p>



<p class="wp-block-paragraph">Protect background: ピクセルをマスクで覆い隠し、保護された領域には補正を施しません。</p>



<h2 class="wp-block-heading">Deconvolution</h2>



<h3 class="wp-block-heading">GraXpert, Deconvolution (Objects), PSF Sizeの数値は？</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Siril,  Dynamic PSF<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/Dynamic-PSF.html">https://siril.readthedocs.io/en/latest/Dynamic-PSF.html</a></p>
</blockquote>



<p class="wp-block-paragraph">FWHM𝑥および FWHM𝑦：</p>



<div class="wp-block-math"><math display="block"><semantics><mrow><msub><mtext>FWHM</mtext><mi>x</mi></msub><mo>=</mo><mn>2</mn><msub><mi>σ</mi><mi>x</mi></msub><msqrt><mrow><mn>2</mn><mrow><mspace width="0.1667em"></mspace><mi>log</mi><mo>⁡</mo><mspace width="0.1667em"></mspace></mrow><mn>2</mn></mrow></msqrt></mrow><annotation encoding="application/x-tex">\text{FWHM}_x = 2\sigma_x\sqrt{2\log 2}</annotation></semantics></math></div>



<div class="wp-block-math"><math display="block"><semantics><mrow><msub><mtext>FWHM</mtext><mi>y</mi></msub><mo>=</mo><mn>2</mn><msub><mi>σ</mi><mi>y</mi></msub><msqrt><mrow><mn>2</mn><mrow><mspace width="0.1667em"></mspace><mi>log</mi><mo>⁡</mo><mspace width="0.1667em"></mspace></mrow><mn>2</mn></mrow></msqrt></mrow><annotation encoding="application/x-tex">\text{FWHM}_y = 2\sigma_y\sqrt{2\log 2}</annotation></semantics></math></div>



<p class="wp-block-paragraph">σx は、x方向の星像のGaussian（ガウス分布）の標準偏差<br>σy はy方向<br></p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">恒星独自のFWHM<br><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/photometry/quickphotometry.html">https://siril.readthedocs.io/en/latest/photometry/quickphotometry.html</a></p>
</blockquote>



<figure class="wp-block-image size-large"><img decoding="async" src="https://siril.readthedocs.io/ja/latest/_images/3D-StarPSF.png" alt=""/></figure>


<p>https://siril.readthedocs.io/ja/latest/_images/3D-StarPSF.png</p>


<p class="wp-block-paragraph">Siril Dynamic PSFの値を見てみます。例えば、</p>



<div class="wp-block-columns is-layout-flex wp-container-core-columns-is-layout-7387b849 wp-block-columns-is-layout-flex">
<div class="wp-block-column is-layout-flow wp-block-column-is-layout-flow">
<p class="wp-block-paragraph">Siril　Average Moffat PSF<br>N: 3 stars (0 saturated and excluded)<br>B: 0.000749<br>A: 0.003588<br>Beta: 3.38<br>FWHMx: 4.27&#8243;<br>FWHMy: 4.11&#8243;<br>r: 0.963<br>Angle: 80.89 deg<br>rmse: 3.413e-05</p>
</div>



<div class="wp-block-column is-layout-flow wp-block-column-is-layout-flow">
<p class="wp-block-paragraph">Moffat PSF<br>測定星数 =&gt;  3<br><br><br>Beta = &gt; 3.38<br>FWHMx =&gt;  4.27秒角<br>FWHMy = &gt; 4.11秒角<br>Roundness：丸み =&gt;  0.963<br>Angle = &gt; 80.89°<br></p>
</div>
</div>



<p class="wp-block-paragraph">4.27″ や 4.11″ は pixel ではなく arcsecです。角度の単位としての秒は、分の1/60の角度です。</p>



<p class="wp-block-paragraph">Graxpertは、PSF Sizeの単位は、ピクセルです。なので、単純にFWHMx:とFWHMy:の平均値を入れても単位が違うので意味を成しません。</p>



<p class="wp-block-paragraph">画像の1 pixelが、空の上で何秒角（arcsec）に相当するかを求める必要があります。</p>



<p class="wp-block-paragraph">FITSデータの、&#8221;CDELT1&#8243;, &#8220;CDELT2&#8243;に注目します。</p>



<p class="wp-block-paragraph">CDELT1: X軸（横方向、一般的には赤経 ：RA 側）の1ピクセルあたりの変化量<br>CDELT2: Y軸（縦方向、一般的には赤緯 ：Dec 側）の1ピクセルあたりの変化量<br>単位は 度（degree） が使われます。</p>



<p class="wp-block-paragraph">例えば、</p>



<pre class="wp-block-code"><code>CDELT1 = -0.000216049837916824
CDELT2 = 0.000216165447682517</code></pre>



<pre class="wp-block-code"><code>CDELT1: X方向：|−0.0002160498379|*3600 = 0.7777794"/pixel
CDELT2: Y方向：  0.0002161654477 *3600 = 0.7781956"/pixel</code></pre>



<pre class="wp-block-code"><code>X方向　=&gt; 0.7778"/pixel
Y方向  =&gt; 0.7782"/pixel</code></pre>



<p class="wp-block-paragraph">1 pixel は、約 0.778 arcsec です。</p>



<pre class="wp-block-code"><code>X方向：FWHMx=4.27" → 4.27"/0.7778"/pixel = 5.49 pixel →　FWHMx 約 5.49 pixel
Y方向：FWHMy=4.11" → 4.11"/0.7782"/pixel = 5.28 pixel →　FWHMy 約 5.28 pixel</code></pre>



<p class="wp-block-paragraph">ここでのPSFは、r: 0.963で円形に近いので、単純に平均します。<br>FWHM平均　→　(5.49 pixel + 5.28 pixel)/2 = 5.385 pixel</p>



<p class="wp-block-paragraph">GraXpert, Deconvolution (Objects), PSF Size に入れる値は、[5.4] になります。</p>



<p class="wp-block-paragraph">今まで単純にSiril Dynamic PSF,  Moffat PSF の値を入れていましたが、FWHMとPSFの関係が分かったようです。Moffat PSF値である 4.2、や、PSF平均値である 5.4を入れて Deconvolutionを行っても大きな差が見られない感じがしています。</p>



<p class="wp-block-paragraph">さらに深掘りするならば、<br>$ find /usr/local/share/GraXpert-linux -iname &#8216;deconv&#8217; -o -iname &#8216;onnx&#8217;<br>/usr/local/share/GraXpert-linux/lib/graxpert/deconvolution.pyc<br>このファイルを逆コンパイルして ↓</p>



<h4 class="wp-block-heading">decompyle3　を用いて逆コンパイル（は、行っていません）</h4>



<p class="wp-block-paragraph">decompyle3を使う仮想環境を作ります。</p>



<pre class="wp-block-code"><code>python3.12 -m venv /tmp/decompyle-env
source /tmp/decompyle-env/bin/activate
pip install decompyle3</code></pre>



<pre class="wp-block-code"><code>decompyle3 \
  /usr/local/share/GraXpert-linux/lib/graxpert/deconvolution.pyc \
  &gt; /tmp/deconvolution.py</code></pre>



<pre class="wp-block-code"><code># Unsupported bytecode in file /usr/local/share/GraXpert-linux/lib/graxpert/deconvolution.pyc
# Unsupported Python version, 3.11, for decompilation</code></pre>



<pre class="wp-block-code"><code>file /usr/local/share/GraXpert-linux/lib/graxpert/deconvolution.pyc
/usr/local/share/GraXpert-linux/lib/graxpert/deconvolution.pyc: Byte-compiled Python module for CPython 3.11</code></pre>



<p class="wp-block-paragraph">python 3.11が必要か&#8230;</p>



<h2 class="wp-block-heading">恒星除去</h2>



<h3 class="wp-block-heading">StarNet CLI tools</h3>



<h3 class="wp-block-heading">SyQon Starless</h3>



<p class="wp-block-paragraph">Stretch Options, Method: MTF Stretch </p>



<p class="wp-block-paragraph">M82の画像を、SyQon Starless, Zenith Model を用いて分離した。その際にMTF Stretchを使った結果が下画像です。</p>



<div class="wp-block-columns is-layout-flex wp-container-core-columns-is-layout-7387b849 wp-block-columns-is-layout-flex">
<div class="wp-block-column is-layout-flow wp-block-column-is-layout-flow">
<figure class="wp-block-image size-full"><img decoding="async" width="469" height="532" src="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.4-LogD-12.30_VeraLux_StarComposer.jpg" alt="" class="wp-image-6119" srcset="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.4-LogD-12.30_VeraLux_StarComposer.jpg 469w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.4-LogD-12.30_VeraLux_StarComposer-300x340.jpg 300w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.4-LogD-12.30_VeraLux_StarComposer-150x170.jpg 150w" sizes="(max-width: 469px) 100vw, 469px" /></figure>



<p class="wp-block-paragraph">MTF = 0.2<br>ガス塊や星雲の細かく淡い構造は保護された。</p>
</div>



<div class="wp-block-column is-layout-flow wp-block-column-is-layout-flow">
<figure class="wp-block-image size-full"><img loading="lazy" decoding="async" width="472" height="533" src="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.05-LogD-12.30_VeraLux_StarComposer.jpg" alt="" class="wp-image-6120" srcset="https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.05-LogD-12.30_VeraLux_StarComposer.jpg 472w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.05-LogD-12.30_VeraLux_StarComposer-300x339.jpg 300w, https://globe3.ddns.net/wp/wp-content/uploads/2026/09/M82_MTF0.05-LogD-12.30_VeraLux_StarComposer-150x169.jpg 150w" sizes="(max-width: 472px) 100vw, 472px" /></figure>



<p class="wp-block-paragraph">MTF = 0.05<br>淡い構造のコントラストが強調された。恒星状になってしまう。</p>
</div>
</div>



<h4 class="wp-block-heading">MTF Stretch / Median</h4>


<div class="hide-content">


<ul class="wp-block-list">
<li>MTF Stretch / Median<br>MTF = Midtone Transfer Function<br>Median Targetは、MTF Stretch後の画像の「中央値（Median）」をどの明るさにするかを指定します。<br><code>0.1 → 中央値を10%程度</code><br><code>0.2 → 中央値を20%程度</code><br><code>0.3 → 中央値を30%程度</code><br><br>例）MTF Stretch<br>　　　Linked stretch (preserve star colors) → ON<br>　　　Median Target → 0.1～0.2<br><br>Median = 0.050, 0.075<br>弱いストレッチ　：銀河の淡い構造をできるだけ守りたい<br>星も銀河も比較的暗い状態のままZenithに渡されます。<br><code>明るい星 → 認識できる</code><br><code>暗い星 → 認識されにくくなる可能性</code><br><code>淡い銀河の腕 → 保護されやすい</code><br><code>淡いハロー → 保護されやすい</code><br><code>星の除去 → やや控えめ</code><br><br>Median = 0.15 以上<br>強いストレッチ：星をできるだけ徹底的に除去したい<br>暗い部分までかなり持ち上げます。<br><code>明るい星 → よく認識</code><br><code>暗い星 → 認識しやすくなる</code><br><code>微小な星 → より検出される可能性</code><br><code>淡い銀河構造 → 星と誤認されるリスクが増える</code><br><code>星除去 → より積極的</code></li>



<li>Statistical Stretch：統計的ストレッチ<br>画像の統計情報（主に中央値や分布）を解析して、Zenithが星を認識しやすい明るさまで自動的にストレッチする方式</li>



<li>IHS</li>
</ul>



<h2 class="wp-block-heading">ノイズ除去</h2>



<h3 class="wp-block-heading">VeraLux Silentium</h3>



<ul class="wp-block-list">
<li>Noise intensity (Log S) : 強度</li>



<li>Detail guard</li>



<li>Adaptive noise model</li>



<li>Enabled chroma noise</li>



<li>Chroma strength</li>



<li>Deep space smoothness：輝度に関係なく、構造のぼやけを防ぎます。</li>



<li>Shadow smoothness</li>



<li>Starfield handling</li>



<li>Use findstar</li>



<li>Auto starless detection</li>
</ul>



<p class="wp-block-paragraph">Aggressive &#8216;Shadow Authority&#8217; for the deep background.<br>Features a Hard Cutoff: it automatically shuts down on signal (&gt;25%),preventing blur on structures regardless of intensity.</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">VeraLux — Silentium<br>Linear-Phase Noise Suppression Engine</p>



<p class="wp-block-paragraph">&#8212;&#8211; 使い方 &#8212;&#8211;<br>1. 前提条件：画像は（伸長処理を行う前に）必ずリニアでなければなりません。<br>2. 設定：<br>  &#8211; PSF保護機能を活用するために、「Use findstar」にチェックが入っていることを確認してください。<br>  -「Adaptive Noise Model」を選択してください（デフォルトでオン）。<br>3. 輝度キャリブレーション：<br>  &#8211; 「Noise Intensity」：一般的な粒状感低減のための全体的なしきい値を設定します。<br>  &#8211; 「Detail Guard」：微弱な構造を保護します。この値を上げると、背景ノイズを再導入することなく、微細なディテールのマイクロコントラストを復元します。<br> -「Shadow Smoothness」：背景に対して、対象を特定した強力なノイズ除去を行います。<br>4. プレビュー操作：<br> &#8211; 全体表示：クリックして領域を1:1で確認します。<br> &#8211; ルーペモード（1:1）：ドラッグしてパンします。クリック（ドラッグせずに）すると全体表示に戻ります。<br>5. 処理：[PROCESS] をクリックして、フル解像度の画像に適用します。</p>



<p class="wp-block-paragraph">概要</p>



<p class="wp-block-paragraph">• 理論的根拠：定常ウェーブレット変換（SWT）及び、ドブシーウェーブレット(Daubechies Wavelets)<br>• ロジック：「シャドウ・オーソリティ」適応型マスキングおよび相対ペデスタル検出<br>• 天体写真の信号対雑音比の最適化（リニアステージ）</p>



<p class="wp-block-paragraph">Silentiumは、恒星の形状や星雲の微細な構造を損なうことなく、物理信号から確率的ノイズを分離するように設計された線形位相ノイズ抑制エンジンです。</p>



<p class="wp-block-paragraph">Silentiumは、低信号領域（背景）の処理と高信号構造の処理を分離することで、線形処理における新たなパラダイムを確立します。「シャドウ・オーソリティ」ロジックと「フィジックス・アウェア」アーキテクチャを通じて、深宇宙の背景ノイズに対して数学的に強力なクリーニングを適用しつつ、星や星雲の周囲には透過しない保護シェルを形成し、データの本来の測光的完全性を維持します。</p>



<ul class="wp-block-list">
<li>定常ウェーブレット変換（SWT）：<br>Daubechies（db2）ウェーブレットを用いた非デシメーション型マルチスケール分解を採用しています。このシフト不変の手法により、位相シフトゼロかつエイリアシングのない正確な周波数分離が保証され、標準的なDWTと比較して優れたテクスチャ品質を実現します。</li>



<li>測光信号ゲーティング：<br>本エンジンは、線形データ解析（中央値＋シグマ閾値）から導出された統計的な「信号確率マップ」を利用しています。これにより、画像を厳密に異なる処理領域に分離し、背景の平滑化と構造のシャープ化の間の相互干渉を防止します。</li>



<li>シャドウ・オーソリティおよびエクスクルージョン・ゲート：<br>背景除去のために厳格な「ウィナー・テイク・オール」ロジックを実装しています。シャドウ スムースネス・エンジンは、信号構造が検出されると直ちに<br>ノイズ除去の強度をゼロに強制するハードな測光カットオフ（エクスクルージョン・ゲート）を備えています。<br>これにより、保護設定にかかわらず、ハイライト部分の「ぼやけ」を防ぎます。</li>



<li>PSF補正付き構造ガード：<br>ディテールガードは、線形信号確率と形態学的拡張、およびシーイング補正を組み合わせています。局所的なFWHMに応じて構造保護を線形にスケーリングし、大気によるぼやけで光学的に柔らかく見えるフィラメントの侵食を防ぎ、ノイズ除去の圧力をバランスさせます。</li>



<li>ルーペUX：<br>グローバルビューは静的（表示のみ）です。クリックすると1:1の「ルーペモード」にズームインし、ピクセル単位での詳細確認が可能になります。ドラッグするとルーペをパンできます。</li>



<li>シャドウレポート：<br>処理完了時に、ノイズ低減、SNRの向上、およびペデスタルの保持状況に関する統計的な内訳を提供します。</li>
</ul>



<p class="wp-block-paragraph">(c) 2025 Riccardo Paterniti<br>SPDX-License-Identifier: GPL-3.0-or-later<br><a rel="noopener" target="_blank" href="https://gitlab.com/free-astro/siril-scripts/-/blob/main/VeraLux/VeraLux_Silentium.py">https://gitlab.com/free-astro/siril-scripts/-/blob/main/VeraLux/VeraLux_Silentium.py</a></p>
</blockquote>



<p class="wp-block-paragraph">VeraLux_Silentium.pyにある説明とコードから、ノイズ除去を追ってみます。</p>



<p class="wp-block-paragraph">ノイズ除去を行う前工程で、ウェーブレットを行います。<br>ウェーブレット変換では、画像を「大きな構造 ←　→ 小さな構造」に分解できます。</p>



<p class="wp-block-paragraph">天体画像<br>大きな構造 : 星雲の大まかな濃淡 ← 中くらいの構造 : 星雲の細かな模様 → 小さな構造 : 星,細部,ノイズ</p>



<p class="wp-block-paragraph">天体画像では、構造はピクセルの集まりと表現できます。<br>天体画像での構造を、ピクセルの集まりとし、その大きさをスケールと表現すると、</p>



<ul class="wp-block-list">
<li>大きいスケール：  数十～数百ピクセル以上の変化 → 星雲の大まかな形、背景の勾配など</li>



<li>中間スケール： 数～数十ピクセル程度の構造 → 星雲の細かなフィラメントなど</li>



<li>小さいスケール → 1～数ピクセル程度の細かな変化 → ノイズ、細かい星、微細なディテール</li>
</ul>



<p class="wp-block-paragraph">天体画像での、ノイズ除去は以下の順序で行われます。</p>



<ol class="wp-block-list">
<li>ドブシーウェーブレットを用いて画像を異なるスケールの成分（ウェーブレット係数）に分解</li>



<li>各スケールの係数を分析して、ノイズと判断される成分を弱める</li>



<li>逆変換によって画像を再構成することでノイズ除去を行う。</li>
</ol>



<p class="wp-block-paragraph">597行目辺りの</p>



<pre class="wp-block-code"><code>        """
        SWT implementation (Stationary Wavelet Transform).
        Shadow Exclusion Gate (Hard Cutoff at 25% Signal)
        """
        max_levels = 4
        wavelet = "db2"

        # Pad image for SWT requirements
        padded_channel, (ph, _, pw, _) = SilentiumCore._pad_for_swt(channel, max_levels)

        # Decompose using Undecimated Wavelet Transform (SWT)
        coeffs = pywt.swt2(padded_channel, wavelet=wavelet, level=max_levels)</code></pre>



<p class="wp-block-paragraph">ノイズとして扱ってよい領域を先に限定し、その領域だけWavelet係数を強く削る</p>



<pre class="wp-block-code"><code>                 ┌─ Noise Map  ──────┐
                 │                          │
Linear Image ──┼─ Signal Map ──────┤
                 │                          │
                 ├─ Edge Map ───────┤
                 │                          │
                 └─ PSF/FWHM Map ─────┘
                            │
                            ▼
                    Adaptive Threshold
                            │
                            ▼
                 SWT / db2 / 4 levels
                            │
                            ▼
                    Soft Thresholding
                            │
                            ▼</code></pre>



<h2 class="wp-block-heading">再合成</h2>



<h3 class="wp-block-heading">VeraLux Star Composer</h3>



<p class="wp-block-paragraph"><a rel="noopener" target="_blank" href="https://www.youtube.com/watch?v=Pen9aOy2NmA">https://www.youtube.com/watch?v=Pen9aOy2NmA</a><br>Rich Stevensonさんが解説しています。</p>



<ul class="wp-block-list">
<li>1.Input: どちらかを好みで使っても良いようです。
<ul class="wp-block-list">
<li>Linear Add<br>Linear Add（線形加算）を使用するとコントラストが最も高くなりますが、背景が非常に明るい場合、ハイライトが飽和するかもしれません。</li>



<li>Screen (Safe)<br>Screen Safeは、中心部がかなり明るい銀河に適用するのが最適で、その明るさを抑えるのに役立ちますが、その一方でコントラストも低下してしまいます。</li>
</ul>
</li>



<li>2.Sensor Profile</li>



<li>3.VeraLux Stretch
<ul class="wp-block-list">
<li>Star Intensity (Log D)</li>



<li>Profile Hardness (b)</li>



<li>Adaptive Anchor</li>
</ul>
</li>



<li>4.Physics (Hybrid)
<ul class="wp-block-list">
<li>Color Grip (Blend)<br>Physics<br>Color Grip (Hybrid Blend):<br>Controls blending between per-channel (Scalar) and luminance-driven (Vector) mapping.<br>0% (Scalar): Per-channel tone mapping. Stronger white cores, higher perceived sharpness.<br>100% (Vector): Luminance-based mapping with RGB ratio preservation.<br>More chromatic fidelity, slightly softer cores.<br>カラーグリップ（ハイブリッドブレンド）：<br>チャンネルごとの（スカラー）マッピングと、輝度に基づく（ベクトル）マッピングのブレンド比率を調整します。<br>0%（スカラー）：チャンネルごとのトーンマッピング。白のコアが強調され、シャープさがより強く感じられます。<br>100%（ベクトル）：RGBの比率を維持する、輝度ベースのマッピング。<br>色再現性が高く、白のコアがやや柔らかくなります。</li>



<li>Shadow Conv (Hide Aetifacts)</li>
</ul>
</li>



<li>5.Star Surgery
<ul class="wp-block-list">
<li>Core Rejection(LSR)</li>



<li>Reduction(Erosin)</li>



<li>Optical Healing(Halos)</li>
</ul>
</li>
</ul>


</div>]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/siril-image-processing-workflow-2026-06/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>&#8220;No matching distribution found for&#8221;</title>
		<link>https://globe3.ddns.net/wp/no-matching-distribution-found-for/</link>
					<comments>https://globe3.ddns.net/wp/no-matching-distribution-found-for/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Tue, 25 Aug 2026 07:48:55 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=5086</guid>

					<description><![CDATA[Sirilの仮想環境で、PyTorchや関連パッケージのインストール時にエラーが発生する。 SyQon Starless スクリプトを起動させると以下のエラー この春から夏は曇って、星空が見られないので、SirilやGU [&#8230;]]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">Sirilの仮想環境で、PyTorchや関連パッケージのインストール時にエラーが発生する。</p>



<p class="wp-block-paragraph">SyQon Starless スクリプトを起動させると以下のエラー</p>



<pre class="wp-block-code"><code>: PyQt6 is installed
: PySide6 is installed
: astropy is installed
: scipy is installed
: Installing PyTorch with backend: intel
: Installing: torch torchvision intel-extension-for-pytorch
: Looking in indexes: https://download.pytorch.org/whl/xpu
: Collecting torch
:   Using cached torch-2.13.0%2Bxpu-cp312-cp312-win_amd64.whl.metadata (40 kB)
: Collecting torchvision

:   Using cached torchvision-0.28.0%2Bxpu-cp312-cp312-win_amd64.whl.metadata (5.7 kB)
: ERROR: Could not find a version that satisfies the requirement intel-extension-for-pytorch (from versions: none)
: ERROR: No matching distribution found for intel-extension-for-pytorch
: Failed to install PyTorch: Command '&#91;'C:\\Users\\USERNAME\\AppData\\Local\\siril\\venv\\Scripts\\python.exe', '-m', 'pip', 'install', 'torch', 'torchvision', 'intel-extension-for-pytorch', '--index-url', 'https://download.pytorch.org/whl/xpu']' returned non-zero exit status 1.

: Error installing Torch: Command '&#91;'C:\\Users\\USERNAME\\AppData\\Local\\siril\\venv\\Scripts\\python.exe', '-m', 'pip', 'install', 'torch', 'torchvision', 'intel-extension-for-pytorch', '--index-url', 'https://download.pytorch.org/whl/xpu']' returned non-zero exit status 1.

: Traceback (most recent call last):
:   File "C:\Users\USENAME\AppData\Local\siril-scripts\SyQon/Starless.py", line 108, in &lt;module&gt;
:     import torch
: ModuleNotFoundError: No module named 'torch'
: Python process (PID: 10120) exited with status 1
: Python communication worker finished...</code></pre>



<p class="wp-block-paragraph">この春から夏は曇って、星空が見られないので、SirilやGUN/Linuxばかり弄ってます。</p>



<div style="height:22px" aria-hidden="true" class="wp-block-spacer"></div>



<p class="wp-block-paragraph">軽量版のCPU専用（whl/cpu）を入れ、Disable GPUで行う方法</p>



<p class="wp-block-paragraph">Sirilの仮想環境（venv）を有効化<br>cmd</p>



<pre class="wp-block-code"><code>C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\activate.bat</code></pre>



<p class="wp-block-paragraph">pip自体をアップデート</p>



<pre class="wp-block-code"><code>python -m pip install --upgrade pip</code></pre>



<p class="wp-block-paragraph">CPU専用PyTorchをインストール<br>エラーの原因だったIntel拡張パッケージを無視し、余計なGPU用ライブラリを除外したCPU版を上書きします。</p>



<p class="wp-block-paragraph">&#8211;dry-run オプションでインストールした際の振る舞いを確認します。</p>



<pre class="wp-block-code"><code>python -m pip install torch torchvision --index-url https://download.pytorch.org/whl --dry-run</code></pre>



<pre class="wp-block-code"><code>python -m pip install torch torchvision --index-url https://pytorch.org --dry-run
Looking in indexes: https://pytorch.org
ERROR: Could not find a version that satisfies the requirement torch (from versions: none)
ERROR: No matching distribution found for torch</code></pre>



<pre class="wp-block-code"><code>python -m pip install torch torchvision --index-url <a rel="noopener" target="_blank" href="https://download.pytorch.org/whl/cpu">https://download.pytorch.org/whl/cpu</a> --dry-run</code></pre>



<pre class="wp-block-code"><code>Would install Jinja2-3.1.6 MarkupSafe-3.0.3 filelock-3.32.3 fsspec-2026.7.0 mpmath-1.3.0 networkx-3.6.1 pillow-12.3.0 setuptools-78.1.0 sympy-1.14.0 torch-2.13.0+cpu torchvision-0.28.0+cpu</code></pre>



<pre class="wp-block-code"><code>Successfully installed MarkupSafe-3.0.3 filelock-3.32.3 fsspec-2026.7.0 jinja2-3.1.6 mpmath-1.3.0 networkx-3.6.1 pillow-12.3.0 setuptools-78.1.0 sympy-1.14.0 torch-2.13.0+cu132 torchvision-0.28.0+xpu ###← xpuが入っちゃった場合は削除</code></pre>



<pre class="wp-block-code"><code>### torchvision-0.28.0+xpu ← xpuが入っちゃった場合は削除
python -m pip uninstall torch torchvision</code></pre>



<pre class="wp-block-code"><code>python -m pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu --force-reinstall</code></pre>



<pre class="wp-block-code"><code>Successfully installed MarkupSafe-3.0.3 filelock-3.32.3 fsspec-2026.7.0 jinja2-3.1.6 mpmath-1.3.0 networkx-3.6.1 numpy-2.5.2 pillow-12.3.0 setuptools-78.1.0 sympy-1.14.0 torch-2.13.0+cpu torchvision-0.28.0+cpu typing-extensions-4.16.0</code></pre>



<p class="wp-block-paragraph">動作確認テスト</p>



<pre class="wp-block-code"><code>python -c "import torch; print('PyTorch Version:', torch.__version__)</code></pre>



<p class="wp-block-paragraph">仮想環境を終了<br>Successfully installed&#8230;が出たら、環境を終えます<br>deactivate</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<div style="height:13px" aria-hidden="true" class="wp-block-spacer"></div>



<p class="wp-block-paragraph">Intel Iris Xe環境<br>xpuを入れるらしいが、試していない。</p>



<p class="wp-block-paragraph">Sirilの仮想環境（venv）を有効化（アクティベート）<br>cmd<br>C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\activate.bat</p>



<p class="wp-block-paragraph">pip自体をアップデートし、古い依存関係が残らないよう、念のため仮想環境内のpipを最新にします。<br>cmd<br>python -m pip install &#8211;upgrade pip</p>



<p class="wp-block-paragraph">エラー原因を省いたPyTorchのみの強制再インストール問題を引き起こしていた intel-extension-for-pytorch の指定を完全に省き、ログで正常に掴めていた Intel XPU 対応の PyTorch 本体のみを上書き配置します。<br>cmd<br>python -m pip install torch torchvision &#8211;index-url <a rel="noopener" target="_blank" href="https://download.pytorch.org/whl/xpu">https://download.pytorch.org/whl/xpu</a> &#8211;dry-run<br><br>python -m pip install torch torchvision &#8211;index-url <a rel="noopener" target="_blank" href="https://download.pytorch.org/whl/xpu">https://download.pytorch.org/whl/xpu</a> &#8211;force-reinstall</p>



<p class="wp-block-paragraph">動作確認テスト<br>cmd<br>python -c &#8220;import torch; print(&#8216;PyTorchバージョン:&#8217;, torch.<strong>version</strong>); print(&#8216;XPU利用可能か:&#8217;, torch.xpu.is_available() if hasattr(torch, &#8216;xpu&#8217;) else &#8216;非対応&#8217;)&#8221;</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/no-matching-distribution-found-for/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Siril 1.5 ビルドエラー / Invalid character escape &#8216;\m&#8217;.</title>
		<link>https://globe3.ddns.net/wp/siril-build-error-character-escape-m/</link>
					<comments>https://globe3.ddns.net/wp/siril-build-error-character-escape-m/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Sun, 23 Aug 2026 08:28:55 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=4973</guid>

					<description><![CDATA[Siril 1.5 をソースからビルドするとエラー表示 二回、表示されてる、文字エスケープが以下です。 Windowsのファイルパスである C:\msys64&#8230; の \m という部分（バックスラッシュ＋m）を [&#8230;]]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">Siril 1.5 をソースからビルドするとエラー表示</p>



<pre class="wp-block-code"><code>| -- Could NOT find NSIS (missing: NSIS_MAKE)
| -- Found PkgConfig: C:/msys64/ucrt64/bin/pkg-config.exe (found version "3.0.5")
| CMake Warning (policy) at CMakeLists.txt:74 (add_custom_target):
| Policy CMP0219 is not set: Macro invocations preserve backslashes in
| arguments.  Run "cmake --help-policy CMP0219" for policy details.  Use the
| cmake_policy command to set the policy and suppress this warning.

| Command "add_custom_target" called with arguments containing backslashes.

| Since the policy is not set, backslashes in the arguments:

| "C:\msys64\ucrt64\bin/ninja.exe"

| will be interpreted as escape sequences for compatibility.

| Set the policy to NEW to instead pass

| "C:\\msys64\\ucrt64\\bin/ninja.exe"

| so that argument parsing will preserve the original values.
| This warning is for project developers.  Use -Wno-author or -Wno-policy to
| suppress it.

| CMake Error at C:/msys64/ucrt64/lib/python3.14/site-packages/mesonbuild/cmake/data/preload.cmake:39 (_add_custom_target):
| Syntax error in cmake code when parsing string

| dist;COMMAND;C:\msys64\ucrt64\bin/ninja.exe;package_source

| Invalid character escape '\m'.
| Call Stack (most recent call first):
| CMakeLists.txt:74 (add_custom_target)


| CMake Error at C:/msys64/ucrt64/lib/python3.14/site-packages/mesonbuild/cmake/data/preload.cmake:39 (_add_custom_target):
| Syntax error in cmake code at

| C:/msys64/ucrt64/lib/python3.14/site-packages/mesonbuild/cmake/data/preload.cmake:39

| when parsing string

| dist;COMMAND;C:\msys64\ucrt64\bin/ninja.exe;package_source

| Invalid character escape '\m'.
| Call Stack (most recent call first):
| CMakeLists.txt:74 (add_custom_target)


| -- Configuring incomplete, errors occurred!

librtprocess| CMake configuration: FAILED

meson.build:343:19: ERROR: Failed to configure the CMake subproject
</code></pre>



<p class="wp-block-paragraph">二回、表示されてる、文字エスケープが以下です。</p>



<pre class="wp-block-code"><code>Invalid character escape '\m'.</code></pre>



<p class="wp-block-paragraph">Windowsのファイルパスである C:\msys64&#8230; の \m という部分（バックスラッシュ＋m）を、mesonプログラムが「改行（\n）やタブ（\t）のような特殊な文字コード（エスケープシーケンス）」だと勘違いし、「そんな特殊文字（\m）は存在しない」とエラーを起こしているらしい。</p>



<p class="wp-block-paragraph">どこで \m が発生したかというと、</p>



<pre class="wp-block-code"><code>| dist;COMMAND;C:\msys64\ucrt64\bin/ninja.exe;package_source</code></pre>



<p class="wp-block-paragraph">ビルドツールである ninja.exe の場所を指定する際、Windows形式のパス（C:\msys64&#8230;）がそのまま入ってしまっている模様です。</p>



<p class="wp-block-paragraph">どの箇所でエラーが起きているのかな？</p>



<pre class="wp-block-code"><code>librtprocess| CMake configuration: FAILED

meson.build:343:19: ERROR: Failed to configure the CMake subproject</code></pre>



<p class="wp-block-paragraph">librtprocess を CMakeを使って組み立てる時に、バックスラッシュを特殊文字と勘違いした様です。</p>



<p class="wp-block-paragraph">Mesonツールが、CMakeツールにバックスラッシュ（\）を含んだWindowsのパスをそのまま渡したので、CMake側がそれを特殊文字（エスケープシーケンス）と勘違いして「そんな特殊文字は知らん！」と止まったようです。</p>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph"><code>\n</code> → 改行<br><code>\t</code> → タブ<br><code>\m</code> → <strong>「そんな特殊文字は知らない！」</strong></p>
</div>



<p class="wp-block-paragraph">C:\msys64　この箇所の、『&#8221;/m&#8221;が解らないよ』とCMakeが言ってます。</p>



<pre class="wp-block-code"><code>dist;COMMAND;C:\msys64\ucrt64\bin/ninja.exe;package_source</code></pre>



<pre class="wp-block-code"><code>| "C:\msys64\ucrt64\bin/ninja.exe"</code></pre>



<p class="wp-block-paragraph">警告箇所があります。</p>



<pre class="wp-block-code"><code> Set the policy to NEW to instead pass

| "C:\\msys64\\ucrt64\\bin/ninja.exe"

| so that argument parsing will preserve the original values.
| This warning is for project developers.  Use -Wno-author or -Wno-policy to
| suppress it.</code></pre>



<p class="wp-block-paragraph">「古いルール（OLD）のまま処理しちゃうから、パスが壊れて \m になっちゃうよ！新しいルール（NEW）に変えたほうがいいよ！」と警告した直後に、パスが壊れてエラーになり、ビルドが停止した様です。</p>



<p class="wp-block-paragraph">バックスラッシュではなくて、スラッシュを使えば良いので、GUN/Linux風にコマンドを打ちます。<br>GUN/Linux風に、スラッシュを用いてninja コマンドを環境変数としてmesonに渡します。</p>



<pre class="wp-block-code"><code>meson setup _build --buildtype release -Dgtk=true　ではなくて、</code></pre>



<pre class="wp-block-code"><code>NINJA=/msys64/ucrt64/bin/ninja.exe meson setup _build --buildtype release -Dgtk=true</code></pre>



<p class="wp-block-paragraph">これでビルドが出来ます。</p>



<p class="wp-block-paragraph">次もありました。</p>



<pre class="wp-block-code"><code>NINJA=c:/local/msys64/ucrt64/bin/ninja.exe  meson setup _build --reconfigure</code></pre>



<p class="wp-block-paragraph"></p>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/siril-build-error-character-escape-m/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Cosmic Clarity Suite, Cosmic Clarity Sharpening &#038; Cosmic Clarity denoise, CosmicClaritySuite_Linux_AlternateBuild, Ubuntu 26.04 + Siril 1.5</title>
		<link>https://globe3.ddns.net/wp/cosmic-clarity-sharpening-denoise-cosmicclaritysuite_linux_alternatebuild-ubuntu2604-siril-15/</link>
					<comments>https://globe3.ddns.net/wp/cosmic-clarity-sharpening-denoise-cosmicclaritysuite_linux_alternatebuild-ubuntu2604-siril-15/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Mon, 10 Aug 2026 05:16:14 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<category><![CDATA[Ubuntu]]></category>
		<category><![CDATA[画像処理]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=4623</guid>

					<description><![CDATA[Siril 1.5 Pythonスクリプトの一つである、Cosmic Clarity Sharpeningを、Ubuntu 26.04 + RTX5060 GPUで使います。 &#8220;SetiAstroCosmic [&#8230;]]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">Siril 1.5 Pythonスクリプトの一つである、Cosmic Clarity Sharpeningを、<br>Ubuntu 26.04 + RTX5060 GPUで使います。</p>



<h2 class="wp-block-heading">&#8220;SetiAstroCosmicClarity&#8221;, &#8220;SetiAstroCosmicClarity_denoise&#8221;エラー</h2>



<p class="wp-block-paragraph">CosmicClaritySuite_Linux_AlternateBuild にある、&#8221;SetiAstroCosmicClarity&#8221;、&#8221;SetiAstroCosmicClarity_denoise&#8221;は単体ではエラーが表示されます。</p>



<p class="wp-block-paragraph">エラー表示</p>



<pre class="wp-block-code"><code>PyInstaller/loader/pyimod02_importers.py:450: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools&lt;81.

/usr/lib/x86_64-linux-gnu/gvfs/libgvfscommon.so: undefined symbol: g_variant_builder_init_static
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgvfsdbus.so
/usr/lib/x86_64-linux-gnu/gvfs/libgvfscommon.so: undefined symbol: g_variant_builder_init_static
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgioremote-volume-monitor.so
Using device: cuda (CUDA)
Loaded TIFF image with dtype: float32
Final bit depth set to: 32-bit floating point
Sharpening Image: result_drizzle_19620s.fit.tif
Total Chunks: 225
Stellar Sharpening:
Traceback (most recent call last):
  File "setiastrocosmicclaritylinux.py", line 1374, in &lt;module&gt;
  File "setiastrocosmicclaritylinux.py", line 1199, in process_images
  File "setiastrocosmicclaritylinux.py", line 921, in sharpen_image
RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

&#91;PYI-7815:ERROR] Failed to execute script 'setiastrocosmicclaritylinux' due to unhandled exception!</code></pre>



<p class="wp-block-paragraph">RuntimeError: CUDA error: no kernel image is available for execution on the device</p>



<ul class="wp-block-list">
<li>RuntimeError: CUDA エラー: デバイス上で実行可能なカーネルイメージがありません。</li>



<li><code>SetiAstroCosmicClarity</code>にパッケージされているPyTorchのバイナリが、システム環境（Ubuntu 26.04 + CUDA 12.8）のGPUを扱えない。</li>
</ul>



<p class="wp-block-paragraph">UserWarning: pkg_resources is deprecated as an API</p>



<ul class="wp-block-list">
<li>pkg_resources は API として非推奨となっています。</li>



<li>Pythonの古い内部ツールが使われているという警告です。</li>
</ul>



<p class="wp-block-paragraph">SetiAstroCosmicClarity&#8221;, &#8220;SetiAstroCosmicClarity_denoise&#8221; を単体で動かし、動作確認を行います。<br>その後に Siril 1.5 Pythonスクリプトから動かします。</p>



<p class="wp-block-paragraph">SetiAstroCosmicClarity&#8221;, &#8220;SetiAstroCosmicClarity_denoise&#8221; それぞれをpyinstxtractor.py を使って展開し動作確認、後に、ラッパーを作成し、Siril 1.5から動かします。</p>



<h2 class="wp-block-heading">pyinstxtractor.py</h2>



<p class="wp-block-paragraph">pyinstxtractor.py を使って、SetiAstroCosmicClarity を展開します。</p>



<h3 class="wp-block-heading">Cosmic Clarity Sharpening を展開</h3>



<pre class="wp-block-code"><code># Siril 1.5 仮想環境で行う
pwd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild

python3.12 pyinstxtractor.py ./SetiAstroCosmicClarity

&#91;+] Processing ./SetiAstroCosmicClarity
&#91;+] Pyinstaller version: 2.1+
&#91;+] Python version: 3.12
&#91;+] Length of package: 57264312 bytes
&#91;+] Found 17 files in CArchive
&#91;+] Beginning extraction...please standby
&#91;+] Possible entry point: pyiboot01_bootstrap.pyc
&#91;+] Possible entry point: pyi_rth_inspect.pyc
&#91;+] Possible entry point: pyi_rth_pkgutil.pyc
&#91;+] Possible entry point: pyi_rth_multiprocessing.pyc
&#91;+] Possible entry point: pyi_rth_pkgres.pyc
&#91;+] Possible entry point: pyi_rth_setuptools.pyc
&#91;+] Possible entry point: pyi_rth_mplconfig.pyc
&#91;+] Possible entry point: pyi_rth_cryptography_openssl.pyc
&#91;+] Possible entry point: pyi_rth_pyqt6.pyc
&#91;+] Possible entry point: pyi_rth__tkinter.pyc
&#91;+] Possible entry point: setiastrocosmicclaritylinux.pyc
&#91;+] Found 7078 files in PYZ archive
&#91;+] Successfully extracted pyinstaller archive: ./SetiAstroCosmicClarity

You can now use a python decompiler on the pyc files within the extracted directory</code></pre>



<p class="wp-block-paragraph">SetiAstroCosmicClarity_extracted ディレクトリにある、<br>setiastrocosmicclaritylinux.pyc を<br>CosmicClaritySuite_Linux_AlternateBuild ディレクトリに配置</p>



<pre class="wp-block-code"><code># 動作確認
python3.12 ./setiastrocosmicclaritylinux.pyc

Using device: cuda (CUDA)
Identified 32bit floating point FITS iamge.
Sharpening Image: result_drizzle_19620s.fit
Total Chunks: 225
Stellar Sharpening:
Progress: 100.00% (225/225 chunks processed)
Stellar sharpening complete. Shape: (2877, 2877)
Saved 32-bit float sharpened image to: /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/./output/result_drizzle_19620s_sharpened.fit</code></pre>



<h4 class="wp-block-heading">Cosmic Clarity Sharpening ラッパーを作成</h4>



<pre class="wp-block-code"><code>### ラッパーファイルを作成 run_Sharpen.sh
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/</code></pre>



<p class="wp-block-paragraph">以下が、run_Sharpen.sh</p>



<pre class="wp-block-code"><code>#!/bin/bash
source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/
python3.12 ./setiastrocosmicclaritylinux.pyc "$@"

# 実行権
chmod +x ./run_Sharpen.sh</code></pre>



<p class="wp-block-paragraph"><code>"$@"</code>　について<br></p>



<h4 class="wp-block-heading">Cosmic Clarity Sharpening を設定</h4>



<p class="wp-block-paragraph">Cosmic Clarity Sharpening Settings<br>Cosmic Clarity Executable:  run_Sharpen.sh (上記ラッパー)を指定する。</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h4 class="wp-block-heading">RTX 5060 動作確認</h4>



<pre class="wp-block-code"><code>source /home/$USER/.local/share/siril/venv/bin/activate
pip list |grep torch
torch                    2.11.0+cu128
torchaudio               2.11.0+cu128
torchvision              0.26.0+cu128

pip list |grep torch-cuda80
pip list |grep torch-model-archiver
pip list |grep torchrec-cpu</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 -c "import torch; print(torch.__version__, torch.cuda.is_available(), torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"
2.11.0+cu128 True NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 -c "import torch; print('Version:', torch.__version__); print('CUDA:', torch.cuda.is_available()); print('Device Name:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"
Version: 2.11.0+cu128
CUDA: True
Device Name: NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<h3 class="wp-block-heading">Cosmic Clarity denoise を展開</h3>



<p class="wp-block-paragraph">python3.12 pyinstxtractor.py /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/SetiAstroCosmicClarity_denoise</p>



<pre class="wp-block-code"><code># Siril 1.5 仮想環境で行う
pwd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild
python3.12 pyinstxtractor.py ./SetiAstroCosmicClarity_denoise</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 ./setiastrocosmicclaritylinux.pyc
</code></pre>



<h4 class="wp-block-heading">Cosmic Clarity denoise ラッパーを作成</h4>



<pre class="wp-block-code"><code>### ラッパーファイルを作成 run_Denoise.sh
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/</code></pre>



<p class="wp-block-paragraph">以下が、run_Denoise.sh</p>



<pre class="wp-block-code"><code>#!/bin/bash

### RTX 5060（12.0）
### export TORCH_CUDA_ARCH_LIST="9.0"
### export TORCH_CUDA_ARCH_LIST="12.0"

source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/
python3.12 ./SetiAstroCosmicClarity_denoise.pyc "$@"</code></pre>



<h4 class="wp-block-heading">Cosmic Clarity Denoise を設定</h4>



<p class="wp-block-paragraph">Cosmic Clarity Executable<br>run_Sharpen.sh (上記ラッパー)を指定する。</p>



<p class="wp-block-paragraph">🥂</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/cosmic-clarity-sharpening-denoise-cosmicclaritysuite_linux_alternatebuild-ubuntu2604-siril-15/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Siril 1.5 + starnet2_linux_2.5.4-0214_ORT_x64_cli + Ubuntu 26.04 + GPU</title>
		<link>https://globe3.ddns.net/wp/siril-15-starnet2_linux_2-5-4-0214_ort_x64_cli_ubuntu2604/</link>
					<comments>https://globe3.ddns.net/wp/siril-15-starnet2_linux_2-5-4-0214_ort_x64_cli_ubuntu2604/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Thu, 06 Aug 2026 06:01:40 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<category><![CDATA[Ubuntu]]></category>
		<category><![CDATA[画像処理]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=4534</guid>

					<description><![CDATA[onnxruntime-linux-x64-gpu-1.26.0に含まれる、libonnxruntime.so.1.26.0を用いるとStarnet2-2.5.4がGPU (RTX5060) を使えます。 CUDA To [&#8230;]]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">onnxruntime-linux-x64-gpu-1.26.0に含まれる、libonnxruntime.so.1.26.0を用いると<br>Starnet2-2.5.4がGPU (RTX5060) を使えます。</p>



<p class="wp-block-paragraph">CUDA Toolkit 12.8.2 (April 2026)runfile + Ubuntu 26.04 + nvidia-driver 580 + python version 3.12.13 + PyTorch は、あらかじめインストールし動作するようにして下さい。</p>



<p class="wp-block-paragraph">例えば、<br><a target="_self" href="https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-cosmic-clarity-suite-siril-15/">https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-cosmic-clarity-suite-siril-15/</a></p>



<p class="wp-block-paragraph">Open Neural Network Exchange (ONNX)<br>推論・学習高速化を行う実行エンジンを、GPU適応に置き換えて、GPU（ここではRTX5060）が動作するように設定します。</p>



<p class="wp-block-paragraph">Windows 11 OS環境での、<br>Siril 1.5 + Windows 11 + starnet2_win_2.5.4-0214_ORT_x64_cli + RTX5060については、後述しています。</p>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">アップデートがありました。<br>2026-09-15　StarNet2 2.6.2 and DeepSNR 1.3.2 command-line release<br><br>starnet2_linux_2.5.4-0214_ORT_x64_cli を、<br>starnet2_linux_2.6.2-0241_ORT_x64_cli へアップデートできます。</p>
</div>



<p class="wp-block-paragraph">starnet2_linux_2.6.2-0241 では以前に感じていた色調の違和感が無くなりました。有り難い。</p>



<h2 class="wp-block-heading">StarNet.py</h2>



<h3 class="wp-block-heading">Siril へ StarNet.pyを導入する</h3>



<p class="wp-block-paragraph">[Set paths] で &#8220;starnet2&#8221; 及び &#8220;StarNet2_weights.onnx &#8220;を指定<br>[Run StarNet] → CPUが動作する。</p>



<pre class="wp-block-code"><code>Linear mode: applied MTF autostretch to the StarNet input.
Reading input from /tmp/siril_starnet_8mjx6poe/input.tif...
Color image is detected!
Image size: 3996x2721
Restoring neural network checkpoint...
Loading ONNX model from file: /usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/StarNet2_weights.onnx
Backend: ONNX Runtime (CPU execution provider) # ← CPUが動いている</code></pre>



<p class="wp-block-paragraph">GPUが動作するように、設定していきます。</p>



<h2 class="wp-block-heading">weights.onnx, libonnxruntime</h2>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">こちらを参考にします。<br>Method 1: System-wide NVIDIA installer setup<br><a rel="noopener" target="_blank" href="https://starnetastro.com/documentation/windows-cuda/">https://starnetastro.com/documentation/windows-cuda/</a></p>
</blockquote>



<p class="wp-block-paragraph">libonnxruntime.so.1 をlibonnxruntime.so.1.26.0に入れ替えます。<br>次の様にシンボリックリンクを用います。</p>



<pre class="wp-block-code"><code>ln -s libonnxruntime.so.1.26.0 libonnxruntime.so.1
ln -s libonnxruntime.so.1 libonnxruntime.so

lrwxrwxrwx 1  libonnxruntime.so@ -> libonnxruntime.so.1
lrwxrwxrwx 1  libonnxruntime.so.1@ -> libonnxruntime.so.1.26.0</code></pre>



<h3 class="wp-block-heading">libonnxruntime.so.1.26.0 を導入</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">then download and extract ONNX Runtime GPU 1.26.0 for Windows x64 from GitHub.<br><a rel="noopener" target="_blank" href="https://starnetastro.com/documentation/windows-cuda/">https://starnetastro.com/documentation/windows-cuda/</a><br></p>
</blockquote>



<p class="wp-block-paragraph">onnxruntime-linux-x64-gpu-1.26.0.tgzを導入します。</p>



<p class="wp-block-paragraph">ONNX Runtime v1.27.0 は、<br>CUDA 12 packages are deprecated, please move to CUDA 13 ASAP.<br>CUDA 12 パッケージは非推奨となりました。できるだけ早く CUDA 13 へ移行してください。</p>



<p class="wp-block-paragraph">ここではCUDA12を用いているので、1.27ではなく、1.26を導入します。</p>



<p class="wp-block-paragraph"><a rel="noopener" target="_blank" href="https://github.com/microsoft/onnxruntime/releases">https://github.com/microsoft/onnxruntime/releases</a><br>onnxruntime-linux-x64-gpu-1.26.0.tgz をダウンロード、展開します。</p>



<pre class="wp-block-code"><code>onnxruntime-linux-x64-gpu-1.26.0/lib を確認
libonnxruntime.so.1.26.0
libonnxruntime_providers_cuda.so
libonnxruntime_providers_shared.so
libonnxruntime_providers_tensorrt.so #TensorRTを使わないので不要？</code></pre>



<p class="wp-block-paragraph">/usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/lib にある、<br>libonnxruntime.so.1 を別名にします。</p>



<pre class="wp-block-code"><code>sudo mv libonnxruntime.so.1 libonnxruntime.so.1.backup</code></pre>



<p class="wp-block-paragraph">/usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/libに、<br>libonnxruntime_providers_shared.so<br>があるならば、これも別名にします。</p>



<p class="wp-block-paragraph"><s>/usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/lib/ にある、</s> （2026,9,18 訂正）<br>onnxruntime-linux-x64-gpu-1.26.0/lib にある、<br>libonnxruntime.so.1.26.0<br>libonnxruntime_providers_cuda.so<br>libonnxruntime_providers_shared.so<br>libonnxruntime_providers_tensorrt.so<br>これらを、<br>/usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/lib へコピーします。</p>



<pre class="wp-block-code"><code>sudo cp libonnxruntime.so.1.26.0 /usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/lib/</code></pre>



<p class="wp-block-paragraph">下記のリンクを作ります。<br>libonnxruntime.so@ -&gt; libonnxruntime.so.1<br>libonnxruntime.so.1@ -&gt; libonnxruntime.so.1.26.0</p>



<pre class="wp-block-code"><code>sudo ln -s libonnxruntime.so.1.26.0 libonnxruntime.so.1
sudo ln -s libonnxruntime.so.1 libonnxruntime.so</code></pre>



<h3 class="wp-block-heading">deepsnr_linux_1.3.2-0147_ORT_x64_cli</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">deepsnr_linux_1.3.2-0147_ORT_x64_cli/README.txt, LICENSE.txt<br>本契約は、お客様とミキタ（ニキタ）・ミシウラ（以下「ライセンサー」という）との間で締結されるものです。<br><br>オプションのGPUアクセラレーション<br>同梱のONNX RuntimeファイルはCPUで実行されるため、GPUの設定は不要です。<br>NVIDIAに精通したユーザーは、同梱のONNX Runtimeファイルを互換性のあるONNX Runtime GPUランタイムに置き換え、必要なCUDA/cuDNNランタイムライブラリをダイナミックローダーから認識できるようにすることができます。CUDAの初期化に成功した場合、DeepSNRは次のように報告します：<br>バックエンド：ONNX Runtime (CUDA実行プロバイダー)<br>CUDAが利用できない、または不完全な場合、DeepSNRは同じ実行ファイルからCPU実行に切り替わります。手動でのGPUランタイムの変更により起動や処理に失敗した場合は、元のパッケージファイルを復元してください。</p>
</blockquote>



<p class="wp-block-paragraph">deepsnr_linux_1.3.2-0147_ORT_x64_cliも、上記 starnet2_linux_2.5.4-0214_ORT_x64_cli と同様にonnxruntime-linux-x64-gpu-1.26.0.tgzを用いてGPUが使えます。</p>



<h3 class="wp-block-heading">環境変数を追加</h3>



<p class="wp-block-paragraph">1: cuda runtime, cudnn, torch 各ライブラリがあるディレクトリにパスを通します。</p>



<p class="wp-block-paragraph">2: NVIDIA TensorRT（ディープラーニングの推論（予測・実行）を高速化するためのソフトウェア開発キット）が使えない場合でもCPU使用を禁止させる。処理できないならば、エラー表示で停止させる。</p>



<p class="wp-block-paragraph">3: 使用する計算エンジンを列挙し指定する。CUDA GPU、NVIDIA GPU、仕方がないからCPUの順番</p>



<pre class="wp-block-code"><code>export LD_LIBRARY_PATH=/home/$USER/.local/share/siril/venv/lib/python3.12/site-packages/nvidia/cuda_runtime/lib:/home/$USER/.local/share/siril/venv/lib/python3.12/site-packages/nvidia/cudnn/lib:/home/$USER/.local/share/siril/venv/lib/python3.12/site-packages/torch/lib:$LD_LIBRARY_PATH

export ORT_TENSORRT_UNAVAILABLE_FALLBACK=0

export ONNXRUNTIME_PROVIDER_PRIORITY="CUDAExecutionProvider,TensorrtExecutionProvider,CPUExecutionProvider"</code></pre>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p class="wp-block-paragraph">動作log</p>



<pre class="wp-block-code"><code>Linear mode: applied MTF autostretch to the StarNet input.
Reading input from /tmp/siril_starnet_qcse_6ma/input.tif...
Color image is detected!
Image size: 3996x2721
Restoring neural network checkpoint...
Loading ONNX model from file: /usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/StarNet2_weights.onnx
&#91;0;93m2026-08-11 19:21:51.880234368 &#91;W:onnxruntime:, session_state.cc:1367 VerifyEachNodeIsAssignedToAnEp] Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.&#91;m
&#91;0;93m2026-08-11 19:21:51.880253284 &#91;W:onnxruntime:, session_state.cc:1369 VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.&#91;m
Backend: ONNX Runtime (CUDA execution provider) # ← CUDA GPUが動いている
Working: Done!
Writing starless image to: /tmp/siril_starnet_qcse_6ma/starless.tif
Writing mask to: /tmp/siril_starnet_qcse_6ma/mask.tif
Saved Starmask to starnetmask_image.fit
StarNet finished.</code></pre>



<pre class="wp-block-code"><code># log
Linear mode: applied MTF autostretch to the StarNet input.
Reading input from /tmp/siril_starnet_tyv_jll2/input.tif...
Color image is detected!
Image size: 3008x3008
Restoring neural network checkpoint...
Loading ONNX model from file: /usr/local/share/starnet2_linux_2.5.4-0214_ORT_x64_cli/StarNet2_weights.onnx
&#91;0;93m2026-08-06 14:39:54.576734905 &#91;W:onnxruntime:, session_state.cc:1367 VerifyEachNodeIsAssignedToAnEp] Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.&#91;m
&#91;0;93m2026-08-06 14:39:54.576760112 &#91;W:onnxruntime:, session_state.cc:1369 VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.&#91;m
Backend: ONNX Runtime (CUDA execution provider) # ← GPUが動いている
Working: Done!
Writing starless image to: /tmp/siril_starnet_tyv_jll2/starless.tif
Writing mask to: /tmp/siril_starnet_tyv_jll2/mask.tif
Saved Starmask to starnetmask_image.fit
StarNet finished.</code></pre>



<h2 class="wp-block-heading">Windows11 と Starnet2</h2>



<p class="wp-block-paragraph">Windows での GPU アクセラレーションの設定手順<br><a rel="noopener" target="_blank" href="https://starnetastro.com/documentation/windows-cuda/">https://starnetastro.com/documentation/windows-cuda/</a></p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Use CUDA 12.x for the current ONNX Runtime 1.26.0 GPU path. Do not use CUDA 13.x runtime DLLs with the current CUDA-12 ONNX Runtime GPU DLLs.<br>現在の ONNX Runtime 1.26.0 の GPU パスでは、CUDA 12.x を使用してください。現在の CUDA-12 ONNX Runtime GPU DLL では、CUDA 13.x ランタイム DLL を使用しないでください。</p>
</blockquote>



<p class="wp-block-paragraph">こちらで (globe3.ddns.net) 確認された内容:</p>



<p class="wp-block-paragraph">以下の設定は、この Windows 11, Siril 1.5 環境で確認しています：</p>



<p class="wp-block-paragraph">オペレーティングシステム: Windows 11<br>GPU およびドライバー: NVIDIA GeForce RTX 5060、NVIDIA ドライバー 610.88 | WHQL<br>CUDA Toolkit: CUDA Toolkit 12.8.2 (April 2026). nvcc &#8211;version の出力は CUDA XX.x / VXX.x.x <br>cuDNN: CUDA 12.8 対応の cuDNN 9.24.0<br>ONNX Runtime GPU: Windows x64 対応の ONNX Runtime GPU 1.26.0</p>



<p class="wp-block-paragraph">NVIDIAドライバー、CUDA Toolkit、およびcuDNNをインストールし、Windowsの$PATH変数を通じて、これらのランタイムライブラリをSirilから使えるようにします。<br>以下を参照します。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Method 1: System-wide NVIDIA installer setup<br><a rel="noopener" target="_blank" href="https://starnetastro.com/documentation/windows-cuda/">https://starnetastro.com/documentation/windows-cuda/</a></p>
</blockquote>



<p class="wp-block-paragraph">1. Install NVIDIA driver</p>



<p class="wp-block-paragraph">2. Install CUDA Toolkit 12.8.2</p>



<pre class="wp-block-code"><code>where.exe nvcc</code></pre>



<p class="wp-block-paragraph">3. Install cuDNN 9.24.0</p>



<p class="wp-block-paragraph">CUDA 13.x フォルダではなく、CUDA 12.9 cuDNN フォルダを使用してください。cuDNN インストーラによって、以下の CUDA 13.3 フォルダが作成される場合がありますが、この ONNX Runtime 1.26.0 CUDA-12 のセットアップでは使用しないでください：</p>



<p class="wp-block-paragraph">4. Replace ONNX Runtime DLLs</p>



<p class="wp-block-paragraph">ONNX Runtime GPU 1.26.0 for Windows x64 from GitHub.<br>onnxruntime-win-x64-gpu-1.26.0.zip をダウンロード、展開<br><a rel="noopener" target="_blank" href="https://github.com/microsoft/onnxruntime/releases/">https://github.com/microsoft/onnxruntime/releases/</a></p>



<p class="wp-block-paragraph">onnxruntime.dll<br> ONNX Runtimeのメインエンジン。全ての計算命令を受け付けます。</p>



<p class="wp-block-paragraph">onnxruntime_providers_cuda.dll<br> NVIDIAのCUDAを使って計算するための専用プラグインです。</p>



<p class="wp-block-paragraph">onnxruntime_providers_shared.dll<br> メインエンジンと、CUDAプラグイン（および他のGPUプラグイン）との間で、データを仲介、共有するための共通ライブラリです。</p>



<p class="wp-block-paragraph">StarNet2 CLI と、Python仮想環境下に3つのdllファイルをそれぞれ置きます。<br>　starnet++.exe がC++アプリとして直接GPUを使うため。<br>　StarNet.py スクリプトが、Pythonを用いてGPUを使うため。</p>



<pre class="wp-block-code"><code>C:\local\starnet2_win_2.5.3-0208_ORT
C:\Users\USERNAME\AppData\Local\siril\venv\Lib\site-packages\onnxruntime\backend</code></pre>



<p class="wp-block-paragraph">5. CUDA 12.8DLLを配置する</p>



<p class="wp-block-paragraph">cublas64_12.dll 、 cudart64_12.dll （あれば cublasLt64_12.dll 、 cufft64_11.dll）を<br>C:\local\starnet2_win_2.5.3-0208_ORTディレクトリに置く。</p>



<pre class="wp-block-code"><code>### CUDA Toolkit 12.8 Update ディレクトリは以下
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8\bin</code></pre>



<p class="wp-block-paragraph">ここからコピーします。</p>



<p class="wp-block-paragraph">6. cudnn64_9.dll、cudnn_ops_infer64_9.dll、cudnn_cnn_infer64_9.dll　を配置する</p>



<p class="wp-block-paragraph">pythonコマンドを使って、それぞれのdllファイルをダウンロードします。<br>これはSiril 1.5仮想環境下ではなく、通常のPowerShell（またはターミナル）で行います。</p>



<pre class="wp-block-code"><code>python -m pip install --target "$env:TEMP\starnet-cuda-libs" "onnxruntime-gpu&#91;cuda,cudnn]==1.26.0"</code></pre>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph">以下のファイルが starnet++.exe があるディレクトリに配置します。<br> ONNX 1.26 GPU版の3つ: onnxruntime.dll, onnxruntime_providers_shared.dll, onnxruntime_providers_cuda.dll<br><br> CUDA 12.8（システムからコピーしたもの）: cublas64_12.dll, cudart64_12.dll<br><br> cuDNN 9（pythonを用いてダウンロード）: cudnn64_9.dll（およびその他の cudnn_*.dll）</p>
</div>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h3 class="wp-block-heading">Windows11での StarNet2.py を使用し、GPUを動かす</h3>



<p class="wp-block-paragraph">Siril Pythonスクリプト、StarNet2.pyをSiril に導入し、GPUを使って画像処理を行う為の設定をします。</p>



<p class="wp-block-paragraph">OS、ツール環境</p>



<pre class="wp-block-code"><code>Windows 11
RTX 5060（Blackwell / Compute Capability 12.0）
CUDA Toolkit 12.8
Python 3.12.13
StarNet2 2.5.4-0214 ORT x64 CLI</code></pre>



<p class="wp-block-paragraph">Windows11での設定手順</p>



<p class="wp-block-paragraph">各種ファイルを確認します。</p>



<pre class="wp-block-code"><code> nvidia-smi
Mon Aug 31 17:25:13 2026
+-----------------------------------------------------------------------------------------
| NVIDIA-SMI 616.56                 KMD Version: 616.56        CUDA UMD Version: 13.4     
+-----------------------------------------+------------------------+----------------------+
| GPU  Name                Driver-Model | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf        Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                       |                        |               MIG M. |
|=======================================+========================+======================|
|   0  NVIDIA GeForce RTX 5060 ...  WDDM  |   00000000:65:00.0 Off |                N/A |</code></pre>



<pre class="wp-block-code"><code>nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Fri_Feb_21_20:42:46_Pacific_Standard_Time_2025
Cuda compilation tools, release 12.8, V12.8.93
Build cuda_12.8.r12.8/compiler.35583870_0</code></pre>



<pre class="wp-block-code"><code>where.exe nvcc
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8\bin\nvcc.exe</code></pre>



<p class="wp-block-paragraph">cuDNN関連</p>



<pre class="wp-block-code"><code>where.exe cudnn64_9.dll
C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64\cudnn64_9.dll</code></pre>



<pre class="wp-block-code"><code>where.exe cudnn_ops64_9.dll
C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64\cudnn_ops64_9.dll</code></pre>



<pre class="wp-block-code"><code>where.exe cudnn_cnn64_9.dll
C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64\cudnn_cnn64_9.dll</code></pre>



<h4 class="wp-block-heading">starnet2_win_2.5.4-0214_ORT_x64_cli を眺める</h4>



<pre class="wp-block-code"><code>starnet2_win_2.5.4-0214_ORT_x64_cli.zip
展開</code></pre>



<p class="wp-block-paragraph">StarNetに同梱されているONNX Runtimeのバージョンを確認します。</p>



<pre class="wp-block-code"><code># powershell 
(Get-Item "C:\local\starnet2_win_2.5.4-0214_ORT_x64_cli\onnxruntime.dll").VersionInfo | Format-List FileVersion,ProductVersion


FileVersion    : 1.24.20260316.9.2d92497
ProductVersion : 1.24.20260316.9.2d92497</code></pre>



<p class="wp-block-paragraph">現在の環境<br>────────────────────────<br>RTX 5060<br>NVIDIA Driver 616.56<br>CUDA Toolkit 12.8<br>cuDNN 9.24 / CUDA 12.9</p>



<p class="wp-block-paragraph">StarNet CUDA構成<br>────────────────────────<br>ONNX Runtime 1.26.0<br>CUDA 12.x<br>cuDNN 9.x</p>



<h3 class="wp-block-heading">ONNX Runtime GPU 1.26.0 をstarnet2_win_2.5.4-0214_ORT_x64_cli へ導入</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">then download and extract ONNX Runtime GPU 1.26.0 for Windows x64 from GitHub.<br><a rel="noopener" target="_blank" href="https://starnetastro.com/documentation/windows-cuda/">https://starnetastro.com/documentation/windows-cuda/</a></p>
</blockquote>



<p class="wp-block-paragraph">onnxruntime-win-x64-gpu-1.26.0.zipを導入します。</p>



<p class="wp-block-paragraph">onnxruntime-win-x64-gpu-1.26.0.zipを展開し、onnxruntime-win-x64-gpu-1.26.0\libにある次のファイルを、starnet2_win_2.5.4-0214_ORT_x64_cliへ配置します。</p>



<p class="wp-block-paragraph">C:\local\starnet2_win_2.5.4-0214_ORT_x64_cli<br>│<br>├─ starnet2.exe<br>├─ StarNet2_weights.onnx<br>│<br>├─ onnxruntime.dll ← 1.26.0版に置換<br>├─ onnxruntime_providers_cuda.dll ← 新規<br>└─ onnxruntime_providers_shared.dll ← 1.26.0版に置換</p>



<pre class="wp-block-code"><code>### パス確認
where.exe cudnn_ops64_9.dll
C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64\cudnn_ops64_9.dll

where.exe cudnn_cnn64_9.dll
C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64\cudnn_cnn64_9.dll

where.exe cublas64_12.dll
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8\bin\cublas64_12.dll</code></pre>



<pre class="wp-block-code"><code>Get-ChildItem "C:\local\starnet2_win_2.5.4-0214_ORT_x64_cli\" -Filter "onnxruntime*"
Mode                 LastWriteTime         Length Name
----                 -------------         ------ ----
-a----        2026/08/31     19:15       15336248 onnxruntime.dll
-a----        2026/08/31     17:53           1094 ONNXRUNTIME_LICENSE.txt
-a----        2026/08/31     19:15      298998584 onnxruntime_providers_cuda.dll
-a----        2026/08/31     19:15          21816 onnxruntime_providers_shared.dll</code></pre>



<p class="wp-block-paragraph">Siril Pythonスクリプト StarNet.py から、starnet2_win_2.5.4-0214_ORT_x64_cliにある、<br>starnet2<br>StarNet2_weights.onnx<br>をそれぞれ指定します。</p>



<p class="wp-block-paragraph">タスクマネージャーから、GPUの動作を確認します。</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/siril-15-starnet2_linux_2-5-4-0214_ort_x64_cli_ubuntu2604/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>CUDA Toolkit 12.8.2 (April 2026)runfile +  Ubuntu 26.04 + nvidia-driver 580 + python version 3.12.13  + PyTorch + GraXpert + Cosmic Clarity Suite + Siril 1.5</title>
		<link>https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-graxpert-3-1-cosmicclaritysuite-siril-15/</link>
					<comments>https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-graxpert-3-1-cosmicclaritysuite-siril-15/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Wed, 05 Aug 2026 01:14:42 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<category><![CDATA[Ubuntu]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=4440</guid>

					<description><![CDATA[Ubuntu 26.04 に、天体画像処理ソフトを入れて、RTX5060で処理を行います。 Note: Starting with version 1.27, GPU packages published to PyPI [&#8230;]]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">Ubuntu 26.04 に、天体画像処理ソフトを入れて、RTX5060で処理を行います。</p>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">CUDA 12.xでは、ONNX Runtime: onnxruntime-gpu 1.28.0 を用いても、<br>GPU (RTX5060) は、動きません。<br><br>ONNX Runtime: 1.26.x-1.21.x<br>CUDA: 12.8<br>cuDNN: 9.x<br>この3つを組み合わせます。</p>
</div>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Note: Starting with version 1.27, GPU packages published to PyPI (onnxruntime-gpu) and NuGet (Microsoft.ML.OnnxRuntime.Gpu) are built with CUDA 13.0 by default. Older GPU package versions are built with CUDA 12.8 by default.<br>バージョン 1.27 以降、PyPI（onnxruntime-gpu）および NuGet（Microsoft.ML.OnnxRuntime.Gpu）に公開されている GPU パッケージは、デフォルトで CUDA 13.0 を使用してビルドされています。それ以前のバージョンの GPU パッケージは、デフォルトで CUDA 12.8 を使用してビルドされています。<br><a rel="noopener" target="_blank" href="https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements">https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements</a></p>
</blockquote>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">CosmicClaritySuite_Linux_AlternateBuildは、ラッパーコマンドを作り、ラッパーを介して、Siril Pythonスクリプトから動作させています。(2026, 08. 08)</p>
</div>



<p class="wp-block-paragraph">CUDA Toolkit 13系では、Cosmic Clarity Suiteが動作しない様なので、<br>CUDA Toolkit 12.8.2を用いて、Ubuntu + CUDA環境を作ります。</p>



<p class="wp-block-paragraph">バージョンの組み合わせは、<br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/deeplearning/cudnn/backend/v9.24.0/reference/support-matrix.html">https://docs.nvidia.com/deeplearning/cudnn/backend/v9.24.0/reference/support-matrix.html</a></p>



<p class="wp-block-paragraph">福山大学 工学部 情報工学科 『金子邦彦研究室へようこそ』<br>こちらで綺麗にまとめられています。<br>「推奨フロー図解」は、作業の流れをつかみ易くしてくれます。<br></p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Ubuntu 24.04 開発・研究環境構築ガイド<br><a rel="noopener" target="_blank" href="https://www.kkaneko.jp/tools/ubuntu/ubuntu.html">https://www.kkaneko.jp/tools/ubuntu/ubuntu.html</a></p>
</blockquote>



<h2 class="wp-block-heading">インストールとインストール後の落ち穂拾い</h2>



<h3 class="wp-block-heading">時刻：ハードウェアクロック</h3>



<p class="wp-block-paragraph">Windows OSとLinux OSが、デュアルブートさせていると、時刻表示が日本標準時から8時間ずれる。<br>これを日本標準時（JST +9h）に指定します。</p>



<pre class="wp-block-code"><code># ハードウェアクロックにローカルタイムを、ローカル時刻に指定
sudo timedatectl set-local-rtc 1

# ハードウェアクロックにローカルタイムを、UTC時刻に指定
sudo timedatectl set-local-rtc 0</code></pre>



<h3 class="wp-block-heading">C/C++コンパイラ、make、パッケージツール</h3>



<pre class="wp-block-code"><code>sudo apt update
sudo apt install build-essential gcc g++ make libtool dpkg-dev pkg-config dkms</code></pre>



<h2 class="wp-block-heading">Ubuntuのリポジトリからnvidia-driver-580-open, CUDA 12.x の runfile をダウンロードし、インストール</h2>



<pre class="wp-block-code"><code># インストールされているCUDA関連パッケージを確認
dpkg -l | grep cuda
dpkg -l | grep nvidia

# NVIDIA関連パッケージを削除
sudo apt --purge remove '*nvidia*'

# CUDA はrunfile を用いてインストールしているので、"cuda-uninstaller"を使う
# sudo apt --purge remove '*cuda*'
sudo apt --purge remove '*cudnn*'
</code></pre>



<pre class="wp-block-code"><code># 不要になった依存関係を自動削除
sudo apt autoremove --purge

# パッケージリストの情報を更新
sudo apt update

# Update device list 必要ならば
sudo update-pciids</code></pre>



<h3 class="wp-block-heading">ドライバのビルドに必要なカーネルヘッダー等をインストールする</h3>



<pre class="wp-block-code"><code>sudo apt update
sudo apt -y install linux-headers-$(uname -r) build-essential dkms</code></pre>



<pre class="wp-block-code"><code># システムが認識しているGPUと利用可能なドライバを表示
ubuntu-drivers devices
driver   : nvidia-driver-???-open - distro non-free recommended

# 2026年7月現在では、580
sudo apt install nvidia-driver-580-open

# initramfsを更新
sudo update-initramfs -u

# ドライバの有効化のためにシステムを再起動
reboot</code></pre>



<pre class="wp-block-code"><code>NVIDIA製GPUの稼働状況やシステム情報を確認します。
nvidia-smi</code></pre>



<h4 class="wp-block-heading">Boot miss ! 起動しなくなったら、Boot Repair</h4>



<pre class="wp-block-code"><code>sudo add-apt-repository ppa:yannubuntu/boot-repair
sudo apt update
sudo apt install boot-repair
boot-repair</code></pre>



<pre class="wp-block-code"><code>sudo chroot "/mnt/boot-sav/sda2" dpkg --configure -a
sudo chroot "/mnt/boot-sav/sda2" apt-get install -fy
sudo chroot "/mnt/boot-sav/sda2" apt-get purge --allow-remove-essential -y grub-com*
sudo chroot "/mnt/boot-sav/sda2" apt-get purge --allow-remove-essential -y grub2-com*
sudo chroot "/mnt/boot-sav/sda2" apt-get purge --allow-remove-essential -y shim-signed
sudo chroot "/mnt/boot-sav/sda2" apt-get purge --allow-remove-essential -y grub-common:*
sudo chroot "/mnt/boot-sav/sda2" apt-get purge --allow-remove-essential -y grub2-common:*
</code></pre>



<pre class="wp-block-code"><code>sudo chroot "/mnt/boot-sav/sda2" apt-get install -y grub-efi</code></pre>



<pre class="wp-block-code"><code>Boot successfully repaired.

A new file (/var/log/boot-repair/20260805_015552/Boot-Info_20260805_0155.txt) will open in your text viewer.

In case you still experience boot problem, indicate its content to:
boot.repair@gmail.com or to your favorite support forum.

Locked-NVram detected. Please do not forget to make your UEFI firmware boot on the Ubuntu 26.04 LTS entry (sda1/efi/ubuntu/grubx64.efi file) !</code></pre>



<h3 class="wp-block-heading">CUDA Toolkit 12.8.2 (April 2026)runfile</h3>



<pre class="wp-block-code"><code>wget https://developer.download.nvidia.com/compute/cuda/12.8.2/local_installers/cuda_12.8.2_570.211.01_linux.run
sudo sh cuda_12.8.2_570.211.01_linux.run</code></pre>



<p class="wp-block-paragraph">先に、nvidia-driver-580-openをインストールしているので、Driverを外す。</p>



<h4 class="wp-block-heading"><strong>環境変数を設定する</strong></h4>



<pre class="wp-block-code"><code>echo 'export CUDA_HOME=/usr/local/cuda-12.8' &gt;&gt; ${HOME}/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:${LD_LIBRARY_PATH}' &gt;&gt; ${HOME}/.bashrc
echo 'export PATH=/usr/local/cuda-12.8/bin:${PATH}' &gt;&gt; ${HOME}/.bashrc

# 確認
tail ${HOME}/.bashrc

# 設定を即時反映（現在のシェルのみ）
source ${HOME}/.bashrc

# 詳細バージョンを確認
cat /usr/local/cuda-12.8/version.json # version.json

# nvcc コマンドでバージョンを確認
nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Fri_Feb_21_20:23:50_PST_2025
Cuda compilation tools, release 12.8, V12.8.93
Build cuda_12.8.r12.8/compiler.35583870_0</code></pre>



<h3 class="wp-block-heading">cuDNN</h3>



<p class="wp-block-paragraph">アーカイブからもダウンロード出来るでしょう。<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cudnn-archive">https://developer.nvidia.com/cudnn-archive</a></p>



<pre class="wp-block-code"><code># ローカルインストール (FULL)
wget https://developer.download.nvidia.com/compute/cudnn/9.24.0/local_installers/cudnn-local-repo-ubuntu2604-9.24.0_1.0-1_amd64.deb
sudo dpkg -i cudnn-local-repo-ubuntu2604-9.24.0_1.0-1_amd64.deb
sudo cp /var/cudnn-local-repo-ubuntu2604-9.24.0/cudnn-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update

# CUDA 12 をインストールするには、パッケージを指定します。
sudo apt-get install cudnn9-cuda-12

### CUDA 13 をインストールするには、パッケージを指定します。
### sudo apt-get install cudnn9-cuda-13</code></pre>



<pre class="wp-block-code"><code># cuDNN　インストールを確認
dpkg -l | grep cudnn</code></pre>



<pre class="wp-block-code"><code>ii  cudnn-local-repo-ubuntu2604-9.24.0         1.0-1                                      amd64        cudnn-local repository configuration files
ii  cudnn9-cuda-12                             9.24.0.43-1                                amd64        NVIDIA cuDNN for CUDA 12
ii  cudnn9-cuda-12-9                           9.24.0.43-1                                amd64        NVIDIA cuDNN for CUDA 12.9
ii  libcudnn9-cuda-12                          9.24.0.43-1                                amd64        cuDNN runtime libraries for CUDA 12.9
ii  libcudnn9-dev-cuda-12                      9.24.0.43-1                                amd64        cuDNN development libraries for CUDA 12.9
ii  libcudnn9-headers-cuda-12                  9.24.0.43-1                                amd64        cuDNN header files for CUDA 12.9
ii  libcudnn9-static-cuda-12                   9.24.0.43-1                                amd64        cuDNN static libraries for CUDA 12.9</code></pre>



<h3 class="wp-block-heading">Siril Pytonスクリプトが GPUモードを設定</h3>



<p class="wp-block-paragraph">CPU内蔵グラフィックスが画面表示を行い、NVIDIA GeForceは計算ユニットとして使います。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">On-demand mode (recommended for laptop users)<br><a rel="noopener" target="_blank" href="https://ubuntu.fan/en/docs/ref/hardware/nvidia#dual-gpu-switching-prime">https://ubuntu.fan/en/docs/ref/hardware/nvidia#dual-gpu-switching-prime</a></p>
</blockquote>



<p class="wp-block-paragraph"></p>



<pre class="wp-block-code"><code># Check current mode
prime-select query

# On-demand mode (recommended for laptop users) #GPUモードを On-Demand に設定
sudo prime-select on-demand

# sudo reboot</code></pre>



<p class="wp-block-paragraph">CUDA、Python、Hashcatなどの場合では、TensorFlowやPyTorchやライブラリ経由でCUDAを使うならば、環境変数などの指定は要らないようです。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Dual GPU Intel-Nvidia / Prime Render Offloading / Ubuntu 20.04 – does not offload<br><a rel="noopener" target="_blank" href="https://forums.developer.nvidia.com/t/dual-gpu-intel-nvidia-prime-render-offloading-ubuntu-20-04-does-not-offload/166864">https://forums.developer.nvidia.com/t/dual-gpu-intel-nvidia-prime-render-offloading-ubuntu-20-04-does-not-offload/166864</a></p>
</blockquote>



<h3 class="wp-block-heading">python version 3.12.13</h3>



<pre class="wp-block-code"><code># sudo make altinstall</code></pre>



<p class="wp-block-paragraph">ソースから構築したpython3.x コマンドはインストールされますが、python3 コマンドはインストールされません。<br>これにより、python3 コマンドでは主に使用するバージョンのPythonを起動し、それ以外のバージョンのPythonは、 python3.12 のように、バージョンを指定して実行します。</p>



<p class="wp-block-paragraph">Ubuntu環境のPython<br><a rel="noopener" target="_blank" href="https://www.python.jp/install/ubuntu/index.html">https://www.python.jp/install/ubuntu/index.html</a></p>



<pre class="wp-block-code"><code>sudo apt update
sudo apt install build-essential libbz2-dev libdb-dev \
  libreadline-dev libffi-dev libgdbm-dev liblzma-dev \
  libncursesw5-dev libsqlite3-dev libssl-dev \
  zlib1g-dev uuid-dev tk-dev</code></pre>



<pre class="wp-block-code"><code>cd Python-3.12.13
./configure
make
sudo make altinstall</code></pre>



<pre class="wp-block-code"><code>python3.12 --version

Python 3.12.13</code></pre>



<h3 class="wp-block-heading">python アップデート (Python version 3.12.14)</h3>



<ul class="wp-block-list">
<li><code>make uninstall</code> できるかな？</li>



<li>手動でインストールされた関連ファイルを削除
<ul class="wp-block-list">
<li><code>cd /usr/local/bin</code></li>



<li><code>実行ファイル（バイナリ）の削除</code>
<ul class="wp-block-list">
<li>sudo rm -f /usr/local/bin/python3.12 /usr/local/bin/python3.12-config /usr/local/bin/pip3.12 /usr/local/bin/pydoc3.12 /usr/local/bin/idle3.12 /usr/local/bin/2to3-3.12</li>
</ul>
</li>



<li>ライブラリ（標準モジュール）の削除
<ul class="wp-block-list">
<li>sudo rm -rf /usr/local/lib/python3.12</li>
</ul>
</li>



<li>ヘッダーファイルの削除
<ul class="wp-block-list">
<li>sudo rm -rf /usr/local/include/python3.12</li>
</ul>
</li>



<li><code>マニュアルページの削除</code>
<ul class="wp-block-list">
<li>sudo rm -f /usr/local/share/man/man1/python3.12.1</li>
</ul>
</li>



<li>共有ライブラリキャッシュの更新
<ul class="wp-block-list">
<li>sudo ldconfig</li>
</ul>
</li>
</ul>
</li>



<li>Python version 3.12.14をソースビルド、インストール
<ul class="wp-block-list">
<li>sudo make altinstallを用いる</li>
</ul>
</li>



<li>Siril 1.5 仮想環境デイレクトリをワーキングディレクトリとする
<ul class="wp-block-list">
<li>仮想環境を 3.12.14 用に更新
<ul class="wp-block-list">
<li><code>python3.12 -m venv --upgrade</code> <code>venv</code></li>
</ul>
</li>
</ul>
</li>
</ul>



<h3 class="wp-block-heading">Siril 1.5 Python仮想環境を作ります</h3>



<p class="wp-block-paragraph">以下の作業は、仮想環境下で行います。<br>全ては、/home/$USER/.local/share/siril 以下にインストールされます。</p>



<pre class="wp-block-code"><code># Siril 1.5 Python仮想環境ディレクトリを作成
mkdir -p /home/$USER/.local/share/siril

# Python 3.12.13 を使って指定パスに仮想環境を作成
python3.12 -m venv /home/$USER/.local/share/siril/venv

# 仮想環境を有効化 (アクティベート)
source /home/$USER/.local/share/siril/venv/bin/activate

### 仮想環境内にpip のベースを生成する
### python -m ensurepip --upgrade

# pip 自身をインストール、最新版にアップデートする
python -m pip install --upgrade pip

### pip 確認
pip3 --version</code></pre>



<h3 class="wp-block-heading">PyTorch のインストール</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">PyTorchの旧バージョンのインストール<br>最新バージョンのインストールを推奨しますが、便宜上、旧バージョンのバイナリおよびインストール手順を以下に掲載しています。<br><a rel="noopener" target="_blank" href="https://pytorch.org/get-started/previous-versions/">https://pytorch.org/get-started/previous-versions/</a></p>
</blockquote>



<p class="wp-block-paragraph">上記を参考にします。</p>



<pre class="wp-block-code"><code># Siril 1.5 仮想環境で行う
python3.12 -m pip install torch torchvision --index-url https://download.pytorch.org/whl/cu128

### python3.12 -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128</code></pre>



<pre class="wp-block-code"><code># インストールログ
Successfully installed MarkupSafe-3.0.3 cuda-bindings-12.9.4 cuda-pathfinder-1.2.2 cuda-toolkit-12.8.1 filelock-3.29.0 fsspec-2026.4.0 jinja2-3.1.6 mpmath-1.3.0 networkx-3.6.1 numpy-2.4.4 nvidia-cublas-cu12-12.8.4.1 nvidia-cuda-cupti-cu12-12.8.90 nvidia-cuda-nvrtc-cu12-12.8.93 nvidia-cuda-runtime-cu12-12.8.90 nvidia-cudnn-cu12-9.19.0.56 nvidia-cufft-cu12-11.3.3.83 nvidia-cufile-cu12-1.13.1.3 nvidia-curand-cu12-10.3.9.90 nvidia-cusolver-cu12-11.7.3.90 nvidia-cusparse-cu12-12.5.8.93 nvidia-cusparselt-cu12-0.7.1 nvidia-nccl-cu12-2.28.9 nvidia-nvjitlink-cu12-12.8.93 nvidia-nvshmem-cu12-3.4.5 nvidia-nvtx-cu12-12.8.90 pillow-12.2.0 setuptools-78.1.0 sympy-1.14.0 torch-2.11.0+cu128 torchaudio-2.11.0+cu128 torchvision-0.26.0+cu128 triton-3.6.0 typing-extensions-4.15.0</code></pre>



<h4 class="wp-block-heading">動作テスト</h4>



<pre class="wp-block-code"><code># インストール確認
python3.12 -c "import torch; print('PyTorch Version:', torch.__version__); print('CUDA Available:', torch.cuda.is_available()); print('GPU Device:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"</code></pre>



<pre class="wp-block-code"><code>PyTorch Version: 2.11.0+cu128
CUDA Available: True
GPU Device: NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<h3 class="wp-block-heading">ONNX Runtime GPU インストール</h3>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">CUDA 12.xでは、ONNX Runtime: onnxruntime-gpu 1.28.0 を用いても、<br>GPU (RTX5060) は、動きません。<br><br>ONNX Runtime: 1.26.x-1.21.x ,  CUDA: 12.8 ,  cuDNN: 9.x<br>この3つの要素を組み合わせます。<br><br>onnxruntime-gpu==1.26.0,  version 1.26をインストールします。</p>
</div>



<p class="wp-block-paragraph">ONNX RuntimeとCUDAのバージョンの互換性<br><a rel="noopener" target="_blank" href="https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements">https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements</a><br>CUDA 12.X<br><a rel="noopener" target="_blank" href="https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#cuda-12x">https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#cuda-12x</a></p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Note: Starting with version 1.27, GPU packages published to PyPI (onnxruntime-gpu) and NuGet (Microsoft.ML.OnnxRuntime.Gpu) are built with CUDA 13.0 by default. Older GPU package versions are built with CUDA 12.8 by default.<br>バージョン 1.27 以降、PyPI（onnxruntime-gpu）および NuGet（Microsoft.ML.OnnxRuntime.Gpu）に公開されている GPU パッケージは、デフォルトで CUDA 13.0 を使用してビルドされています。それ以前のバージョンの GPU パッケージは、デフォルトで CUDA 12.8 を使用してビルドされています。<br><a rel="noopener" target="_blank" href="https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements">https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements</a></p>
</blockquote>



<p class="wp-block-paragraph">この互換性を元に、ONNX Runtime をインストールするのが大切らしい。</p>



<pre class="wp-block-code"><code># Siril 仮想環境で行う
source /home/$USER/.local/share/siril/venv/bin/activate

# 以前にruntimeをインストールしているならば、バージョン確認
pip list | grep onnxruntime
pip show onnxruntime-gpu

### 以前にruntimeをインストールしているならば、その関連パッケージを削除
### python3.12 -m pip uninstall onnxruntime onnxruntime-gpu

# python3.12 -m pip install onnxruntime-gpu

# version 1.26をインストールします。
# これ以降のバージョンは、CUDA 13.0 を使用してビルドされています。
python3.12 -m pip install onnxruntime-gpu==1.26.0

Successfully installed flatbuffers-25.12.19 onnxruntime-gpu-1.26.0 packaging-26.3 protobuf-7.35.1</code></pre>



<pre class="wp-block-code"><code># 動作テスト
python3.12 -c "import onnxruntime as ort; print(ort.get_available_providers())"</code></pre>



<h4 class="wp-block-heading">RTX 5060 動作確認</h4>



<pre class="wp-block-code"><code>source /home/$USER/.local/share/siril/venv/bin/activate
pip list |grep torch
torch                    2.11.0+cu128
torchaudio               2.11.0+cu128
torchvision              0.26.0+cu128</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 -c "import torch; print(torch.__version__, torch.cuda.is_available(), torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"
2.11.0+cu128 True NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 -c "import torch; print('Version:', torch.__version__); print('CUDA:', torch.cuda.is_available()); print('Device Name:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"
Version: 2.11.0+cu128
CUDA: True
Device Name: NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<h3 class="wp-block-heading">CUDA 12.8に適合するPyTorch、ONNX、天体処理用ライブラリを導入</h3>



<pre class="wp-block-code"><code>python3.12 -m pip install xisf tifffile astropy pillow pyqt6 sep lz4 zstandard

### python3.12 -m pip install onnxruntime-gpu xisf tifffile astropy pillow pyqt6 sep lz4 zstandard</code></pre>



<pre class="wp-block-code"><code># インストールログ
Installing collected packages: PyQt6-Qt6, flatbuffers, zstandard, tifffile, sep, PyYAML, PyQt6-sip, pyerfa, protobuf, packaging, lz4, astropy-iers-data, xisf, pyqt6, onnxruntime-gpu, astropy
Successfully installed PyQt6-Qt6-6.11.1 PyQt6-sip-13.12.0 PyYAML-6.0.3 astropy-8.0.1 astropy-iers-data-0.2026.8.3.0.53.6 flatbuffers-25.12.19 lz4-4.4.5 onnxruntime-gpu-1.28.0 packaging-26.3 protobuf-7.35.1 pyerfa-2.0.1.5 pyqt6-6.11.0 sep-1.4.1 tifffile-2026.7.31 xisf-0.9.7 zstandard-0.25.0</code></pre>



<h2 class="wp-block-heading">Siril 1.5</h2>



<h3 class="wp-block-heading">Siril 1.5 リビルド</h3>



<pre class="wp-block-code"><code>sudo apt update
sudo apt install meson git cmake
sudo apt install libglib2.0-dev libgtk-3-dev libgsl-dev liblcms2-dev \
  libcfitsio-dev libfftw3-dev libopencv-dev libgtksourceview-4-dev \
  libexiv2-dev libraw-dev libjxl-dev libheif-dev libxisf-dev \
  libpugixml-dev liblz4-dev libgit2-dev libffms2-dev libcurl4-gnutls-dev

#2026年7月　追加
sudo apt install wcslib-tools libgtk-4-dev libgtksourceview-5-dev

# Sirilソース
git clone --recurse-submodules https://gitlab.com/free-astro/siril.git

# Siril 開発版ビルド、インストール
cd siril
meson setup --prefix /usr/local --buildtype release _build
ninja -C _build
sudo ninja -C _build install
</code></pre>



<h2 class="wp-block-heading">GraXpert</h2>



<h3 class="wp-block-heading">GraXpert 3.1.0rc2</h3>



<p class="wp-block-paragraph">Set GraXpert Executable にて GraXpertファイルを指定して、GraXpert 3.1.0rc2を使います。</p>



<pre class="wp-block-code"><code># graxpert log
Starting script /home/minmin/.local/share/siril-scripts/processing/GraXpert-AI.py
Installing platformdirs. This may take a few seconds...
Collecting platformdirs
Downloading platformdirs-4.11.0-py3-none-any.whl.metadata (5.5 kB)
Downloading platformdirs-4.11.0-py3-none-any.whl (23 kB)
Installing collected packages: platformdirs
Successfully installed platformdirs-4.11.0
Successfully installed platformdirs
Installing onnx. This may take a few seconds...
Collecting onnx
Downloading onnx-1.22.0-cp312-abi3-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.metadata (8.5 kB)
Requirement already satisfied: numpy&gt;=1.23.2 in /home/minmin/.local/share/siril/venv/lib/python3.12/site-packages (from onnx) (2.4.4)
Requirement already satisfied: protobuf&gt;=4.25.1 in /home/minmin/.local/share/siril/venv/lib/python3.12/site-packages (from onnx) (7.35.1)
Requirement already satisfied: typing_extensions&gt;=4.15.0 in /home/minmin/.local/share/siril/venv/lib/python3.12/site-packages (from onnx) (4.15.0)
Collecting ml_dtypes&gt;=0.5.4 (from onnx)
Downloading ml_dtypes-0.5.4-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.metadata (8.9 kB)
Downloading onnx-1.22.0-cp312-abi3-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (19.1 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 19.1/19.1 MB 8.9 MB/s  0:00:02
17:49:44: Downloading ml_dtypes-0.5.4-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (5.0 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.0/5.0 MB 8.4 MB/s  0:00:00
Installing collected packages: ml_dtypes, onnx
Successfully installed ml_dtypes-0.5.4 onnx-1.22.0
Successfully installed onnx
Detected ONNX Runtime with providers: &#91;'TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']
ONNX Runtime is already installed: onnxruntime-gpu
numpy is installed
astropy is installed
Installing appdirs. This may take a few seconds...
Collecting appdirs
Downloading appdirs-1.4.4-py2.py3-none-any.whl.metadata (9.0 kB)
Downloading appdirs-1.4.4-py2.py3-none-any.whl (9.6 kB)
Installing collected packages: appdirs
Successfully installed appdirs-1.4.4
Successfully installed appdirs
Installing opencv-python. This may take a few seconds...
Collecting opencv-python
Downloading opencv_python-5.0.0.93-cp37-abi3-manylinux_2_28_x86_64.whl.metadata (19 kB)
Requirement already satisfied: numpy&gt;=2 in /home/minmin/.local/share/siril/venv/lib/python3.12/site-packages (from opencv-python) (2.4.4)
Downloading opencv_python-5.0.0.93-cp37-abi3-manylinux_2_28_x86_64.whl (73.8 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 73.8/73.8 MB 8.9 MB/s  0:00:08
Installing collected packages: opencv-python
Successfully installed opencv-python-5.0.0.93
Successfully installed opencv-python
PyQt6 is installed
Running command: requires
OK! This script is compatible with this version of Siril.

</code></pre>



<pre class="wp-block-code"><code># Deconvolution log
Python communication worker finished...
Starting script /home/minmin/.local/share/siril-scripts/processing/GraXpert-AI.py
platformdirs is installed
onnx is installed
Detected ONNX Runtime with providers: &#91;'TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']
ONNX Runtime is already installed: onnxruntime-gpu
numpy is installed
astropy is installed
appdirs is installed
opencv-python is installed
PyQt6 is installed
Running command: requires
OK! This script is compatible with this version of Siril.

Starting deconvolution
Calculated normalized PSFsize value: 0.2951167728237792
=== ONNX Execution Provider Tester ===
Creating ONNX model...
Model saved to temporary file: /tmp/tmpnm_0t04i.onnx
Running reference on CPU...
OK: CPU output computed.
Available execution providers:
- TensorrtExecutionProvider
- CUDAExecutionProvider
- CPUExecutionProvider
Testing each provider without fallback...
Testing TensorrtExecutionProvider...
*************** EP Error ***************
EP Error /onnxruntime_src/onnxruntime/python/onnxruntime_pybind_state.cc:534 void onnxruntime::python::RegisterTensorRTPluginsAsCustomOps(PySessionOptions&amp;, const onnxruntime::ProviderOptions&amp;) Please install TensorRT libraries as mentioned in the GPU requirements page, make sure they're in the PATH or LD_LIBRARY_PATH, and that your GPU is supported.

when using &#91;'TensorrtExecutionProvider']
Falling back to &#91;'CPUExecutionProvider'] and retrying.
****************************************
(x) TensorrtExecutionProvider: fallback occurred (used CPUExecutionProvider)

Testing CUDAExecutionProvider...
OK: CUDAExecutionProvider ran successfully
Testing CPUExecutionProvider...
OK: CPUExecutionProvider ran successfully
=== Summary ===
OK: Working providers (in priority order):
- CUDAExecutionProvider
- CPUExecutionProvider

→ Best available provider: CUDAExecutionProvider
Cached execution providers to /home/minmin/.config/siril/siril_onnx.conf
Using inference providers: &#91;'CUDAExecutionProvider', 'CPUExecutionProvider']
Finished deconvolution</code></pre>



<h2 class="wp-block-heading">Cosmic Clarity Suite</h2>



<h3 class="wp-block-heading">CosmicClaritySuite_Linux_AlternateBuildを展開</h3>



<p class="wp-block-paragraph">/use/local/share/CosmicClaritySuite_Linux/<br>├── SetiAstroCosmicClarity.py<br>├── SetiAstroCosmicClarity_denoise.py <br>│<br>├── *.onnx, *.pth (下記にある各種ファイル）<br>│ <br>└── run_Sharpen.sh, run_Sharpen.sh, 及びその他の付属ファイル群</p>



<h4 class="wp-block-heading">Linux Files Excluding _internal を導入</h4>



<p class="wp-block-paragraph">Linux Files Excluding _internal<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity/releases/tag/Linux">https://github.com/setiastro/cosmicclarity/releases/tag/Linux</a><br>ここにあるファイルを/use/local/share/CosmicClaritySuite_Linux/へ配置して、アップデートします。</p>



<p class="wp-block-paragraph">darkstar_*.onnx<br>darkstar_*.pth<br>deep_*.onnx<br>deep_*.pth<br>satellite*.onnx<br>satellite*.pth<br>setiastro*.exe<br>superres_*.onnx<br>superres_*.pth<br>各ファイルを入れ替えます。</p>



<h4 class="wp-block-heading">CosmicClaritySuite_Linux へ SetiAstroCosmicClarity.py を導入</h4>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph"><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity">https://github.com/setiastro/cosmicclarity</a><br>ここにある、&#8221;SetiAstroCosmicClarity.py&#8221; を使います。</p>



<p class="wp-block-paragraph">ラッパーファイルを作成 run_Sharpen.shは次の様です。<br>#!/bin/bash<br>source &#8220;$HOME/.local/share/siril/venv/bin/activate&#8221;<br>cd /usr/local/share/CosmicClaritySuite_Linux/<br>python3.12 ./SetiAstroCosmicClarity.py &#8220;$@&#8221;</p>
</div>



<p class="wp-block-paragraph">setiastrosuiteにある、SetiAstroCosmicClarity.py、SetiAstroCosmicClarity_denoise.pyをダウンロードし配置します。<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity">https://github.com/setiastro/cosmicclarity</a></p>



<p class="wp-block-paragraph">不足するパッケージをインストールします。</p>



<pre class="wp-block-code"><code>python -m pip install xisf astropy sep opencv-python</code></pre>



<h4 class="wp-block-heading">CosmicClaritySuite, Sharpening, denoise のラッパーを作成</h4>



<p class="wp-block-paragraph">[SetiAstroCosmicClarity.py]　へSiril から諸設定などの情報を受け取れるラッパーファイルを作ります。&#8221;run_Sharpen.sh&#8221;とします。</p>



<pre class="wp-block-code"><code>### ラッパーファイルを作成 run_Sharpen.sh
cd /usr/local/share/CosmicClaritySuite_Linux/</code></pre>



<p class="wp-block-paragraph">以下が、run_Sharpen.sh<br>SetiAstroCosmicClarity.pyに対して。</p>



<pre class="wp-block-code"><code>#!/bin/bash
source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux/
python3.12 ./SetiAstroCosmicClarity.py "$@"</code></pre>



<pre class="wp-block-code"><code># 実行権
chmod +x ./run_Sharpen.sh</code></pre>



<p class="wp-block-paragraph">[SetiAstroCosmicClarity_denoise.py]　へSiril から諸設定などの情報を受け取れるラッパーファイルを作ります。&#8221;run_Denoise.sh&#8221; とします。</p>



<p class="wp-block-paragraph">以下が、run_Denoise.sh<br>SetiAstroCosmicClarity_denoise.pyに対して。</p>



<pre class="wp-block-code"><code>#!/bin/bash
source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux/
python3.12 ./SetiAstroCosmicClarity_denoise.py "$@"</code></pre>



<pre class="wp-block-code"><code># 実行権
chmod +x ./run_Denoise.sh</code></pre>



<h4 class="wp-block-heading">SetiAstroCosmicClarity_satellitelinux.pyを導入, satellitelinuxのラッパーを作成</h4>



<pre class="wp-block-code"><code>source "$HOME/.local/share/siril/venv/bin/activate"
python3.12 ./SetiAstroCosmicClarity_satellitelinux.py

Traceback (most recent call last):
  File "/usr/local/share/CosmicClaritySuite_Linux/./SetiAstroCosmicClarity_satellitelinux.py", line 52, in &lt;module>
    import rawpy
ModuleNotFoundError: No module named 'rawpy'

Traceback (most recent call last):
  File "/usr/local/share/CosmicClaritySuite_Linux/./SetiAstroCosmicClarity_satellitelinux.py", line 58, in &lt;module>
    from skimage.transform import resize
ModuleNotFoundError: No module named 'skimage'

</code></pre>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">rawpy は、LibRaw ライブラリ用の使いやすい Python ラッパーです。 また、ホットピクセル/デッドピクセルの検出と修復のための追加機能も備えています。<br><a rel="noopener" target="_blank" href="https://pypi.org/project/rawpy/">https://pypi.org/project/rawpy/</a></p>
</blockquote>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">scikit-image は画像処理のためのアルゴリズムの集合体です。<br><a rel="noopener" target="_blank" href="https://scikit-image.org/">https://scikit-image.org/</a></p>
</blockquote>



<pre class="wp-block-code"><code>python3.12 -m pip install rawpy

Collecting rawpy
  Downloading rawpy-0.27.1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.metadata (9.7 kB)
Requirement already satisfied: numpy>=1.26.0 in /home/minmin/.local/share/siril/venv/lib/python3.12/site-packages (from rawpy) (2.4.4)
Downloading rawpy-0.27.1-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (3.0 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 3.0/3.0 MB 8.9 MB/s  0:00:00
Installing collected packages: rawpy
Successfully installed rawpy-0.27.1

python3.12 -m pip install scikit-image</code></pre>



<pre class="wp-block-code"><code>Error during headless execution: &#91;Errno 2] そのようなファイルやディレクトリはありません: '/usr/local/share/CosmicClaritySuite_Linux/./satellite_trail_detector_AI3.pth'

Error during headless execution: &#91;Errno 2] そのようなファイルやディレクトリはありません: '/usr/local/share/CosmicClaritySuite_Linux/./satellite_trail_detector_mobilenetv2.pth'

Error during headless execution: &#91;Errno 2] そのようなファイルやディレクトリはありません: '/usr/local/share/CosmicClaritySuite_Linux/./satelliteremovalAI3.pth'</code></pre>



<p class="wp-block-paragraph">上記ファイルは、<br>satellite_trail_detector_AI3.pth　→　satellite_trail_detector_AI3.5.pth<br>satellite_trail_detector_mobilenetv2.pth　→　satellite_trail_detector_mobilenetv2.5.pth<br>satelliteremovalAI3.pth　→　satelliteremovalAI3.5.pth</p>



<p class="wp-block-paragraph">satellite_trail_detector_AI3.onnx　→　satellite_trail_detector_AI3.5.onnx<br>satellite_trail_detector_mobilenetv2.onnx　→　satellite_trail_detector_mobilenetv2.5.onnx<br>satelliteremovalAI3.onnx　→　satelliteremovalAI3.5.onnx</p>



<p class="wp-block-paragraph">SetiAstroCosmicClarity_satellitelinux.py　をそれぞれ、書き換えます。</p>



<h3 class="wp-block-heading"><code>"$@"</code>　について</h3>



<p class="wp-block-paragraph">シェルスクリプトにおける &#8220;$@&#8221; は、<br>スクリプトに渡されたすべての引数を、元の区切り（スペースなど）を完全に保ったまま展開する特殊な変数</p>



<p class="wp-block-paragraph"><code>"$@"</code>　について参考にしたサイト<br><a rel="noopener" target="_blank" href="https://qiita.com/ko1nksm/items/98952723d7a71799673c">https://qiita.com/ko1nksm/items/98952723d7a71799673c</a><br></p>



<h4 class="wp-block-heading">Cosmic Clarity Sharpening を設定</h4>



<p class="wp-block-paragraph">Cosmic Clarity Sharpening Settings<br>Cosmic Clarity Executable:  run_Sharpen.sh (上記ラッパー)を指定する。</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h4 class="wp-block-heading">Cosmic Clarity Denoise を設定</h4>



<p class="wp-block-paragraph">Cosmic Clarity Executable<br>run_Sharpen.sh (上記ラッパー)を指定する。</p>



<h2 class="wp-block-heading">SyQon,  Prizm, Starless</h2>



<p class="wp-block-paragraph">/home/$USER/.local/share/siril <br>以下に、次のディレクトリがあり、そこへそれぞれのファイルを配置します。<br>syqon_prism　→　prism_mini.pt<br>syqon_starless　→　zenith.pt</p>



<p class="wp-block-paragraph">それぞれのスクリプトを起動すると、.ptファイルが読み込まれます。</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p class="wp-block-paragraph">ここまでが、本記事の本文になります。</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h2 class="wp-block-heading">参考：SetiAstroCosmicClarityが動かない時の試行錯誤</h2>



<p class="wp-block-paragraph">これ以下は試行錯誤した経過です。全ては、途中で作業を止めています。または用いていません。</p>



<p class="wp-block-paragraph">./SetiAstroCosmicClarity を動作確認</p>



<pre class="wp-block-code"><code>./SetiAstroCosmicClarity

PyInstaller/loader/pyimod02_importers.py:450: UserWarning: pkg_resources is deprecated as an API. See https://setuptools.pypa.io/en/latest/pkg_resources.html. The pkg_resources package is slated for removal as early as 2025-11-30. Refrain from using this package or pin to Setuptools&lt;81.

/usr/lib/x86_64-linux-gnu/gvfs/libgvfscommon.so: undefined symbol: g_variant_builder_init_static
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgvfsdbus.so</code></pre>



<p class="wp-block-paragraph">動かない。。。</p>



<p class="wp-block-paragraph">githubにある [SetiAstroCosmicClarity.py] を使ってみる。<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity">https://github.com/setiastro/cosmicclarity</a></p>



<p class="wp-block-paragraph">[SetiAstroCosmicClarity.py] を /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/へ置く。</p>



<pre class="wp-block-code"><code>source /home/$USER/.local/share/siril/venv/bin/activate
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/
python3.12 SetiAstroCosmicClarity.py</code></pre>



<p class="wp-block-paragraph">CUDA 12.8.2（2026年4月リリース版）は、CUDA 12系のビルドであり、古いONNX RuntimeやPyTorchの安定版との親和性があるらしい。</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p class="wp-block-paragraph">ONNX Runtime: <code>onnxruntime-gpu==1.26.0</code> を用いる</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p class="wp-block-paragraph">直面した問題と原因</p>



<p class="wp-block-paragraph">配布されているポータブルバイナリ（<code>AlternateBuild</code> 版の <code>SetiAstroCosmicClarity</code> や <code>setiastrobenchmark</code>）をそのまま実行した際、問題が発生。</p>



<ol class="wp-block-list">
<li><strong>GLib/GIOライブラリの衝突</strong>:
<ul class="wp-block-list">
<li>現象: <code>libgvfscommon.so: undefined symbol</code> エラー</li>



<li>原因: バイナリに内包された古い <code>libglib</code> が、Ubuntu 26.04の最新デスクトップ環境（GVFS）と競合</li>
</ul>
</li>



<li><strong>CUDAカーネルのミスマッチ（致命的）</strong>:
<ul class="wp-block-list">
<li>現象: <code>RuntimeError: CUDA error: no kernel image is available for execution on the device</code>（コアダンプ）。</li>



<li>原因: バイナリに内包されていたPyTorchが <code>2.4.1+cu124</code> (CUDA 12.4用) であったため、RTX5060 GPUを駆動するための計算データ（カーネル）が不足し、処理開始と同時にクラッシュ。</li>
</ul>
</li>
</ol>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p class="wp-block-paragraph">バイナリの仕様を諦め、公式の生のソースコード（<code>.py</code>）を改変する。</p>



<p class="wp-block-paragraph">１）本体スクリプトの取得と配置</p>



<ol class="wp-block-list">
<li>ディレクトリ <code>/usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/</code> に移動。</li>



<li>公式GitHubより <strong><code>SetiAstroCosmicClarity.py</code></strong> をダウンロード。<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity">https://github.com/setiastro/cosmicclarity</a></li>



<li>同フォルダ内のモデルデータ（<code>models/</code> 内の <code>~_AI4.onnx</code> など）をそのまま利用できる状態に統合。</li>
</ol>



<p class="wp-block-paragraph">２）本体スクリプト（<code>SetiAstroCosmicClarity.py</code>）の改造</p>



<p class="wp-block-paragraph">Sirilからの自動連携（画面を開かないCLI処理）に対応させるため、ファイル最末尾の引数解析ブロックを以下のように書き換えました。</p>



<pre class="wp-block-code"><code># --- 引数に --input を追加 ---
parser.add_argument('--input', type=str, help="Path to the input TIFF/FIT image")
args = parser.parse_args()
use_gpu = not args.disable_gpu

# --- 引数があれば入力パスを上書き ---
if args.input:
    input_dir = args.input
    print(f"&#91;CLI Mode] Input file overrides to: {input_dir}")

# --- 処理を実行 ---
process_images(input_dir, output_dir, args.sharpening_mode, args.nonstellar_strength, args.stellar_amount, args.nonstellar_amount, use_gpu, args.sharpen_channels_separately, args.auto_detect_psf)
</code></pre>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p class="wp-block-paragraph">３）Siril用スクリプトの配置</p>



<p class="wp-block-paragraph">ディレクトリ <code>~/.local/share/siril-scripts/processing/</code> に、新しく <code>CosmicClarity_Sharpen_CUDA.py</code> を作成し、以下の完全版コードを配置（他のファイルと同様、実行権限 <code>+x</code> は不要）。</p>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">~/.local/share/siril-scripts/processing/ は、Sirilにより上書きされます。</p>
</div>



<pre class="wp-block-code"><code>#!/usr/bin/env python3
# -*- coding: utf-8 -*-
import sys, subprocess os

# ユーザー名($USER)を自動取得して最新のCUDA環境を直接指定
current_user = os.environ.get("USER")
PYTHON_ENV = f"/home/{current_user}/.local/share/siril/venv/bin/python3.12"
CC_SCRIPT  = "/usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/SetiAstroCosmicClarity.py"

def main():
    args = sys.argv&#91;1:]
    cleaned_args = &#91;]
    for arg in args:
        if "SetiAstroCosmicClarity" in arg or arg.endswith(".py"):
            continue
        cleaned_args.append(arg)

    # 完璧に構築したPython仮想環境で本体スクリプトをキック
    cmd = &#91;PYTHON_ENV, CC_SCRIPT] + cleaned_args
    print(f"&#91;Siril-CUDA-CC] Redirecting execution via CUDA 12.8.2...")

    try:
        subprocess.run(cmd, check=True, text=True)
        print("&#91;Siril-CUDA-CC] CosmicClarity sharpening process completed successfully!")
    except subprocess.CalledProcessError as e:
        sys.exit(1)

if __name__ == "__main__":
    main()
</code></pre>



<p class="wp-block-paragraph">４）Siril上のGUI設定</p>



<ol class="wp-block-list">
<li>Siril 1.5の設定画面にある「Cosmic Clarity Executable」のBrowseボタンから、ダミーの実行トリガーとして <code>/usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/SetiAstroCosmicClarity.py</code> を指定。</li>



<li>Siril上部メニューの「スクリプト」から、追加された <code>CosmicClarity_Sharpen_CUDA</code> を選択して実行。</li>
</ol>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<p class="wp-block-paragraph">こちらは一度、設定したものです。今は用いていません。</p>



<h4 class="wp-block-heading">Cosmic Clarity Sharpening を展開</h4>



<p class="wp-block-paragraph">pyinstxtractor.py を使って、SetiAstroCosmicClarity を展開します。</p>



<pre class="wp-block-code"><code># Siril 1.5 仮想環境で行う
source /home/$USER/.local/share/siril/venv/bin/activate

cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild

python3.12 pyinstxtractor.py ./SetiAstroCosmicClarity

&#91;+] Processing ./SetiAstroCosmicClarity
&#91;+] Pyinstaller version: 2.1+
&#91;+] Python version: 3.12
&#91;+] Length of package: 57264312 bytes
&#91;+] Found 17 files in CArchive
&#91;+] Beginning extraction...please standby
&#91;+] Possible entry point: pyiboot01_bootstrap.pyc
&#91;+] Possible entry point: pyi_rth_inspect.pyc
&#91;+] Possible entry point: pyi_rth_pkgutil.pyc
&#91;+] Possible entry point: pyi_rth_multiprocessing.pyc
&#91;+] Possible entry point: pyi_rth_pkgres.pyc
&#91;+] Possible entry point: pyi_rth_setuptools.pyc
&#91;+] Possible entry point: pyi_rth_mplconfig.pyc
&#91;+] Possible entry point: pyi_rth_cryptography_openssl.pyc
&#91;+] Possible entry point: pyi_rth_pyqt6.pyc
&#91;+] Possible entry point: pyi_rth__tkinter.pyc
&#91;+] Possible entry point: setiastrocosmicclaritylinux.pyc
&#91;+] Found 7078 files in PYZ archive
&#91;+] Successfully extracted pyinstaller archive: ./SetiAstroCosmicClarity

You can now use a python decompiler on the pyc files within the extracted directory</code></pre>



<p class="wp-block-paragraph">SetiAstroCosmicClarity_extracted ディレクトリにある、<br>setiastrocosmicclaritylinux.pyc を<br>CosmicClaritySuite_Linux_AlternateBuild ディレクトリに配置</p>



<pre class="wp-block-code"><code># 動作確認
python3.12 ./setiastrocosmicclaritylinux.pyc

Using device: cuda (CUDA)
Identified 32bit floating point FITS iamge.
Sharpening Image: result_drizzle_19620s.fit
Total Chunks: 225
Stellar Sharpening:
Progress: 100.00% (225/225 chunks processed)
Stellar sharpening complete. Shape: (2877, 2877)
Saved 32-bit float sharpened image to: /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/./output/result_drizzle_19620s_sharpened.fit</code></pre>



<pre class="wp-block-code"><code>#!/bin/bash
source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/
python3.12 ./setiastrocosmicclaritylinux.pyc "$@"

# 実行権
chmod +x ./run_Sharpen.sh</code></pre>



<h3 class="wp-block-heading">Cosmic Clarity denoise を展開</h3>



<p class="wp-block-paragraph">python3.12 pyinstxtractor.py /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/SetiAstroCosmicClarity_denoise</p>



<pre class="wp-block-code"><code># Siril 1.5 仮想環境で行う
pwd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild
python3.12 pyinstxtractor.py ./SetiAstroCosmicClarity_denoise</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 ./setiastrocosmicclaritylinux.pyc
</code></pre>



<h4 class="wp-block-heading">Cosmic Clarity denoise ラッパーを作成</h4>



<pre class="wp-block-code"><code>### ラッパーファイルを作成 run_Denoise.sh
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/</code></pre>



<p class="wp-block-paragraph">以下が、run_Denoise.sh</p>



<pre class="wp-block-code"><code>#!/bin/bash

### RTX 5060（12.0）
### export TORCH_CUDA_ARCH_LIST="9.0"
### export TORCH_CUDA_ARCH_LIST="12.0"

source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/
python3.12 ./SetiAstroCosmicClarity_denoise.pyc "$@"</code></pre>



<p class="wp-block-paragraph">/use/local/share/Legacy_Linux_2004/<br>├── SetiAstroCosmicClarity.py<br>├── SetiAstroCosmicClarity_denoise.py <br>├── 📁 models/ （← ここにmodelsフォルダごと置く）<br>│ 　　├── 各種AIモデルファイル（.onnx や .pth など）<br>└── run_Sharpen.sh, run_Sharpen.sh, 及びその他の付属ファイル群</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-graxpert-3-1-cosmicclaritysuite-siril-15/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>Ubuntu 26.04 LTS, Windows11, PCを天文セットアップ</title>
		<link>https://globe3.ddns.net/wp/setting-up-ubuntu26-and-windows11-for-astronomy/</link>
					<comments>https://globe3.ddns.net/wp/setting-up-ubuntu26-and-windows11-for-astronomy/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Thu, 16 Jul 2026 02:11:48 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<category><![CDATA[Ubuntu]]></category>
		<category><![CDATA[画像処理]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=3551</guid>

					<description><![CDATA[ノートPCを天文趣味PCにします。インストールをメモします。ノートPC構成は次ですCPU: AMD Ryzen AI 7 350 プロセッサグラフィックス: NVIDIA GeForce RTX™ 5060 Laptop [&#8230;]]]></description>
										<content:encoded><![CDATA[
<p class="wp-block-paragraph">ノートPCを天文趣味PCにします。インストールをメモします。<br>ノートPC構成は次です<br>CPU: AMD Ryzen AI 7 350 プロセッサ<br>グラフィックス: NVIDIA GeForce RTX™ 5060 Laptop GPU　← ここを活用したい。</p>



<div class="wp-block-cocoon-blocks-sticky-box blank-box block-box sticky">
<p class="wp-block-paragraph">「開封」と言えば、梱包やパッケージを開けることでしょうが、私にとっては、中を見てパーツを確認することなので、ノートPCの裏蓋を開けることが「開封」です 😉<br>そんなアイキャッチ画像です</p>
</div>



<p class="wp-block-paragraph">天体画像処理には、Siril 1.5 を使います。ソースからビルドします。</p>



<p class="wp-block-paragraph">以下は、時系列では無く、間にメモがあり、まとまっていなので、とても見にくいかと思います。ごめんなさい。</p>



<p class="wp-block-paragraph">以下はインストール作業をまとめています。<br><a target="_self" href="https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-graxpert-3-1-cosmicclaritysuite-siril-15/">https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-graxpert-3-1-cosmicclaritysuite-siril-15/</a></p>



<p class="wp-block-paragraph"><br>試行錯誤した部分もそのまま残しています。貴方にとって不必要な操作もあります。<br>参考にする際は、前後の操作を確認して下さい。</p>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">CosmicClaritySuite_Linux_AlternateBuild<br>Python共有ライブラリを見てみる。<br>libpython3.12.so*<br>libpython3.12.so.1.0*</p>



<p class="wp-block-paragraph">Active Python releasesを見ると、Python 3.12 は &#8220;security&#8221; とあるので、<br>Python 3.12.13 &#8211; March 3, 2026を入れてみよう。<br><br>Ubuntu 26.04 + RTX5060 + CUDA 12.x + Python 3.12.13 + onnxruntime-gpu 1.26 + GraXPert  + Cosmic Clarity + ラッパーコマンド = Cheers🥂<br>かな？</p>
</div>



<h2 class="wp-block-heading">Ubuntu 26.04 LTS</h2>



<p class="wp-block-paragraph">福山大学 工学部 情報工学科 『金子邦彦研究室へようこそ』<br>こちらで綺麗にまとめられています。公開されていて、とても有り難いです。金子邦彦さん、感謝です。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">金子邦彦研究室へようこそ <a rel="noopener" target="_blank" href="https://www.kkaneko.jp/index.html">https://www.kkaneko.jp/index.html</a><br>Ubuntu 24.04 開発・研究環境構築ガイド<br>【概要】本ガイドでは、Ubuntu 24.04 LTSにおける各種ソフトウェアのインストール手順を説明する。プログラミング環境（C/C++、Python、Java、R）、NVIDIA GPU環境（ドライバ、CUDA Toolkit、cuDNN）、データベース（PostgreSQL）、人工知能・機械学習ライブラリ、3次元コンピュータグラフィックス・地理情報システム・メディア処理ツール、エディタ・統合開発環境などを扱う。<br><a rel="noopener" target="_blank" href="https://www.kkaneko.jp/tools/ubuntu/ubuntu.html">https://www.kkaneko.jp/tools/ubuntu/ubuntu.html</a><br><a rel="noopener" target="_blank" href="https://www.kkaneko.jp/tools/ubuntu/ubuntu_cudnn.html">https://www.kkaneko.jp/tools/ubuntu/ubuntu_cudnn.html</a></p>
</blockquote>



<h3 class="wp-block-heading">インストールできない。。。</h3>



<p class="wp-block-paragraph">インストール不具合があるならば、「セキュアブートを無効」は優先すべき選択肢と思います。</p>



<p class="wp-block-paragraph">SDDに問題があり、次のエラー表示。インストールが始まらない。。。</p>



<pre class="wp-block-code"><code>&#91;8768]:raise ValueError("Failed to find storage volume id='%s' config: %s"
&#91;8768]:ValueError: Failed to find storage volume id='disk-sda' config: {'ptable': 'gpt', 'serial': 'ELECOM_ESD-EYB_055Y00400047-0:0', 'path': '/dev/sda', 'wipe': 'superblock-recursive', 'preserve': False, 'name': '', 'grub_device': False, 'id': 'disk-sda', 'type': 'disk'}</code></pre>



<pre class="wp-block-code"><code>ストレージに対する指定は、

path	/dev/sda	対象となる物理ディスクのパス
wipe	superblock-recursive	ディスク上の古いファイルシステムやLVMのメタデータ（スーパーブロック）を完全に消去する
preserve   False	既存のパーティションやデータを保持しない（すべて上書きする）
name	''	ディスクの名前やラベルを指定しない（空白、指定なし）
grub_device	False	このディスク自体にはGRUB（ブートローダ）をインストールしない（※通常、GRUBはEFIパーティションなどの特定のパーティションにインストールされます）
type	disk	対象のタイプが「ディスク（物理ドライブ）」であることを示す
id	disk-sda	このディスク設定に対するユニークな識別子（ID）</code></pre>



<p class="wp-block-paragraph">インストールストレージのデバイス諸要件と、実際のストレージの要件が一致しないエラーのようです。何度もインストールを試し、パーティションを作り直しているので、このような症状になったのかと思い、インストールメディアの先頭データ512バイトを削除します。<br>これで、そのディスクの構成を記録してあるパーティションテーブルが無くなります。（乱暴。。。）</p>



<pre class="wp-block-code"><code>sudo dd if=/dev/zero of=/dev/sda bs=512 count=1
sudo dd if=/dev/zero of=/dev/sda bs=1M count=100 conv=fdatasync

sync</code></pre>



<p class="wp-block-paragraph"><code>/dev/sda</code> に残っているファイルシステムやパーティションなどの識別情報（署名）を削除します。</p>



<pre class="wp-block-code"><code>sudo wipefs -a /dev/sda</code></pre>



<p class="wp-block-paragraph">/dev/sda GPTパーティションテーブルを新しく作ります。</p>



<pre class="wp-block-code"><code>sudo parted /dev/sda mklabel gpt</code></pre>



<p class="wp-block-paragraph">この後、インストールを行います。</p>



<h3 class="wp-block-heading">インストール後の「落ち穂拾い」</h3>



<pre class="wp-block-code"><code>sudo apt update
sudo apt upgrade
reboot</code></pre>



<h4 class="wp-block-heading">バッテリーの話し</h4>



<p class="wp-block-paragraph">TLP: 電源管理</p>



<pre class="wp-block-code"><code>sudo apt install tlp tlp-rdw
REMOVING:
  power-profiles-daemon</code></pre>



<p class="wp-block-paragraph">TLP有効化</p>



<pre class="wp-block-code"><code>sudo systemctl enable tlp
sudo systemctl start tlp</code></pre>



<p class="wp-block-paragraph">TLP設定ファイルを編集</p>



<p class="wp-block-paragraph">こちらのノートPCでは、GNOME 拡張機能ではバッテリーのコントロールが出来ませんでした。<br>Windows11にてバッテリーマネージャー で設定すると、Ubuntu 26でも引き継がれている。<br>制御チップに指定がされ、残っているのだろうか？</p>



<pre class="wp-block-code"><code>電源デバイスのパスを表示
upower -e
/org/freedesktop/UPower/devices/battery_BAT0
/org/freedesktop/UPower/devices/line_power_AC0

upower -i /org/freedesktop/UPower/devices/battery_BAT0
charge-cycles:       N/A
</code></pre>



<pre class="wp-block-code"><code>/sys/class/power_supply/BAT
charge_full         6400000
charge_full_design  6400000

充電80% の時
charge_now　　　　　5120000

77%
current_now         2040000

voltage_min_design 15480000
voltage_now        16795000</code></pre>



<p class="wp-block-paragraph">GNOME 拡張機能を使用する場合は、以下の指定をします。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">

<a rel="noopener" target="_blank" href="https://www.youtube.com/watch?v=yDHEKpmSLO0" title="How to Limit Battery Charging to 80% in Ubuntu Linux" class="blogcard-wrap external-blogcard-wrap a-wrap cf"><div class="blogcard external-blogcard eb-left cf"><div class="blogcard-label external-blogcard-label"><span class="fa"></span></div><figure class="blogcard-thumbnail external-blogcard-thumbnail"><img loading="lazy" decoding="async" src="https://globe3.ddns.net/wp/wp-content/uploads/cocoon-resources/blog-card-cache/b99086c1ba963e2f2442d4f69908ca1e.jpg" alt="" class="blogcard-thumb-image external-blogcard-thumb-image" width="320" height="180" /></figure><div class="blogcard-content external-blogcard-content"><div class="blogcard-title external-blogcard-title">How to Limit Battery Charging to 80% in Ubuntu Linux</div><div class="blogcard-snippet external-blogcard-snippet">Keeping your laptop constantly charged at 100% may reduce its battery health over time.So, in this quick tutorial, I’ll ...</div></div><div class="blogcard-footer external-blogcard-footer cf"><div class="blogcard-site external-blogcard-site"><div class="blogcard-favicon external-blogcard-favicon"><img loading="lazy" decoding="async" src="https://www.google.com/s2/favicons?domain=https://www.youtube.com/watch?v=yDHEKpmSLO0" alt="" class="blogcard-favicon-image external-blogcard-favicon-image" width="16" height="16" /></div><div class="blogcard-domain external-blogcard-domain">www.youtube.com</div></div></div></div></a>



<ul class="wp-block-list">
<li>sudo apt install gnome-shell-extension-manager ：拡張機能マネージャーをインストール</li>



<li>「アプリを表示」→「Extension Manager」を開く→「Battery Health Charging」を検索</li>



<li>拡張機能の設定画面（歯車アイコン）を開き、 Install user privileges for this user</li>



<li>sudo 機能が必要なので、Install からパスワードを入れます。</li>



<li>パスワードを入力して権限が適用されると、画面右上のステータスバーから充電モード（100%、80%、60%など）を切り替えられる。</li>
</ul>
</blockquote>



<ul class="wp-block-list">
<li>sudo apt install gnome-shell-extension-manager ：拡張機能マネージャーをインストール</li>



<li>Extension Manager」アプリを開き、「Browse」 タブから 「Battery Health Charging」 を検索</li>



<li>拡張機能の設定画面（歯車アイコン）を開き、一番下にある Install user privileges</li>



<li>パスワードを入力して権限が適用されると、画面右上のステータスバーから充電モード（100%、80%、60%など）を切り替えられる</li>
</ul>



<h4 class="wp-block-heading">ファンコントロール</h4>



<pre class="wp-block-code"><code>lsmod | grep -E 'ec_sys|wmi|asus|dell|ideapad|thinkpad|hp'</code></pre>



<pre class="wp-block-code"><code>asus_wmi              106496  0
sparse_keymap          12288  1 asus_wmi
wmi_bmof               12288  0
nvidia_wmi_ec_backlight    12288  0
snd_rawmidi            57344  1 snd_seq_midi
snd_seq_device         16384  3 snd_seq,snd_seq_midi,snd_rawmidi
snd                   147456  23 snd_hda_codec_generic,snd_seq,snd_hda_codec_conexant,snd_seq_device,snd_hda_codec_hdmi,snd_hwdep,snd_hda_intel,snd_hda_codec,snd_sof,snd_soc_sdca,snd_timer,snd_compress,snd_soc_core,snd_pcm,snd_rawmidi
platform_profile       20480  2 amd_pmf,asus_wmi
video                  77824  4 nvidia_wmi_ec_backlight,asus_wmi,amdgpu,nvidia_modeset
wmi                    36864  4 video,nvidia_wmi_ec_backlight,asus_wmi,wmi_bmof</code></pre>



<p class="wp-block-paragraph"><code>asus_wmi</code> がロードされています。</p>



<p class="wp-block-paragraph">EC（Embedded Controller：エンベデッド・コントローラー）は、パソコンのマザーボード上に搭載され、キーボードの入力監視や電源管理、温度・ファン制御などのシステム基本動作を常時監視・制御する小型のマイクロコントローラーです。</p>



<p class="wp-block-paragraph">Linuxカーネルは、このECをASUS系のWMIインターフェースとして認識しているかもしれない。</p>



<p class="wp-block-paragraph">DMI情報は、PCのメーカー名・機種名・製品型番などを、ファームウェア（BIOS/UEFI）からLinuxが読み取った情報です。</p>



<h4 class="wp-block-heading">時刻表示が8時間ずれる</h4>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">ハードウェアクロック:  PC集積回路上あるCMOSに保存されている、時刻情報<br>システムクロック: OSのカーネルが管理している時刻情報<br><br>OSの起動時にハードウェアクロックの情報をシステムクロックに反映させることで、時刻情報を取得している。<br><br>Linux : UTC（世界標準時）をNTPサーバから取得し、<strong>UTC時刻情報をハードウェアクロックに設定している</strong>。その後システムクロックは、ハードウェアクロックから取得した時刻情報はUTCと解釈しているため、その時刻情報をローカルタイムに合わせて調整する。<br><br>Windows : ハードウェアクロックに設定する時刻情報はローカルタイムであるため、システムクロックはその時刻情報をローカルタイムと解釈して、そのままの値で取得したものを設定している。<br><a rel="noopener" target="_blank" href="https://qiita.com/Juice/items/63279e93f818ba74312d">https://qiita.com/Juice/items/63279e93f818ba74312d</a></p>
</blockquote>



<p class="wp-block-paragraph">SSDなどを替えて、OSを同じPCで起動した場合、Windowsが世界標準時になり、Linuxが日本時刻（JST）になる。</p>



<pre class="wp-block-code"><code># ハードウェアクロックにローカルタイムを、ローカル時刻に指定
sudo timedatectl set-local-rtc 1

# ハードウェアクロックにローカルタイムを、UTC時刻に指定
sudo timedatectl set-local-rtc 0</code></pre>



<h4 class="wp-block-heading">Linux GPU Control Application</h4>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Linux GPU Control Application<br><a rel="noopener" target="_blank" href="https://github.com/ilya-zlobintsev/LACT">https://github.com/ilya-zlobintsev/LACT</a></p>
</blockquote>



<p class="wp-block-paragraph">Ubuntu 26.04 へインストールします。</p>



<pre class="wp-block-code"><code>wget https://github.com/ilya-zlobintsev/LACT/releases/download/v0.10.0/lact-0.10.0-0.amd64.ubuntu-2604.deb
sudo apt install ./lact-0.10.0-0.amd64.ubuntu-2604.deb</code></pre>



<pre class="wp-block-code"><code>sudo systemctl enable --now lactd</code></pre>



<h3 class="wp-block-heading">C/C++コンパイラ、make、パッケージツール</h3>



<pre class="wp-block-code"><code>sudo apt update
sudo apt install build-essential gcc g++ make libtool dpkg-dev pkg-config dkms</code></pre>



<pre class="wp-block-code"><code>### sudo apt update
### sudo apt install build-essential gcc g++ make libtool pkg-config</code></pre>



<h3 class="wp-block-heading">NVIDIA関連ソフトウェア（ドライバ、CUDA Toolkit、cuDNN）</h3>



<p class="wp-block-paragraph">ここが今回の主役とも言えます。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">NVIDIA ドライバ、NVIDIA CUDA ツールキット 11.8、NVIDIA cuDNN 8 のインストール<br><a rel="noopener" target="_blank" href="https://www.kkaneko.jp/tools/ubuntu/ubuntu_cudnn.html">https://www.kkaneko.jp/tools/ubuntu/ubuntu_cudnn.html</a></p>



<p class="wp-block-paragraph">26.04 LTSとNVIDIA/AMDのGPGPU事情<br><a rel="noopener" target="_blank" href="https://gihyo.jp/admin/serial/01/ubuntu-recipe/0912">https://gihyo.jp/admin/serial/01/ubuntu-recipe/0912</a><br><br>Linux向けCUDAインストールガイド<br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html">https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html</a></p>
</blockquote>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">NVIDIA CUDA ツールキットとは？<br>NVIDIA® CUDA® ツールキットは、高性能な GPU アクセラレーション対応アプリケーションを作成するための開発環境を提供します。<br>このツールキットには、GPU アクセラレーション対応ライブラリ、デバッグおよび最適化ツール、C/C++ コンパイラ、およびランタイムライブラリが含まれています。<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cuda/toolkit">https://developer.nvidia.com/cuda/toolkit</a></p>
</blockquote>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">cuDNN とは？<br>NVIDIA® CUDA® ディープニューラルネットワークライブラリ（cuDNN）は、ディープニューラルネットワーク用のGPUアクセラレーション対応プリミティブライブラリです。cuDNNは、順方向および逆方向畳み込み、アテンション、行列乗算、プーリング、正規化などの標準ルーチンに対して、高度に最適化された実装を提供します。<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cudnn">https://developer.nvidia.com/cudnn</a></p>
</blockquote>



<p class="wp-block-paragraph">インストールの流れは、</p>



<ul class="wp-block-list">
<li>目的にあった<code>nvidia-driver</code>が入っているかを、確認します。
<ul class="wp-block-list">
<li>R595がインストールされる （2026年7月現在）</li>



<li>NVIDIAのプロプライエタリなドライバーが入っているならば、削除</li>



<li><code>nvidia-driver</code>を入れ直す</li>
</ul>
</li>



<li>Ubuntu 公式リポジトリの <code>nvidia-driver-???-open</code> をインストール</li>



<li>CUDA ツールキット　<code>cuda-toolkit-</code>12.x.x runfileにてインストール</li>



<li>cuDNN</li>



<li>GPUアーキテクチャをサポートする最新のPyTorch、CUDA（CUDA 12.x）の組み合わせにアップデート</li>
</ul>



<pre class="wp-block-code"><code># システム内容を確認します。
dpkg -l | grep nvidia

nvidia-driver-595-open
linux-modules-nvidia-595-open-7.0.0-28-generic
linux-modules-nvidia-595-open-generic-hwe-26.04
libnvidia-compute-595:amd64
nvidia-utils-595</code></pre>



<pre class="wp-block-code"><code>nvidia-smi

+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 595.71.05              Driver Version: 595.71.05      CUDA Version: 13.2   |  
+-----------------------------------------+------------------------+--------------------+
|   0  NVIDIA GeForce RTX
| N/A   41C    P8              6W /   75W |      26MiB /   8151MiB |      0%      Default </code></pre>



<pre class="wp-block-code"><code>nvidia-smi

+--------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.173.02             Driver Version: 580.173.02     CUDA Version: 13.0  |
+-----------------------------------------+------------------------+-------------------+</code></pre>



<p class="wp-block-paragraph">ここに表示されている、&#8221;CUDA Version: 13.2&#8243; や、&#8221;CUDA Version: 13.0&#8243;は、インストールされているCUDAではなくて、GPUがサポート出来るCUDAのバージョンです。</p>



<h4 class="wp-block-heading">どこの配布元のドライバを入れれば良いのか？</h4>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph"><a rel="noopener" target="_blank" href="https://www.nemotos.net/?p=5792">https://www.nemotos.net/?p=5792</a><br>Ubuntu 公式リポジトリと NVIDIA 公式リポジトリの両方に NVIDIA 関連パッケージが存在する状態になります。<code>apt-get full-upgrade</code> 時に、どちらのリポジトリからパッケージが引かれるかは優先度（priority）設定に依存し、意図しないバージョンが入る可能性があります。</p>
</blockquote>



<p class="wp-block-paragraph">アップデートなどで意図しないバージョンが更新されたり、ビルドに失敗することがあるらしい。</p>



<p class="wp-block-paragraph">2026年7月現在は、&#8221;Ubuntu 公式リポジトリの <code>restricted</code> セクションに <code>nvidia-driver-XXX-open</code> パッケージが収録され、NVIDIAのリポジトリを追加しなくても、nvidia-driverがインストールできます。</p>



<p class="wp-block-paragraph">Ubuntu 公式リポジトリの <code>nvidia-driver-580-open</code> を使えば、ドライバをインストールでき、カーネル更新との互換性も高まるそうです。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">NVIDIA ドライバのブランチ体系<br>NVIDIA は Linux GPU ドライバを “Long Term Support Branch”, “Production Branch”, “New Feature Branch” にわけてリリースしている<br><a rel="noopener" target="_blank" href="https://www.nemotos.net/?p=5792">https://www.nemotos.net/?p=5792</a></p>
</blockquote>



<p class="wp-block-paragraph">Ubuntu 26 04 インストール時に、プロプライエタリなドライバーを選択すると、<br>R595がインストールされます。（2026年7月19日現在）<br>しかし、R580を使いたいと思います。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Supported Drivers and CUDA Toolkit Versions を見ると、<br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/datacenter/tesla/drivers/supported-drivers-and-cuda-toolkit-versions.html">https://docs.nvidia.com/datacenter/tesla/drivers/supported-drivers-and-cuda-toolkit-versions.html</a><br><br>R595は、2027年3月までのサポート<br>R580は、2028年6月までのサポート</p>
</blockquote>



<p class="wp-block-paragraph">Long Term Support Branch である 580が良いな。<br>サポート期間が終わったならば、NVIDIA ドライバを次のLong Term Support Branchへ移行し、<br>CUDAもNVIDIA ドライバに合わせて、runファイルを使って、アップデートを行う予定にする。</p>



<h4 class="wp-block-heading">CUDAバージョンに関して</h4>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">GPUの型番にあったCUDAバージョンと、CUDA バージョンに対応した cuDNNをインストールすることが、「鍵」の一つになります。</p>
</div>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">GPUの型番にあったCUDAバージョンの選び方<br><a rel="noopener" target="_blank" href="https://zenn.dev/yumizz/articles/73d6c7d1085d2f">https://zenn.dev/yumizz/articles/73d6c7d1085d2f</a></p>
</blockquote>



<p class="wp-block-paragraph">こちらの表を参照します。<br>Wiki CUDA<br><a rel="noopener" target="_blank" href="https://en.wikipedia.org/wiki/CUDA">https://en.wikipedia.org/wiki/CUDA</a></p>



<p class="wp-block-paragraph">GPUが、RTX 5060 だと、Compute capability (version)は [12.0]、[Blackwell]</p>



<p class="wp-block-paragraph">CUDA SDK support vs. microarchitecture (cell: compute capability)表を見ると、<br>&#8220;Blackwell&#8221; → [12.0] → &#8220;CUDA SDK version(s)&#8221; → 12.8 と辿れます。</p>



<p class="wp-block-paragraph">CUDA 12.8 をインストールし、使用すれば良いと判ります。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">PyTorchはすでに CUDA 13.0 を標準（Stable）としてサポートするフェーズ<br><a rel="noopener" target="_blank" href="https://dev-discuss.pytorch.org/t/introducing-cuda-13-2-and-deprecating-cuda-12-8-release-2-12/3337">https://dev-discuss.pytorch.org/t/introducing-cuda-13-2-and-deprecating-cuda-12-8-release-2-12/3337</a></p>
</blockquote>



<h2 class="wp-block-heading">Ubuntuのリポジトリからnvidia-driver-580-open, <br>CUDA 12.x の runfile をダウンロードし、インストール</h2>



<h3 class="wp-block-heading">現在のNVIDIA関連パッケージを設定ファイルを含めて削除</h3>



<p class="wp-block-paragraph">Ubuntu 26インストール時に、プロプライエタリなドライバーインストールを選択したならば、</p>



<pre class="wp-block-code"><code># インストールされているCUDA関連パッケージを確認
dpkg -l | grep cuda
dpkg -l | grep nvidia

# NVIDIA関連パッケージを削除
sudo apt --purge remove '*nvidia*'
# CUDA はrunfile を用いてインストールしているので、"cuda-uninstaller"を使う
# sudo apt --purge remove '*cuda*'
sudo apt --purge remove '*cudnn*'

# 若しくは、
sudo apt --purge remove \
  'nvidia-driver-*' \
  'nvidia-dkms-*' \
  'nvidia-kernel-*' \
  'libnvidia-*' \
  'nvidia-utils-*' \
  'nvidia-compute-utils-*' \
  'nvidia-modprobe' \
  'nvidia-settings' \
  'xserver-xorg-video-nvidia-*'</code></pre>



<pre class="wp-block-code"><code># 削除結果
# nvidia-settings control panelが残る
dpkg -l | grep nvidia
ii  screen-resolution-extr 0.18.5build1 all Extension for the nvidia-settings control panel

# nvidia-settings control panelを削除
sudo apt --purge remove screen-resolution-extra
</code></pre>



<h3 class="wp-block-heading">現在のNVIDIA CUDA 関連パッケージの設定ファイルを含めて削除</h3>



<pre class="wp-block-code"><code>sudo apt --purge remove \
  'cuda*' \
  'libcudnn*' \
  'libnccl*' \
  'libcublas*' \
  'libcufft*' \
  'libcurand*' \
  'libcusolver*' \
  'libcusparse*' \
  'libnpp*' \
  'libnvjpeg*' \
  'libnvjitlink*' \
  'nsight-*'</code></pre>



<h3 class="wp-block-heading">NVIDIA repositoryの削除 、更新が固定されたNVIDIA, CUDA関連の固定解除</h3>



<pre class="wp-block-code"><code>ファイルがあれば、sudo rm -f　を用いて削除
/etc/apt/sources.list.d/cuda-ubuntu${UBUNTU_VER}-x86_64.list
/etc/apt/sources.list.d/cuda-*.list
/usr/share/keyrings/cuda-archive-keyring.gpg
/etc/apt/trusted.gpg.d/cuda-*.gpg</code></pre>



<pre class="wp-block-code"><code># 更新が固定されているNVIDIA、又は、CUDA関連のパッケージを検索し、
# 変数の"hold" を "remaining" に変更する
# Author: K. Nemoto
# Date: 2026-05-23
# Show remaining holds (should be empty)
echo "--- Remaining apt-mark holds for nvidia/cuda ---"
remaining=$(apt-mark showhold | grep -E 'nvidia|cuda' || true)
if &#91; -z "$remaining" ]; then
  echo "(none)"
else
  echo "$remaining"
fi

#”https://gitlab.com/kytk/lin4neuro-jammy/-/raw/main/installer-scripts/cuda_cleanup.sh”</code></pre>



<h3 class="wp-block-heading">依存関係の削除と、パッケージリスト情報を更新</h3>



<pre class="wp-block-code"><code># 不要になった依存関係を自動削除
sudo apt autoremove --purge

# パッケージリストの情報を更新
sudo apt update

# Update device list 必要ならば
sudo update-pciids</code></pre>



<h3 class="wp-block-heading">ドライバのビルドに必要なカーネルヘッダー等をインストールする</h3>



<pre class="wp-block-code"><code>sudo apt update
sudo apt install linux-headers-$(uname -r) build-essential dkms

### ヘッダー無しならば
### sudo apt update
### sudo apt install build-essential dkms</code></pre>



<h3 class="wp-block-heading">Ubuntuが推奨するドライバを指定してインストール</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">NVIDIA が GPU カーネルモジュールのソースコードを open source（GPL/MIT デュアルライセンス）で公開したことに伴い、Ubuntu は公式リポジトリの <code>restricted</code> セクションに <code>nvidia-driver-XXX-open</code> パッケージを収録するようになりました。つまり、NVIDIAのリポジトリを追加しなくても、nvidia-driverがインストールできるようになったということです。<br><a rel="noopener" target="_blank" href="https://www.nemotos.net/?p=5792">https://www.nemotos.net/?p=5792</a></p>
</blockquote>



<p class="wp-block-paragraph">Ubuntuのリポジトリからnvidia-driver-580-open　をインストールします。</p>



<pre class="wp-block-code"><code>Ubuntu 26インストール時に、プロプライエタリなドライバーインストールを選択せず、
コマンドでnvidiaドライバをインストールするならば、以下を行います。

# システムが認識しているGPUと利用可能なドライバを表示
ubuntu-drivers devices
driver   : nvidia-driver-???-open - distro non-free recommended

# "recommended" 推奨されるドライバをインストール
sudo apt install nvidia-driver-???-open

# 2026年7月現在では、580
sudo apt install nvidia-driver-580-open

# initramfsを更新
# /bootにある既存のInitramfsの更新や削除、/bootへの新規Initramfsの作成に使うコマンド
sudo update-initramfs -u

# 内容を確認
#  sudo lsinitramfs /boot/initrd.img-`uname -r`

# ドライバの有効化のためにシステムを再起動
sudo /sbin/shutdown -r now</code></pre>



<p class="wp-block-paragraph">モジュール関連</p>



<pre class="wp-block-code"><code>Building module(s)........... done.
Signing module /var/lib/dkms/nvidia/580.173.02/build/nvidia.ko
Signing module /var/lib/dkms/nvidia/580.173.02/build/nvidia-modeset.ko
Signing module /var/lib/dkms/nvidia/580.173.02/build/nvidia-drm.ko
Signing module /var/lib/dkms/nvidia/580.173.02/build/nvidia-uvm.ko
Signing module /var/lib/dkms/nvidia/580.173.02/build/nvidia-peermem.ko
Installing /lib/modules/7.0.0-28-generic/updates/dkms/nvidia.ko.zst
Installing /lib/modules/7.0.0-28-generic/updates/dkms/nvidia-modeset.ko.zst
Installing /lib/modules/7.0.0-28-generic/updates/dkms/nvidia-drm.ko.zst
Installing /lib/modules/7.0.0-28-generic/updates/dkms/nvidia-uvm.ko.zst
Installing /lib/modules/7.0.0-28-generic/updates/dkms/nvidia-peermem.ko.zst
Running depmod... done.
xserver-xorg-video-nvidia-580 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
libnvidia-compute-580:amd64 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
libnvidia-fbc1-580:amd64 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
nvidia-compute-utils-580 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
libnvidia-gl-580:amd64 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
libnvidia-decode-580:amd64 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
nvidia-utils-580 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
libnvidia-encode-580:amd64 (580.173.02-0ubuntu0.26.04.1) を設定しています ...
nvidia-driver-580-open (580.173.02-0ubuntu0.26.04.1) を設定しています ...
gnome-menus (3.38.1-1ubuntu1) のトリガを処理しています ...
libc-bin (2.43-2ubuntu2) のトリガを処理しています ...
man-db (2.13.1-1build1) のトリガを処理しています ...
dbus (1.16.2-2ubuntu4) のトリガを処理しています ...
dracut (110-11) のトリガを処理しています ...
update-initramfs: Generating /boot/initrd.img-7.0.0-28-generic
desktop-file-utils (0.28-1build1) のトリガを処理しています ...</code></pre>



<pre class="wp-block-code"><code>何がインストールされたか確認
dpkg -l | grep nvidia
libnvidia-cfg1-580:amd64   580.159.03-0ubuntu0.26.04.1   amd64   NVIDIA binary OpenGL/GLX configuration library
libnvidia-common-580      580.159.03-0ubuntu0.26.04.1    amd64   Shared files used by the NVIDIA libraries
libnvidia-compute-580:amd64   580.159.03-0ubuntu0.26.04.1   amd64    NVIDIA libcompute package
libnvidia-decode-580:amd64  580.159.03-0ubuntu0.26.04.1  amd64   NVIDIA Video Decoding runtime libraries
libnvidia-egl-wayland1:amd64  1:1.1.21-1  amd64  Wayland EGL External Platform library -- shared library
libnvidia-encode-580:amd64   580.159.03-0ubuntu0.26.04.1   amd64   NVENC Video Encoding runtime library
libnvidia-extra-580:amd64     580.159.03-0ubuntu0.26.04.1     amd64   Extra libraries for the NVIDIA driver
libnvidia-fbc1-580:amd64    580.159.03-0ubuntu0.26.04.1    amd64   NVIDIA OpenGL-based Framebuffer Capture runtime library
libnvidia-gl-580:amd64  580.159.03-0ubuntu0.26.04.1   amd64  NVIDIA OpenGL/GLX/EGL/GLES GLVND libraries and Vulkan ICD
nvidia-compute-utils-580 　580.159.03-0ubuntu0.26.04.1   amd64  NVIDIA compute utilities
nvidia-dkms-580-open    580.159.03-0ubuntu0.26.04.1  amd64    NVIDIA DKMS package (open kernel module)
nvidia-driver-580-open  580.159.03-0ubuntu0.26.04.1  amd64 NVIDIA driver (open kernel) metapackage
nvidia-firmware-580-580.159.03  580.159.03-0ubuntu0.26.04.1   amd64  Firmware files used by the kernel module
nvidia-kernel-common-580  580.159.03-0ubuntu0.26.04.1   amd64    Shared files used with the kernel module
nvidia-kernel-source-580-open  580.159.03-0ubuntu0.26.04.1  amd64    NVIDIA kernel source package
nvidia-prime      0.8.17.2build1       all       Tools to enable NVIDIA's Prime
nvidia-settings    510.47.03-0ubuntu7    amd64    Tool for configuring the NVIDIA graphics driver
nvidia-utils-580    580.159.03-0ubuntu0.26.04.1    amd64     NVIDIA driver support binaries
screen-resolution-extra    0.18.5build1  all    Extension for the nvidia-settings control panel
xserver-xorg-video-nvidia-580 580.159.03-0ubuntu0.26.04.1   amd64    NVIDIA binary Xorg driver</code></pre>



<pre class="wp-block-code"><code>NVIDIA製GPUの稼働状況やシステム情報を確認します。
nvidia-smi

+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.173.02             Driver Version: 580.173.02     CUDA Version: 13.0   |
+---------------------------------------+------------------------+----------------------+</code></pre>



<h3 class="wp-block-heading">CUDA Toolkit runfile</h3>



<p class="wp-block-paragraph">CUDA Toolkit 12.8.2 (April 2026) runfile をインストールする。</p>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph"><code>--toolkit</code> : オプションを指定して、ドライバーをインストールしない。<br>との案内がありましたが、2026年7月では、CUDA 13.x , 12.xともドライバーにチェックが入っています。<br><br>ドライバは Ubuntu 公式リポジトリの deb で管理し、CUDA Toolkit は runfile で管理することができる。<br>とありますが、インストールツールには、マークされているので、ドライバーが不要ならば、マークを外します。</p>
</div>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">NVIDIA CUDA Toolkit <a rel="noopener" target="_blank" href="https://developer.nvidia.com/cuda-downloads">https://developer.nvidia.com/cuda-downloads</a> で利用可能です。<br><br>archive からでもファイルをダウンロードできます。<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cuda-toolkit-archive">https://developer.nvidia.com/cuda-toolkit-archive</a><br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html">https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html</a></p>
</blockquote>



<pre class="wp-block-code"><code># CUDA Toolkit 12.8.2
wget https://developer.download.nvidia.com/compute/cuda/12.8.2/local_installers/cuda_12.8.2_570.211.01_linux.run
sudo sh cuda_12.8.2_570.211.01_linux.run --toolkit</code></pre>



<pre class="wp-block-code"><code># CUDA Toolkit 13.0.3

wget https://developer.download.nvidia.com/compute/cuda/13.0.3/local_installers/cuda_13.0.3_580.126.20_linux.run
sudo sh cuda_13.0.3_580.126.20_linux.run</code></pre>



<pre class="wp-block-code"><code>wget https://developer.download.nvidia.com/compute/cuda/12.8.2/local_installers/cuda_12.8.2_570.211.01_linux.run</code></pre>



<p class="wp-block-paragraph">「nvidia-fs をインストールするには、mofedを先にインストールしなさいよ」とインストールログ</p>



<pre class="wp-block-code"><code># インストールログ
&#91;INFO]: previous version of nvidia-fs is not installed, nvidia-fs version: 2.24.3 will be installed.
&#91;INFO]: getting mofed Status
&#91;INFO]: installation status shows that mofed is not installed,please install mofed before continuing nvidia_fs install.
&#91;ERROR]: Install of nvidia-fs failed, quitting</code></pre>



<h4 class="wp-block-heading">nvidia-fs と mofed</h4>



<p class="wp-block-paragraph">複数のGPUユニットがある場合に有効でしょう。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">nvidia-fs とは、<br><a rel="noopener" target="_blank" href="https://jp.ubuntu.com/blog/canonical-announces-it-will-distribute-nvidia-doca-ofed-in-ubuntu-jp">https://jp.ubuntu.com/blog/canonical-announces-it-will-distribute-nvidia-doca-ofed-in-ubuntu-jp</a></p>



<p class="wp-block-paragraph">RDMA（リモートダイレクトメモリアクセス）やNVIDIA GPUDirectといった高度な機能により、CPUのオフロード、テールレイテンシの短縮と予測、負荷時におけるスループットの維持を実現します。<br><br>NVIDIA GPUDirect とは、<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/gpudirect">https://developer.nvidia.com/gpudirect</a><br><br>システムメモリを介さずに、NVMeやNVMe over Fabric（NVMe-oF）などのローカルまたはリモートストレージと、GPUメモリとの間に、直接的なデータパスを実現します。</p>
</blockquote>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Canonical、UbuntuでのNVIDIA DOCA-OFEDの配布を発表　by Canonical on 21 April 2026<br><a rel="noopener" target="_blank" href="https://jp.ubuntu.com/blog/canonical-announces-it-will-distribute-nvidia-doca-ofed-in-ubuntu-jp">https://jp.ubuntu.com/blog/canonical-announces-it-will-distribute-nvidia-doca-ofed-in-ubuntu-jp</a><br><br>インストール<br>Install and maintain DOCA-OFED<br><a rel="noopener" target="_blank" href="https://ubuntu.com/server/docs/how-to/networking/install-doca-ofed/">https://ubuntu.com/server/docs/how-to/networking/install-doca-ofed/</a></p>
</blockquote>



<pre class="wp-block-code"><code># インストール
uname -r
sudo add-apt-repository ppa:canonical-nvidia/doca-stable
sudo apt update
sudo apt install -y doca-ofed-&lt;your_kernel_flavor&gt;
## もしも、uname -r -&gt; 7.0.0-28-generic などならば、以下
sudo apt install doca-ofed-generic
Finally, reboot the machine.!!</code></pre>



<pre class="wp-block-code"><code># インストールログ
Installing: doca-ofed-generic

Installing dependencies: doca-ofed-userspace-26.01  ibsim-doc-26.01  libibmad5-26.01 libnl-3-dev  linux-main-modules-doca-ofed-26.01-7.0.0-28-generic  mlnx-iproute2-26.01
opensm-doc-26.01  ucx-26.01

dpcp-26.01  ibutils2-26.01  libibnetdisc5-26.01 libnl-route-3-dev linux-modules-doca-ofed-26.01-generic
mlnx-ofed-kernel-utils-26.01  perftest-26.01

ibacm-26.01  ibverbs-providers-26.01  libibumad-dev-26.01   libopensm-26.01
mft-26.01    mlnx-tools-26.01    rdma-core-26.01

ibarr-26.01  ibverbs-utils-26.01  libibumad3-26.01  libopensm-devel-26.01
mft-mlx5-26.01    ofed-scripts-26.01   rdmacm-utils-26.01

ibdump-26.01 infiniband-diags-26.01 libibverbs-dev-26.01  librdmacm-dev-26.01
mft-nvredfish-26.01    openmpi-26.01  rshim-26.01

ibsim-26.01  libibmad-dev-26.01  libibverbs1-26.01  librdmacm1-26.01  mlnx-ethtool-26.01
opensm-26.01    srptools-26.01
</code></pre>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">OSの再起動をお忘れなく</p>
</div>



<pre class="wp-block-code"><code>DOCA-OFED をアップグレード
apt list --upgradable
sudo apt update &amp;&amp; sudo apt full-upgrade</code></pre>



<pre class="wp-block-code"><code>#--silent 使用許諾やオプション表示がされない
sudo sh cuda_12.8.2_570.211.01_linux.run --silent --toolkit</code></pre>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph">既存のパッケージマネージャーによるドライバのインストールが検出されました。続行する前に、これを削除することを強くお勧めします。</p>



<p class="wp-block-paragraph">エンドユーザー使用許諾契約書<br>NVIDIA ソフトウェア使用許諾契約書およびソフトウェア使用許諾契約書の CUDA 補足条項。CUDA ツールキットエンドユーザー使用許諾契約書は、NVIDIA CUDA ツールキット、NVIDIA CUDA サンプル、NVIDIA ディスプレイドライバ、NVIDIA Nsight ツール（Visual Studio Edition）、および CUDA API、プログラミングモデル、開発ツールに関する関連ドキュメントに適用されます。</p>
</div>



<pre class="wp-block-code"><code># インストールログ
===========
= Summary =
===========
Driver:   Not Selected
Toolkit:  Installed in /usr/local/cuda-12.8/

Please make sure that
 -   PATH includes /usr/local/cuda-12.8/bin
 -   LD_LIBRARY_PATH includes /usr/local/cuda-12.8/lib64, or, add /usr/local/cuda-12.8/lib64 to /etc/ld.so.conf and run ldconfig as root

To uninstall the CUDA Toolkit, run cuda-uninstaller in /usr/local/cuda-12.8/bin
***WARNING: Incomplete installation! This installation did not install the CUDA Driver. A driver of version at least 570.00 is required for CUDA 12.8 functionality to work.
To install the driver using this installer, run the following command, replacing &lt;CudaInstaller&gt; with the name of this run file:
    sudo &lt;CudaInstaller&gt;.run --silent --driver

警告：インストールが不完全です！このインストールでは、CUDA ドライバーがインストールされませんでした。CUDA 12.8 の機能を利用するには、バージョン 570.00 以上のドライバーが必要です。
このインストーラーを使用してドライバーをインストールするには、以下のコマンドを実行してください。この際、&lt;CudaInstaller&gt; をこの実行ファイルの名前に置き換えてください：
    sudo &lt;CudaInstaller&gt;.run --silent --driver

Logfile is /var/log/cuda-installer.log
</code></pre>



<pre class="wp-block-code"><code>sudo sh cuda_12.8.2_570.211.01_linux.run --silent --driver

 Existing package manager installation of the driver found. It is strongly recommended that you remove this before continuing.
 Override this check by passing --override-driver-check
</code></pre>



<pre class="wp-block-code"><code># CUDA Toolkit 13.3
wget https://developer.download.nvidia.com/compute/cuda/13.3.1/local_installers/cuda_13.3.1_610.43.02_linux.run
sudo sh cuda_13.3.1_610.43.02_linux.run --toolkit

Logfile is /var/log/cuda-installer.log</code></pre>



<h4 class="wp-block-heading"><strong>環境変数を設定する</strong></h4>



<pre class="wp-block-code"><code>echo 'export CUDA_HOME=/usr/local/cuda-12.8' &gt;&gt; ${HOME}/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:${LD_LIBRARY_PATH}' &gt;&gt; ${HOME}/.bashrc
echo 'export PATH=/usr/local/cuda-12.8/bin:${PATH}' &gt;&gt; ${HOME}/.bashrc

# 確認
tail ${HOME}/.bashrc

# 設定を即時反映（現在のシェルのみ）
source ${HOME}/.bashrc

# 詳細バージョンを確認
cat /usr/local/cuda-12.8/version.json # version.json

# nvcc コマンドでバージョンを確認
nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Wed_Aug_20_01:58:59_PM_PDT_2025
Cuda compilation tools, release 13.0, V13.0.88
Build cuda_13.0.r13.0/compiler.36424714_0</code></pre>



<pre class="wp-block-code"><code>echo 'export CUDA_HOME=/usr/local/cuda-13.0' &gt;&gt; ${HOME}/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-13.0/lib64:${LD_LIBRARY_PATH}' &gt;&gt; ${HOME}/.bashrc
echo 'export PATH=/usr/local/cuda-13.0/bin:${PATH}' &gt;&gt; ${HOME}/.bashrc</code></pre>



<pre class="wp-block-code"><code>export CUDA_HOME=/usr/local/cuda-12.8/
echo 'export CUDA_HOME=/usr/local/cuda-12.8' &gt;&gt; ${HOME}/.bashrc
export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:${LD_LIBRARY_PATH}
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:${LD_LIBRARY_PATH}' &gt;&gt; ${HOME}/.bashrc
export PATH=/usr/local/cuda-12.8/bin:${PATH}
echo 'export PATH=/usr/local/cuda-12.8/bin:${PATH}' &gt;&gt; ${HOME}/.bashrc</code></pre>



<pre class="wp-block-code"><code>echo 'export CUDA_HOME=/usr/local/cuda-13.3' &gt;&gt; ${HOME}/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-13.3/lib64:${LD_LIBRARY_PATH}' &gt;&gt; ${HOME}/.bashrc
echo 'export PATH=/usr/local/cuda-13.3/bin:${PATH}' &gt;&gt; ${HOME}/.bashrc</code></pre>



<pre class="wp-block-code"><code># .bashrc
export CUDA_HOME=/usr/local/cuda-12.8
export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:${LD_LIBRARY_PATH}
export PATH=/usr/local/cuda-12.8/bin:${PATH}</code></pre>



<h3 class="wp-block-heading">CUDA Toolkit を削除する場合</h3>



<pre class="wp-block-code"><code>sudo ./cuda-uninstaller
./cuda-uninstaller: error while loading shared libraries: libxml2.so.2: cannot open shared object file: No such file or directory

sudo ln -s /usr/lib/x86_64-linux-gnu/libxml2.so.16 /usr/lib/x86_64-linux-gnu/libxml2.so.2

sudo ./cuda-uninstaller
./cuda-uninstaller: /lib/x86_64-linux-gnu/libxml2.so.2: no version information available (required by ./cuda-uninstaller)
./cuda-uninstaller: /lib/x86_64-linux-gnu/libxml2.so.2: no version information available (required by ./cuda-uninstaller)
./cuda-uninstaller: /lib/x86_64-linux-gnu/libxml2.so.2: no version information available (required by ./cuda-uninstaller)
cat: /var/log/nvidia/.uninstallManifests/kernelobjects-components/uninstallManifest-nvidia_fs: No such file or directory
make: *** ターゲット 'uninstall' を make するルールがありません.  中止.
rmmod: ERROR: Module nvidia_fs is not currently loaded

 Successfully uninstalled</code></pre>



<h3 class="wp-block-heading">cuDNN</h3>



<p class="wp-block-paragraph">バージョンの互換性がとても重要らしい。多くの方がサイトで苦労話を綴っています。<br>CUDA バージョンに対応した cuDNNを探します。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">cuDNN version9.24.0 は、CUDA Toolkit Version 13と12に互換しています。<br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/deeplearning/cudnn/backend/latest/reference/support-matrix.html">https://docs.nvidia.com/deeplearning/cudnn/backend/latest/reference/support-matrix.html</a></p>
</blockquote>



<p class="wp-block-paragraph">こちらから、cuDNNインストール手順を表示させます。<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cudnn-downloads">https://developer.nvidia.com/cudnn-downloads</a></p>



<p class="wp-block-paragraph">アーカイブからもダウンロード出来るでしょう。<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cudnn-archive">https://developer.nvidia.com/cudnn-archive</a></p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">FULL と JIT の違いは、NVIDIA Ampere GPU アーキテクチャおよびそれ以降GPUに適応しています。ダウンロードサイズ及び、インストール後のバイナリサイズを大幅に削減します。cuDNN JIT は、グラフ API を通じてランタイム・フュージョン・エンジンのみをサポートすることでこのサイズ削減を実現していますが、その代わりにサポート範囲やパフォーマンスが若干低下します。<br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/deeplearning/cudnn/latest/developer/graph-api.html#cudnn-jit">https://docs.nvidia.com/deeplearning/cudnn/latest/developer/graph-api.html#cudnn-jit</a></p>
</blockquote>



<pre class="wp-block-code"><code>libcudnn9-headers-cuda-12 9.25.0.15-1 &#91;75.2 kB]
libcudnn9-cuda-12 9.25.0.15-1 &#91;505 MB]</code></pre>



<pre class="wp-block-code"><code># ローカルインストール (FULL)
wget https://developer.download.nvidia.com/compute/cudnn/9.24.0/local_installers/cudnn-local-repo-ubuntu2604-9.24.0_1.0-1_amd64.deb
sudo dpkg -i cudnn-local-repo-ubuntu2604-9.24.0_1.0-1_amd64.deb
sudo cp /var/cudnn-local-repo-ubuntu2604-9.24.0/cudnn-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update

# CUDA 12 をインストールするには、パッケージを指定します。
sudo apt-get install cudnn9-cuda-12

# CUDA 13 をインストールするには、パッケージを指定します。
sudo apt-get install cudnn9-cuda-13</code></pre>



<pre class="wp-block-code"><code># keyring 使用 ネットワークインストール
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2604/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update

# FULLをインストール
# CUDA 12 用に&#91;FULL]インストールするには、CUDA 12 固有のパッケージをインストールします。
# sudo apt-get -y install cudnn
sudo apt-get -y install cudnn9-cuda-12

CUDA 13 用に&#91;FULL]インストールするには、CUDA 13 の特定のパッケージをインストールします。
# sudo apt-get -y install cudnn9-cuda-13</code></pre>



<pre class="wp-block-code"><code>sudo dpkg -i cudnn-local-repo-ubuntu2604-9.24.0_1.0-1_amd64.deb

#The public cudnn-local-repo-ubuntu2604-9.24.0 GPG key does not appear to be installed.
#To install the key, run this command:
#sudo cp /var/cudnn-local-repo-ubuntu2604-9.24.0/cudnn-local-B991B0EF-keyring.gpg /usr/share/keyrings/</code></pre>



<p class="wp-block-paragraph">cuDNN　インストールを確認</p>



<pre class="wp-block-code"><code>dpkg -l | grep cudnn</code></pre>



<pre class="wp-block-code"><code>cudnn9-cuda-12     9.25.0.15-1 　　amd64        NVIDIA cuDNN for CUDA 12
cudnn9-cuda-12-9   9.25.0.15-1     amd64        NVIDIA cuDNN for CUDA 12.9
libcudnn9-cuda-12  9.25.0.15-1     amd64        cuDNN runtime libraries for CUDA 12.9
libcudnn9-dev-cuda-12      9.25.0.15-1   amd64   cuDNN development libraries for CUDA 12.9
libcudnn9-headers-cuda-12  9.25.0.15-1    amd64    cuDNN header files for CUDA 12.9
libcudnn9-static-cuda-12   9.25.0.15-1   amd64   cuDNN static libraries for CUDA 12.9
</code></pre>



<h2 class="wp-block-heading">Python は、どのバージョンを入れるのか？</h2>



<h3 class="wp-block-heading"> PyTorch の対応</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Python 3.14 でのいくつかの変更が PyTorch 全体に影響を与えました：<br><a rel="noopener" target="_blank" href="https://dev-discuss.pytorch.org/t/torch-compile-support-for-python-3-14-completed/3276">https://dev-discuss.pytorch.org/t/torch-compile-support-for-python-3-14-completed/3276</a><br><br>Python 3.13 のサポートにおいて直面した重要な技術的課題について説明します。<br><a rel="noopener" target="_blank" href="https://dev-discuss.pytorch.org/t/torch-compile-support-for-python-3-13-completed/2738">https://dev-discuss.pytorch.org/t/torch-compile-support-for-python-3-13-completed/2738</a></p>
</blockquote>



<h3 class="wp-block-heading">Python 3.12.x + Cosmic Clarity</h3>



<p class="wp-block-paragraph">CosmicClaritySuite_Linux_AlternateBuild.tar.xz<br>　&#8221;AlternateBuildパッケージ&#8221;<br>/usr/local/share/CosmicClaritySuite_Linux/_internal/ にあるPython関連</p>



<pre class="wp-block-code"><code>/usr/local/share/Legacy_CosmicClaritySuite_Linux/_internal/
libpython3.12.so*
libpython3.12.so.1.0*</code></pre>



<p class="wp-block-paragraph">Python 3.12.x をインストールします。</p>



<pre class="wp-block-code"><code># sudo make altinstall</code></pre>



<p class="wp-block-paragraph">ソースから構築した&#8221;python3.x&#8221; コマンドはインストールされますが、&#8221;python3&#8243; コマンドはインストールされません。<br>これにより、python3 コマンドでは主に使用するバージョンのPythonを起動し、それ以外のバージョンのPythonは、 python3.12 のように、バージョンを指定して実行します。</p>



<p class="wp-block-paragraph">Python 3.12 をインストールします。<br><br>Python 3.12.14 Release date: Aug. 12, 2026<br><a rel="noopener" target="_blank" href="https://www.python.org/downloads/release/python-31214/">https://www.python.org/downloads/release/python-31214/</a><br>Python 3.12.10 Release date: April 8, 2025<br><a rel="noopener" target="_blank" href="https://www.python.org/downloads/release/python-31210/">https://www.python.org/downloads/release/python-31210/</a></p>



<p class="wp-block-paragraph">Ubuntu環境のPython<br><a rel="noopener" target="_blank" href="https://www.python.jp/install/ubuntu/index.html">https://www.python.jp/install/ubuntu/index.html</a></p>



<pre class="wp-block-code"><code>sudo apt update
sudo apt install build-essential libbz2-dev libdb-dev \
  libreadline-dev libffi-dev libgdbm-dev liblzma-dev \
  libncursesw5-dev libsqlite3-dev libssl-dev \
  zlib1g-dev uuid-dev tk-dev</code></pre>



<pre class="wp-block-code"><code>cd Python-3.12.13
./configure
make
sudo make altinstall</code></pre>



<pre class="wp-block-code"><code>python3 --version

python3.12 --version
Python 3.12.13</code></pre>



<h3 class="wp-block-heading">Siril 1.5 Python仮想環境を作ります</h3>



<p class="wp-block-paragraph">以下の作業は、仮想環境下で行います。<br>全ては、/home/$USER/.local/share/siril 以下にインストールされます。</p>



<pre class="wp-block-code"><code># Siril 1.5 Python仮想環境ディレクトリを作成
mkdir -p /home/$USER/.local/share/siril

# Python 3.12.13 を使って指定パスに仮想環境を作成
python3.12 -m venv /home/$USER/.local/share/siril/venv

# 仮想環境を有効化 (アクティベート)
source /home/$USER/.local/share/siril/venv/bin/activate

# 仮想環境内にpip のベースを生成する
python -m ensurepip --upgrade

# pip 自身を最新版にアップデートする
python -m pip install --upgrade pip

### pip 確認
pip3 --version</code></pre>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">WARNING: Running pip as the &#8216;root&#8217; user can result in broken permissions and conflicting behaviour with the system package manager, possibly rendering your system unusable. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv. Use the &#8211;root-user-action option if you know what you are doing and want to suppress this warning.<br>警告：「root」ユーザーとして pip を実行すると、権限設定が破損したり、システムのパッケージマネージャーとの競合が発生したりし、システムが使用不能になる可能性があります。代わりに仮想環境を使用することをお勧めします：https://pip.pypa.io/warnings/venv。操作内容を十分に理解しており、この警告を抑制したい場合は、&#8211;root-user-action オプションを使用してください。</p>
</div>



<h2 class="wp-block-heading">GPUアーキテクチャをサポートする、PyTorch、CUDA（CUDA 12.x）の組み合わせにする</h2>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">th.install_torch を用いたPyTorchとcudaインストール方法を試した経過を掲載しますが、<br>途中で止めました。<br>こちらでは、Siril仮想環境下で、pipを用いてPyTorchをインストールしました。<br></p>
</div>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Sirilのスクリプトエディタにソースコードをコピーして、[run]<br><a rel="noopener" target="_blank" href="https://www.facebook.com/groups/841870683930320/posts/1375756583875058/">https://www.facebook.com/groups/841870683930320/posts/1375756583875058/</a></p>
</blockquote>



<p class="wp-block-paragraph">Sirilの画面上部にある Scripts をクリック、スクリプトエディタを表示します。<br>次のコードをコピーしました。<br>2026年7月現在、このスクリプトは試していません。</p>



<pre class="wp-block-code"><code>import sirilpy as s
th = s.TorchHelper()
th.install_torch(cuda_version='cu128', force_reinstall=True)
th.test_torch()</code></pre>



<pre class="wp-block-code"><code># エラーコード cu130
Traceback (most recent call last):
16:50:11:   File "/tmp/siril-script-S7EUS3.py", line 3, in &lt;module&gt;
16:50:11:     th.install_torch(cuda_version='cu130', force_reinstall=True)
16:50:11: TypeError: TorchHelper.install_torch() got an unexpected keyword argument 'force_reinstall'
16:50:11: Python process (PID: 53902) exited with status 1
16:50:11: Python communication worker finished...</code></pre>



<h3 class="wp-block-heading">force_reinstall 引数が削除、cuda versionを指定</h3>



<p class="wp-block-paragraph">Sirilのアップデート（バージョン1.4.2以降〜1.5）に伴い、<code>sirilpy</code> 内の <code>TorchHelper.install_torch()</code> から <code>force_reinstall</code><strong> </strong>が削除</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Facebookで見つけたんですが、残念ながらもう使えなくなってしまいました。<br><a rel="noopener" target="_blank" href="https://discuss.pixls.us/t/syqon-starless-and-pytorch-install-failure/56239">https://discuss.pixls.us/t/syqon-starless-and-pytorch-install-failure/56239</a><br><a rel="noopener" target="_blank" href="https://www.facebook.com/groups/841870683930320/posts/1398410764942973/">https://www.facebook.com/groups/841870683930320/posts/1398410764942973/</a><br><br><a rel="noopener" target="_blank" href="https://blog.qualiteg.com/pytorch_and_supported_gpu_version/">https://blog.qualiteg.com/pytorch_and_supported_gpu_version/</a><br><a rel="noopener" target="_blank" href="https://discuss.pixls.us/t/syqon-starless-and-pytorch-install-failure/56239">https://discuss.pixls.us/t/syqon-starless-and-pytorch-install-failure/56239</a></p>
</blockquote>



<p class="wp-block-paragraph">以下をSirilのPython script editor にペースト、-&gt; run</p>



<pre class="wp-block-code"><code>import sirilpy as s
th = s.TorchHelper()
th.install_torch(cuda_version='cu128')
th.test_torch()</code></pre>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph"># cu126 の場合<br>cuda_version=&#8217;cu126&#8242;<br><a rel="noopener" target="_blank" href="https://discuss.pytorch.org/t/when-i-install-torch-2-6-0-with-whl-cu126-none-of-cuda-dependencies-get-installed-i-have-cu126-in-the-environment-already-is-this-expected/218014">https://discuss.pytorch.org/t/when-i-install-torch-2-6-0-with-whl-cu126-none-of-cuda-dependencies-get-installed-i-have-cu126-in-the-environment-already-is-this-expected/218014</a><br><br><a rel="noopener" target="_blank" href="https://techjacksolutions.com/ai-tools/pytorch/how-to-install-pytorch/">https://techjacksolutions.com/ai-tools/pytorch/how-to-install-pytorch/</a><br></p>
</blockquote>



<pre class="wp-block-code"><code># インストール　ログ
16:59:15: &#91;notice] A new release of pip is available: 25.0.1 -&gt; 26.2
16:59:15: &#91;notice] To update, run: pip install --upgrade pip
16:59:15: PyTorch installation completed successfully
16:59:15: Torch is already installed
16:59:15: === PyTorch GPU Test ===
16:59:18: OK: PyTorch 2.13.0+cu130 imported successfully!
16:59:18: === Device Information ===
16:59:18: PyTorch version: 2.13.0+cu130
16:59:18: CUDA available: True
16:59:18: CUDA version: 13.0
16:59:18: GPU count: 1
16:59:18:   GPU 0: NVIDIA GeForce RTX 5060 Laptop GPU
16:59:18: Creating model and test data...
16:59:18: Model parameters: 459,904
16:59:18: Input shape: torch.Size(&#91;64, 256])
16:59:18: Testing CPU execution (for accuracy reference)...
16:59:18: Using CUDA (NVidia / AMD)
16:59:18: Testing GPU execution...
16:59:19: OK: GPU execution successful!
16:59:19: OK: CPU and GPU outputs match within tolerance!
16:59:19: ==================================================
16:59:19: === Tensor Operations Test ===
16:59:19: OK: PyTorch 2.13.0+cu130 imported successfully!
16:59:19: Creating 2048x2048 tensors...
16:59:19: OK: Results match within tolerance!
16:59:19: CPU time: 0.0343s
16:59:19: GPU time: 0.0020s
16:59:19: Speedup: 17.51x
16:59:19: ==================================================
16:59:20: Python communication worker finished...</code></pre>



<h3 class="wp-block-heading">コマンドラインからPyTorchをインストール</h3>



<pre class="wp-block-code"><code># 参考
# 1. 既存の torch パッケージを完全にアンインストール
# pip3 uninstall torch torchvision torchaudio -y

# 2. CUDA 12.8 専用インデックスを指定してクリーンインストール
# pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128</code></pre>



<pre class="wp-block-code"><code>python3.12 -m venv ~/siril_cosmic_env
source ~/siril_cosmic_env/bin/activate
pip install --upgrade pip
pip install torch torchvision torchaudio --index-url https://pytorch.org</code></pre>



<p class="wp-block-paragraph">Pytorch バージョン確認</p>



<pre class="wp-block-code"><code>### Siril 1.5 仮想環境で行う
python3.12 -c "import torch; print(torch.__version__)"

2.11.0+cu128</code></pre>



<h3 class="wp-block-heading">onnxruntime-gpu,  NVIDIA CUDA</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">onnxruntime-gpu,  NVIDIA CUDA<br><a rel="noopener" target="_blank" href="https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-or-tensorrt">https://onnxruntime.ai/docs/install/#install-onnx-runtime-gpu-cuda-or-tensorrt</a></p>
</blockquote>



<p class="wp-block-paragraph">デフォルトのインストールは、pip install onnxruntime-gpu</p>



<pre class="wp-block-code"><code># CUDA 12.x
# The default CUDA version for onnxruntime-gpu in pypi is 12.x since 1.19.0.
# pip install onnxruntime-gpu
</code></pre>



<h4 class="wp-block-heading">Siril仮想環境へ、onnxruntime-gpuパッケージをインストール</h4>



<pre class="wp-block-code"><code># Siril 仮想環境で行う
source /home/$USER/.local/share/siril/venv/bin/activate

pip3 uninstall -y onnxruntime onnxruntime-gpu
pip3 install onnxruntime-gpu --upgrade

# onnxruntime-gpu==1.26.0 を指定
python3.12 -m pip install onnxruntime-gpu==1.26.0
Successfully installed flatbuffers-25.12.19 onnxruntime-gpu-1.26.0 packaging-26.3 protobuf-7.35.1</code></pre>



<pre class="wp-block-code"><code># 動作テスト
python3.12 -c "import onnxruntime as ort; print(ort.get_available_providers())"</code></pre>



<pre class="wp-block-code"><code>#インストールログ
Collecting onnxruntime-gpu
  Downloading onnxruntime_gpu-1.28.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl.metadata (5.8 kB)
Collecting flatbuffers (from onnxruntime-gpu)
  Downloading flatbuffers-25.12.19-py2.py3-none-any.whl.metadata (1.0 kB)
Requirement already satisfied: numpy&gt;=1.21.6 in ./lib/python3.12/site-packages (from onnxruntime-gpu) (2.5.1)
Requirement already satisfied: packaging in ./lib/python3.12/site-packages (from onnxruntime-gpu) (26.2)
Requirement already satisfied: protobuf&gt;=4.25.8 in ./lib/python3.12/site-packages (from onnxruntime-gpu) (7.35.1)
Downloading onnxruntime_gpu-1.28.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl (249.8 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 249.8/249.8 MB 8.1 MB/s eta 0:00:00
Downloading flatbuffers-25.12.19-py2.py3-none-any.whl (26 kB)
Installing collected packages: flatbuffers, onnxruntime-gpu
Successfully installed flatbuffers-25.12.19 onnxruntime-gpu-1.28.0

&#91;notice] A new release of pip is available: 25.0.1 -&gt; 26.2
&#91;notice] To update, run: pip install --upgrade pip</code></pre>



<h4 class="wp-block-heading">Siril仮想環境へ CUDA パスを固定する</h4>



<p class="wp-block-paragraph">/home/$USER/.local/share/siril/venv/bin/activate<br>ここに、CUDA パスと環境変数を追記</p>



<pre class="wp-block-code"><code>echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/lib/nvidia-580:${LD_LIBRARY_PATH}' &gt;&gt; /home/$USER/.local/share/siril/venv/bin/activate
echo 'export CUDA_MODULE_LOADING=LAZY' &gt;&gt; /home/$USER/.local/share/siril/venv/bin/activate</code></pre>



<p class="wp-block-paragraph">アクティブ中ならば、アクティブを解除してから、再度アクティブし、Sirilを起動させCosmic Clarityを動かしてみる。</p>



<pre class="wp-block-code"><code>deactivate
cd /home/$USER/.local/share/siril/venv
source ./bin/activate
siril</code></pre>



<h2 class="wp-block-heading">PyTorch と Siril</h2>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph"><br>こちらの環境では、torch-2.11.0+cu128　をインストールしています。<br>ログでは、torch-2.13.0+cu130をインストールを表示しています。</p>
</div>



<p class="wp-block-paragraph">Siril Pythonスクリプトにある、&#8221;GPU Firmware Manager&#8221;を使って、cuda 13対応をインストール</p>



<pre class="wp-block-code"><code>Successfully installed cuda-bindings-13.0.3 cuda-toolkit-13.0.3.0 nvidia-cublas-13.1.1.3 nvidia-cuda-cupti-13.0.85 nvidia-cuda-nvrtc-13.0.88 nvidia-cuda-runtime-13.0.96 nvidia-cudnn-cu13-9.20.0.48 nvidia-cufft-12.0.0.61 nvidia-cufile-1.15.1.6 nvidia-curand-10.4.0.35 nvidia-cusolver-12.0.4.66 nvidia-cusparse-12.6.3.3 nvidia-cusparselt-cu13-0.8.1 nvidia-nccl-cu13-2.29.7 nvidia-nvjitlink-13.2.78 nvidia-nvshmem-cu13-3.4.5 nvidia-nvtx-13.0.85 torch-2.13.0+cu130 torchvision-0.28.0+cu130 triton-3.7.1

PyTorch installation completed successfully
✓ Successfully installed torch+cu130 (CUDA 13.0 - GTX 16xx-RTX 50xx)
Running post-installation tests...

============================================================
Running PyTorch tests...
Torch is already installed
=== PyTorch GPU Test ===
OK: PyTorch 2.13.0+cu130 imported successfully!
=== Device Information ===
PyTorch version: 2.13.0+cu130
CUDA available: True
CUDA version: 13.0
GPU count: 1
  GPU 0: NVIDIA GeForce RTX 5060 Laptop GPU
Creating model and test data...
Model parameters: 459,904
Input shape: torch.Size(&#91;64, 256])
Testing CPU execution (for accuracy reference)...
Using CUDA (NVidia / AMD)
Testing GPU execution...
OK: GPU execution successful!
OK: CPU and GPU outputs match within tolerance!
==================================================
=== Tensor Operations Test ===
OK: PyTorch 2.13.0+cu130 imported successfully!
Creating 2048x2048 tensors...
OK: Results match within tolerance!
CPU time: 0.0343s
GPU time: 0.0020s
Speedup: 16.82x
==================================================
✓ All PyTorch tests passed!</code></pre>



<pre class="wp-block-code"><code># 画面表示のバックエンドとして従来のX11（XCB）を使用するよう強制する環境変数
QT_QPA_PLATFORM=xcb siril

# Wayland環境でSirilを実行する際、Qtプラグインのエラーを防ぐためには、QT_QPA_PLATFORM 環境変数を使用
QT_QPA_PLATFORM=wayland siril</code></pre>



<h2 class="wp-block-heading">Siril をソースからビルド / Ubuntu</h2>



<pre class="wp-block-code"><code># Siril インストール後に、Siril仮想環境を確認
less /home/${USER}/.local/share/siril/venv/pyvenv.cfg

home = /usr/local/bin
include-system-site-packages = false
version = 3.12.13
executable = /usr/local/bin/python3.12
command = /usr/local/bin/python3.12 -m venv /home/minmin/.local/share/siril/venv
#↑ビルドしたPythonを用いていることを確認します

### 3.14.6の場合
### home = /usr/local/bin
### include-system-site-packages = false
### version = 3.14.6
### executable = /usr/local/bin/python3.14
### command = /usr/local/bin/python3 -m venv /home/minmin/.local/share/siril/venv</code></pre>



<h3 class="wp-block-heading">Ubuntu 26.04 で、Sirilビルドに必要なパッケージをインストール</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Siril 1.5 ビルド案内<br><a rel="noopener" target="_blank" href="https://gitlab.com/free-astro/siril">https://gitlab.com/free-astro/siril</a></p>
</blockquote>


<a target="_self" href="https://globe3.ddns.net/wp/try-siril-development-edition-1-4-0/" title="Siril 開発版を試してみよう" class="blogcard-wrap internal-blogcard-wrap a-wrap cf"><div class="blogcard internal-blogcard ib-left cf"><div class="blogcard-label internal-blogcard-label"><span class="fa"></span></div><figure class="blogcard-thumbnail internal-blogcard-thumbnail"><img loading="lazy" decoding="async" width="160" height="90" src="https://globe3.ddns.net/wp/wp-content/uploads/2025/04/Siril-1.4.0-beta1-160x90.jpg" class="blogcard-thumb-image internal-blogcard-thumb-image wp-post-image" alt="" srcset="https://globe3.ddns.net/wp/wp-content/uploads/2025/04/Siril-1.4.0-beta1-160x90.jpg 160w, https://globe3.ddns.net/wp/wp-content/uploads/2025/04/Siril-1.4.0-beta1-300x169.jpg 300w, https://globe3.ddns.net/wp/wp-content/uploads/2025/04/Siril-1.4.0-beta1-120x68.jpg 120w, https://globe3.ddns.net/wp/wp-content/uploads/2025/04/Siril-1.4.0-beta1-320x180.jpg 320w, https://globe3.ddns.net/wp/wp-content/uploads/2025/04/Siril-1.4.0-beta1.jpg 702w" sizes="(max-width: 160px) 100vw, 160px" /></figure><div class="blogcard-content internal-blogcard-content"><div class="blogcard-title internal-blogcard-title">Siril 開発版を試してみよう</div><div class="blogcard-snippet internal-blogcard-snippet">Siril開発版が使えます。SPCC、GHS、モザイクに有用なプレートソルビングの歪み補正が使えます。</div></div><div class="blogcard-footer internal-blogcard-footer cf"><div class="blogcard-site internal-blogcard-site"><div class="blogcard-favicon internal-blogcard-favicon"><img loading="lazy" decoding="async" src="https://www.google.com/s2/favicons?domain=https://globe3.ddns.net/wp" alt="" class="blogcard-favicon-image internal-blogcard-favicon-image" width="16" height="16" /></div><div class="blogcard-domain internal-blogcard-domain">globe3.ddns.net</div></div><div class="blogcard-date internal-blogcard-date"><div class="blogcard-post-date internal-blogcard-post-date">2025.04.18</div></div></div></div></a>


<pre class="wp-block-code"><code># 2026年7月　追加
sudo apt install libgtk-4-dev libgtksourceview-5-dev

### sudo apt install wcslib-tools libgtk-4-dev libgtksourceview-5-dev kmplot \
  libsqlite3-dev</code></pre>



<h4 class="wp-block-heading">OpenCV5</h4>



<h3 class="wp-block-heading">meson ビルドオプション</h3>



<pre class="wp-block-code"><code>meson setup --prefix /usr/local --buildtype release _build
ninja -C _build
ninja -C _build install</code></pre>



<pre class="wp-block-code"><code>meson setup _build --buildtype release -Dgtk=true
cd _build
ninja
ninja install</code></pre>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph"><strong><code>-prefix /usr/local</code></strong>: <code>meson install</code> を実行した際、ビルドされたファイルが格納されるディレクトリ</p>



<p class="wp-block-paragraph"><strong><code>--buildtype release</code></strong>: デバッグシンボルを削除し、コンパイラの最適化（最適化レベル <code>-O3</code> など）を有効にしてビルド</p>



<p class="wp-block-paragraph"><strong><code>_build</code></strong>: 設定ファイルや中間ファイルを出力するディレクトリ</p>
</div>



<pre class="wp-block-code"><code>meson setup --buildtype=release builddir -Dcpp_args="-march=native" -Dc_args="-march=native"</code></pre>



<p class="wp-block-paragraph">ん〜、有効にならない？</p>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph"><code><strong>-march=native</strong></code>: コンパイルを実行しているマシンに搭載されているCPUが持つ全機能を有効にします</p>



<p class="wp-block-paragraph"><code><strong>-mtune=native</strong></code>: コードの命令スケジュールを特定のCPU向けに最適化します（生成されたバイナリは他の機種でも動作します）。</p>



<p class="wp-block-paragraph"><code><strong>-O3</strong></code>: コンパイラが可能な限りの高度な最適化を行います（<code>-O2</code> がデフォルトですが、速度を追求する場合に使用します）</p>
</div>



<pre class="wp-block-code"><code>meson setup --prefix /usr/local --buildtype release _build -Dcpp_args="-march=native" -Dc_args="-march=native"</code></pre>



<pre class="wp-block-code"><code>meson setup --prefix /usr/local --buildtype release _build -Dcpp_args="-march=native" -Dfortran_args="-march=native"</code></pre>



<pre class="wp-block-code"><code>meson setup --prefix /usr/local --buildtype release _build -Dcpp_args="-march=native -O3" -Dfortran_args="-march=native -O3"</code></pre>



<p class="wp-block-paragraph">ん〜、有効にならない？</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Meson <br><a rel="noopener" target="_blank" href="https://mesonbuild.com/Commands.html#configure">https://mesonbuild.com/Commands.html#configure</a></p>
</blockquote>



<h2 class="wp-block-heading">Siril Qtでの不具合</h2>



<p class="wp-block-paragraph">Sirilに関しては致命的なエラーではないようです。X11を使用する場合には、XWaylandを用いるようです。</p>



<p class="wp-block-paragraph">Ubuntu 26.04で採用のGNOME 48は、Wayland上での動作を前提に最適化されている。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Qt 5 で Wayland のサポートを有効化するには、qt5-wayland パッケージをインストールしてください。これで、Qt 5 アプリケーションは Wayland セッション上では Wayland で動作するようになります。</p>



<p class="wp-block-paragraph">必須ではないはずですが、明示的に Wayland プラグインを使って Qt 5 アプリケーションを動作させるには、[7] にあるように、-platform wayland を使用するか、環境変数 QT_QPA_PLATFORM=wayland をセットして下さい。</p>



<p class="wp-block-paragraph">一方、Wayland セッション上で X11 を使用させるには、<br>QT_QPA_PLATFORM=xcb と設定してください。<br>これは、システムの Qt の実装を使用しない一部のプロプライエタリアプリケーションで必要になる場合があります。また、QT_QPA_PLATFORM=&#8221;wayland;xcb&#8221; を使用すると、Qt が Wayland を利用できない場合に、代わりに xcb (X11)プラグインを使用させることができます。<br><br><a rel="noopener" target="_blank" href="https://wiki.archlinux.jp/index.php/Wayland">https://wiki.archlinux.jp/index.php/Wayland</a></p>
</blockquote>



<h2 class="wp-block-heading">Seti Astro Cosmic Clarity linux</h2>



<p class="wp-block-paragraph">SetiAstroCosmicClarity”, “SetiAstroCosmicClarity_denoise” をデコンパイルする方法をこちらに記載しています。</p>



<p class="wp-block-paragraph">SetiAstroCosmicClarity”, “SetiAstroCosmicClarity_denoise” それぞれをpyinstxtractor.py を使って展開し動作確認、後に、ラッパーを作成し、Siril 1.5から動かします。</p>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">こちらでは公開されている、pythonコードをラッパーから利用しています。</p>


<a target="_self" href="https://globe3.ddns.net/wp/cudatoolkit-12-8-2runfile-ubuntu-2604-nvidiadriver580-python3-12-13-graxpert-3-1-cosmicclaritysuite-siril-15/#toc20" title="CUDA Toolkit 12.8.2 (April 2026)runfile +  Ubuntu 26.04 + nvidia-driver 580 + python version 3.12.13  + PyTorch + GraXpert + Cosmic Clarity Suite + Siril 1.5" class="blogcard-wrap internal-blogcard-wrap a-wrap cf"><div class="blogcard internal-blogcard ib-left cf"><div class="blogcard-label internal-blogcard-label"><span class="fa"></span></div><figure class="blogcard-thumbnail internal-blogcard-thumbnail"><img loading="lazy" decoding="async" width="320" height="180" src="https://globe3.ddns.net/wp/wp-content/uploads/2026/08/Ryzen7_350-320x180.jpg" class="blogcard-thumb-image internal-blogcard-thumb-image wp-post-image" alt="" srcset="https://globe3.ddns.net/wp/wp-content/uploads/2026/08/Ryzen7_350-320x180.jpg 320w, https://globe3.ddns.net/wp/wp-content/uploads/2026/08/Ryzen7_350-240x135.jpg 240w, https://globe3.ddns.net/wp/wp-content/uploads/2026/08/Ryzen7_350-640x360.jpg 640w" sizes="(max-width: 320px) 100vw, 320px" /></figure><div class="blogcard-content internal-blogcard-content"><div class="blogcard-title internal-blogcard-title">CUDA Toolkit 12.8.2 (April 2026)runfile +  Ubuntu 26.04 + nvidia-driver 580 + python version 3.12.13  + PyTorch + GraXpert + Cosmic Clarity Suite + Siril 1.5</div><div class="blogcard-snippet internal-blogcard-snippet">CUDA Toolkit 12.8.2を用いて、Ubuntu + CUDA環境を作ります。</div></div><div class="blogcard-footer internal-blogcard-footer cf"><div class="blogcard-site internal-blogcard-site"><div class="blogcard-favicon internal-blogcard-favicon"><img loading="lazy" decoding="async" src="https://www.google.com/s2/favicons?domain=https://globe3.ddns.net/wp" alt="" class="blogcard-favicon-image internal-blogcard-favicon-image" width="16" height="16" /></div><div class="blogcard-domain internal-blogcard-domain">globe3.ddns.net</div></div><div class="blogcard-date internal-blogcard-date"><div class="blogcard-post-date internal-blogcard-post-date">2026.08.05</div></div></div></div></a></div>



<h3 class="wp-block-heading">pyinstxtractor.py を使って、SetiAstroCosmicClarity を展開</h3>



<pre class="wp-block-code"><code># Siril 1.5 仮想環境で行う
pwd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild
python3.12 pyinstxtractor.py ./SetiAstroCosmicClarity

&#91;+] Processing ./SetiAstroCosmicClarity
&#91;+] Pyinstaller version: 2.1+
&#91;+] Python version: 3.12
&#91;+] Length of package: 57264312 bytes
&#91;+] Found 17 files in CArchive
&#91;+] Beginning extraction...please standby
&#91;+] Possible entry point: pyiboot01_bootstrap.pyc
&#91;+] Possible entry point: pyi_rth_inspect.pyc
&#91;+] Possible entry point: pyi_rth_pkgutil.pyc
&#91;+] Possible entry point: pyi_rth_multiprocessing.pyc
&#91;+] Possible entry point: pyi_rth_pkgres.pyc
&#91;+] Possible entry point: pyi_rth_setuptools.pyc
&#91;+] Possible entry point: pyi_rth_mplconfig.pyc
&#91;+] Possible entry point: pyi_rth_cryptography_openssl.pyc
&#91;+] Possible entry point: pyi_rth_pyqt6.pyc
&#91;+] Possible entry point: pyi_rth__tkinter.pyc
&#91;+] Possible entry point: setiastrocosmicclaritylinux.pyc
&#91;+] Found 7078 files in PYZ archive
&#91;+] Successfully extracted pyinstaller archive: ./SetiAstroCosmicClarity

You can now use a python decompiler on the pyc files within the extracted directory</code></pre>



<p class="wp-block-paragraph">SetiAstroCosmicClarity_extracted ディレクトリにある、<br>setiastrocosmicclaritylinux.pyc を<br>CosmicClaritySuite_Linux_AlternateBuild ディレクトリに配置</p>



<pre class="wp-block-code"><code># 動作確認
python3.12 ./setiastrocosmicclaritylinux.pyc

Using device: cuda (CUDA)
Identified 32bit floating point FITS iamge.
Sharpening Image: result_drizzle_19620s.fit
Total Chunks: 225
Stellar Sharpening:
Progress: 100.00% (225/225 chunks processed)
Stellar sharpening complete. Shape: (2877, 2877)
Saved 32-bit float sharpened image to: /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/./output/result_drizzle_19620s_sharpened.fit</code></pre>



<h4 class="wp-block-heading">Cosmic Clarity Sharpening ラッパーを作成</h4>



<pre class="wp-block-code"><code>### ラッパーファイルを作成 run_Sharpen.sh
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/</code></pre>



<p class="wp-block-paragraph">以下が、run_Sharpen.sh</p>



<pre class="wp-block-code"><code>#!/bin/bash
source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/
python3.12 ./setiastrocosmicclaritylinux.pyc "$@"

# 実行権
chmod +x ./run_Sharpen.sh</code></pre>



<p class="wp-block-paragraph">シェルスクリプトにおける “$@” は、<br>スクリプトに渡されたすべての引数を、元の区切り（スペースなど）を完全に保ったまま展開する特殊な変数</p>



<h4 class="wp-block-heading">Cosmic Clarity Sharpening を設定</h4>



<p class="wp-block-paragraph">Cosmic Clarity Sharpening Settings<br>Cosmic Clarity Executable:  run_Sharpen.sh (上記ラッパー)を指定する。</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h4 class="wp-block-heading">RTX 5060 動作確認</h4>



<pre class="wp-block-code"><code>source /home/$USER/.local/share/siril/venv/bin/activate
pip list |grep torch
torch                    2.11.0+cu128
torchaudio               2.11.0+cu128
torchvision              0.26.0+cu128

pip list |grep torch-cuda80
pip list |grep torch-model-archiver
pip list |grep torchrec-cpu</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 -c "import torch; print(torch.__version__, torch.cuda.is_available(), torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"
2.11.0+cu128 True NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 -c "import torch; print('Version:', torch.__version__); print('CUDA:', torch.cuda.is_available()); print('Device Name:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"
Version: 2.11.0+cu128
CUDA: True
Device Name: NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<h3 class="wp-block-heading">Cosmic Clarity denoise を展開</h3>



<p class="wp-block-paragraph">python3.12 pyinstxtractor.py /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/SetiAstroCosmicClarity_denoise</p>



<pre class="wp-block-code"><code># Siril 1.5 仮想環境で行う
pwd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild
python3.12 pyinstxtractor.py ./SetiAstroCosmicClarity_denoise</code></pre>



<pre class="wp-block-code"><code># 動作確認
python3.12 ./setiastrocosmicclaritylinux.pyc
</code></pre>



<h4 class="wp-block-heading">Cosmic Clarity denoise ラッパーを作成</h4>



<pre class="wp-block-code"><code>### ラッパーファイルを作成 run_Denoise.sh
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/</code></pre>



<p class="wp-block-paragraph">以下が、run_Denoise.sh</p>



<pre class="wp-block-code"><code>#!/bin/bash

### RTX 5060（12.0）
### export TORCH_CUDA_ARCH_LIST="9.0"
### export TORCH_CUDA_ARCH_LIST="12.0"

source "$HOME/.local/share/siril/venv/bin/activate"
cd /usr/local/share/CosmicClaritySuite_Linux_AlternateBuild/
python3.12 ./SetiAstroCosmicClarity_denoise.pyc "$@"</code></pre>



<h4 class="wp-block-heading">Cosmic Clarity Denoise を設定</h4>



<p class="wp-block-paragraph">Cosmic Clarity Executable<br>run_Sharpen.sh (上記ラッパー)を指定する。</p>



<h2 class="wp-block-heading">StarNet2 2.5.3</h2>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph"><a rel="noopener" target="_blank" href="https://www.williamliphotos.com/pages/cuda-starnet">https://www.williamliphotos.com/pages/cuda-starnet</a><br><br><a rel="noopener" target="_blank" href="https://blog.0xbadc0de.be/archives/464">https://blog.0xbadc0de.be/archives/464</a></p>
</blockquote>



<p class="wp-block-paragraph"></p>



<h2 class="wp-block-heading">SyQon</h2>



<p class="wp-block-paragraph">syqon.it → syqon.eu<br>スクリプトを書き換えても動きません。モジュールを別途配置して動かしています。</p>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph">Revert to a Previous Script VersionIf you are downloading scripts natively through Siril&#8217;s interface, the automated script updates might be linking to the down server.Go to the Get Scripts drop-down menu in Siril.Right-click on the problematic script (e.g., Starless or Prism Mini).Select Go back 1 version.Save the rolled-back version into your designated local scripts folder and try running it again.</p>
</div>



<h2 class="wp-block-heading">Windows11</h2>



<p class="wp-block-paragraph">ウェッブを辿っていくと、それぞれの特徴がまとめられていて、理解の手助けになります。有難うございます。</p>



<p class="wp-block-paragraph">CUDAツールキット、NVIDIA CUDA Toolkit、cuDNN</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">こちらを参考にしました。有難うございます。<br><a rel="noopener" target="_blank" href="https://docs.ailia.ai/setup/cuda/">https://docs.ailia.ai/setup/cuda/</a></p>
</blockquote>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">以下の設定は、Windows11, Siril 1.5 環境で確認しています：</p>



<p class="wp-block-paragraph">・オペレーティングシステム : Windows 11<br>・GPU およびドライバー : NVIDIA GeForce RTX 5060、NVIDIA ドライバー 610.88 | WHQL<br>　nvidia-smi によって報告された CUDA 機能レベル : 13.3<br>・CUDA Toolkit : CUDA Toolkit 12.8.2 (April 2026). nvcc &#8211;version の出力は CUDA XX.x / VXX.x.x <br>・cuDNN : CUDA 12.8 対応の cuDNN 9.24.0<br>・ONNX Runtime GPU : Windows x64 対応の ONNX Runtime GPU 1.26.0</p>
</div>



<h3 class="wp-block-heading">サインインを回避</h3>



<p class="wp-block-paragraph">サインインの画⾯が表⽰されたら、Shift + F10 を同時に押します。</p>



<pre class="wp-block-code"><code>コマンドプロンプトに、
start ms-cxh:localonly</code></pre>



<h3 class="wp-block-heading">Windows11 「落ち穂ひろい」、Windows 開発者モード</h3>



<h4 class="wp-block-heading">Line</h4>



<p class="wp-block-paragraph">C:\Users\USERNAME\AppData\Local\LINE<br>ディレクトリを丸ごとコピー、Lineを再インストール。トーク履歴が反映される。<br></p>



<h4 class="wp-block-heading">Caps と Ctrl 入れ替え</h4>



<p class="wp-block-paragraph">Ctrl2Cap v3.0を使ってみます。<br>このツールは両キーを入れ替えるのではなく、CapsLockキーをCtrlに変更するツールらしく、CapsLockは存在しなくなるようです。ん、その様です。</p>



<p class="wp-block-paragraph">コマンドプロンプトを管理者として実行し、Ctrl2Cap ファイルを展開したディレクトリからコマンド ctrl2cap /install を実行してインストールします。 アンインストールは、ctrl2cap /uninstall</p>



<h4 class="wp-block-heading">Windows 開発者モード</h4>



<p class="wp-block-paragraph">Siril は天体画像の前処理で、シンボリックリンクを多く使います。<br>管理者権限なしで、シンボリックリンクを扱えるように、「開発者モード」を有効にすると良いらしいです。</p>



<p class="wp-block-paragraph">開発者向けの設定、開発者モードを有効にする<br><a rel="noopener" target="_blank" href="https://learn.microsoft.com/ja-jp/windows/advanced-settings/developer-mode">https://learn.microsoft.com/ja-jp/windows/advanced-settings/developer-mode</a></p>



<h3 class="wp-block-heading">Python 3 バージョン</h3>



<p class="wp-block-paragraph">Pythonバージョンはこちらを参照します。<br><a rel="noopener" target="_blank" href="https://devguide.python.org/versions/#versions">https://devguide.python.org/versions/#versions</a></p>



<p class="wp-block-paragraph">インストール画面の最初にある &#8220;Add python.exe to PATH&#8221;（環境変数への追加）に必ずチェック</p>



<p class="wp-block-paragraph">Python 3.12.10, Release date: April 8, 2025<br><a rel="noopener" target="_blank" href="https://www.python.org/downloads/release/python-31210/">https://www.python.org/downloads/release/python-31210/</a></p>



<p class="wp-block-paragraph">Windows installer (64-bit)を展開、インストール<br>(Python 3.12.14 exeはあるかな？)</p>



<h3 class="wp-block-heading">NVIDIA ドライバのインストール</h3>



<h4 class="wp-block-heading">Windows11 NVIDIA グラフィックスボードを確認</h4>



<p class="wp-block-paragraph">Windows 11の検索に、&#8221;dxdiag&#8221; を入力<br>DirectX 診断ツールが起動し、「システム」「ディスプレイ」等が表示されます。</p>



<p class="wp-block-paragraph">もしくは、Power Shell</p>



<pre class="wp-block-code"><code>nvidia-smi
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 616.56                 KMD Version: 616.56        CUDA UMD Version: 13.4   | 
+---------------------------------+------------------------+----------------------------+</code></pre>



<h4 class="wp-block-heading">NVIDIA Studio ドライバー？ GeForce Game Ready ドライバー？</h4>



<p class="wp-block-paragraph">NVIDIA Studio　2026年8月現在</p>



<p class="wp-block-paragraph">Driver Version: 616.56 | WHQL, Release Date: Wed Aug 26, 2026<br>OS: Windows 10 64-bit,　Windows 11<br>言語: Japanese　ファイル サイズ: 984.32 MB</p>



<h4 class="wp-block-heading">Siril Pytonスクリプトが GPUを使えるように、Nvidia を設定</h4>



<p class="wp-block-paragraph">CPU内蔵グラフィックスが画面表示を行い、NVIDIA GeForceは計算ユニットとして使います。</p>



<p class="wp-block-paragraph">BIOSでは 「MSHybrid」 または 「Optimus」（モデルによっては 「IGPU」 や 「Hybrid」）を指定<br>　画面表示はCPU内蔵グラフィックスが行う。<br>NVIDIAコントロールパネルを開く<br>　NVIDIAコントロールパネル：左メニューの 「3D設定の管理」 をクリック<br>　NVIDIA App：NVIDIA Appの「グラフィックス」→「プログラム設定」<br>　画面中央の 「グローバル設定」 タブが開いていることを確認<br>　項目一覧から 「OpenGL レンダリング GPU」 をクリック<br>　設定値を「自動選択」から GeForce RTX 5060 に変更<br>3D設定の管理の一覧から 「CUDA &#8211; GPU」をクリック<br>設定値が 「すべて」 になっていることを確認</p>



<p class="wp-block-paragraph">Windows 11 の「グラフィック設定」を行う。Windows 11では、システムのグラフィック設定で「高パフォーマンス」に指定されたアプリが優先してNVIDIA GPUに割り当てられます。<br>　スタートメニューから 「設定」（歯車マーク）を開く。<br>　「システム」 ＞ 「ディスプレイ」 ＞ 画面下部にある 「グラフィック」 をクリックする。<br>　アプリへ[Siril]を追加する。<br>　追加した[Siril]をクリックし、「オプション」 を開く。<br>　GPUのユーザー設定に、&#8221;高パフォーマンス（GPU: NVIDIA GeForce &#8230;)&#8221;を指定する。</p>



<p class="wp-block-paragraph">電源管理モード<br> 「標準」から 「パフォーマンス最大化を優先」 に変更<br>　最後に、画面右下の 「適用」 ボタンを必ずクリックして設定を保存</p>



<p class="wp-block-paragraph">ディスクトップの場合は、オンボードのHDMI出力に、ディスプレイケーブルを挿す。</p>



<pre class="wp-block-code"><code>ドライバファイルの保存先
C:\Windows\Temp\b8fa40a0-5b98-476e-ade9-3bd454ada083</code></pre>



<h3 class="wp-block-heading">NVIDIA CUDA Toolkit</h3>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">CUDA、Microsoft Windows向けインストールガイド<br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html">https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html</a></p>
</blockquote>



<p class="wp-block-paragraph">CUDA Toolkit 12.8 をインストールします。</p>



<p class="wp-block-paragraph">CUDA 13系は、ライブラリが対応していないかもしれないので、安定動作をめざすならば、使用しません。</p>



<p class="wp-block-paragraph">また、Starnet2の案内に、<br>『現在の ONNX Runtime 1.26.0 の GPU パスでは、CUDA 12.x を使用してください。現在の CUDA-12 ONNX Runtime GPU DLL では、CUDA 13.x ランタイム DLL を使用しないでください』、とあります。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">Windows GPU acceleration setup instructions<br><a rel="noopener" target="_blank" href="https://starnetastro.com/documentation/windows-cuda/">https://starnetastro.com/documentation/windows-cuda/</a></p>
</blockquote>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">CUDA 12.xを使います。<br>GPUドライバーは、ここではインストールしません。<br>カスタムインストールを選び、ドライバーの印を外します。</p>
</div>



<p class="wp-block-paragraph">CUDA Toolkit 12.8 Update 2 Downloads<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cuda-12-8-0-download-archive">https://developer.nvidia.com/cuda-12-8-0-download-archive</a></p>



<pre class="wp-block-code"><code>nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2025 NVIDIA Corporation
Built on Fri_Feb_21_20:42:46_Pacific_Standard_Time_2025
Cuda compilation tools, release 12.8, V12.8.93
Build cuda_12.8.r12.8/compiler.35583870_0</code></pre>



<h4 class="wp-block-heading">Build Tools for Visual Studio 2026</h4>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">サポートされるプロジェクト: ASP.NET、Azure、C++ デスクトップ、ClickOnce、コンテナー、.NET Core、.NET Desktop、Node.js、Office と SharePoint、Python、TypeScript、単体テスト、UWP、WCF、Xamarin。 このツールを使用するには、プロジェクトのオープン ソースの依存関係をビルドしている場合を除き、有効な Visual Studio ライセンスが必要です。<br><a rel="noopener" target="_blank" href="https://visualstudio.microsoft.com/ja/downloads/">https://visualstudio.microsoft.com/ja/downloads/</a></p>
</blockquote>



<p class="wp-block-paragraph">Build Tools for Visual Studio 2026 をインストールします。<br>C/C++のコンパイラ機能のみを導入します。[C++によるデスクトップ開発]をチェックします。<br>18.9.2では、6.68GBらしいです。</p>



<p class="wp-block-paragraph">C:\Program Files (x86)\Microsoft Visual Studio\18\BuildTools<br>このディレクトリへインストールされます。</p>



<h3 class="wp-block-heading">NVIDIA cuDNN</h3>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph">StarNet2をRTX5060で動かします。その為には、ONNX RuntimeのCUDA Execution Providerでは、現在のCUDA 12.x系に対応した cuDNN 9.x を使います。</p>
</div>



<p class="wp-block-paragraph">cuDNN Archive<br><a rel="noopener" target="_blank" href="https://developer.nvidia.com/cudnn-archive">https://developer.nvidia.com/cudnn-archive</a></p>



<p class="wp-block-paragraph">cuDNN 9.25.0 が2026年7月にリリースされました。が、cuDNN 9.24.0 をインストールします。<br>cuDNNアーカイブから、&#8221;cuDNN 9.24.0 (July 2026)&#8221; を導入します。</p>



<p class="wp-block-paragraph">Cosmic Clarityのスタンドアロン版やSirilのPython連携（SetiAstroSuiteなど）は、内部に必要なライブラリを内蔵（Frozen Executables）しているため、cuDNNをインストールする必要はないらしい。</p>



<p class="wp-block-paragraph">もしも、ソフト起動時に「cudnn64_8.dll または cudnn64_9.dll が見つかりません」といったエラーが出る場合は、NVIDIA cuDNNページから対応するv9.xのZipをダウンロードしインストールします。<br>（中身をCUDAのインストール先 [C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.x] の各フォルダに上書きコピーします。）</p>



<pre class="wp-block-code"><code>C:\Program Files\NVIDIA\CUDNN\v9.24</code></pre>



<pre class="wp-block-code"><code>C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64</code></pre>



<pre class="wp-block-code"><code>findstr "CUDNN_M CUDNN_P" "C:\Program Files\NVIDIA\CUDNN\v9.24\include\12.9\cudnn_version.h"</code></pre>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">ここでは、CUDA Toolkit 12.8 Update 2を使うので、<br>cuDNN for CUDA cuda 12.9をカスタムインストールで選びます。</p>
</div>



<h4 class="wp-block-heading">cuDNN パスを通す</h4>



<p class="wp-block-paragraph">cuDNNライブラリ (<code>cudnn64_*.dll</code>) にパスを通します。</p>



<p class="wp-block-paragraph">「設定」→「システム」→ → 「関連リンク」 → 「システムの詳細設定」をクリックして「システムのプロパティ」を表示します。「環境変数」ボタンをクリックします。</p>



<p class="wp-block-paragraph">もしくは、<br>スタートボタンを右クリックし、「システム」→「バージョン情報」→「システムの詳細設定」→「環境変数」を選びます。</p>



<p class="wp-block-paragraph"><strong>システム環境変数</strong>の「Path」を選択し、「編集」をクリックします。「新規」をクリックして、cuDNN の <code>bin</code> フォルダのパスを追加します。</p>



<p class="wp-block-paragraph">CUDA Toolkit 12.8を導入しているので、以下のパスを通します。</p>



<pre class="wp-block-code"><code>C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64</code></pre>



<pre class="wp-block-code"><code># コマンドプロンプトで、確認します
where cudnn64_*.dll
C:\Program Files\NVIDIA\CUDNN\v9.24\bin\12.9\x64\cudnn64_9.dll</code></pre>



<h2 class="wp-block-heading">ソースから Siril をビルド / Windows 11</h2>



<p class="wp-block-paragraph"><a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/installation/windows.html#building-on-windows-with-msys2">https://siril.readthedocs.io/en/latest/installation/windows.html#building-on-windows-with-msys2</a><br>こちらの案内に従い、MSYS2 64bit をインストールし、これを用いて、Siril 開発版を構築します。</p>



<p class="wp-block-paragraph">MSYS2<br>Software Distribution and Building Platform for Windows<br><a rel="noopener" target="_blank" href="https://www.msys2.org/">https://www.msys2.org/</a></p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">MSYS2　各シェルの違い<br><a rel="noopener" target="_blank" href="https://qiita.com/koyukitukimino/items/df76f7fdb8512920d75b">https://qiita.com/koyukitukimino/items/df76f7fdb8512920d75b</a><br><br>MSYS2 MSYS : Unix向けソフトウェアのビルドに使用する なし<br>MSYS2 MINGW64 : 64Bit Windowsアプリケーションをgccでビルドするための環境 あり<br>MSYS2 UCRT64 : Universal C Runtime（UCRT）を使用して64Bit Windowsアプリケーションをgccでビルドするための環境 あり<br>MSYS2 CLANG64 : 64Bit WindowsアプリケーションをClangでビルドするための環境 あり<br>MSYS2 CLANGARM64 : ARM64(AArch64)アプリケーションをClangでビルドするための環境 ARM64版Windowsで動作</p>
</blockquote>



<p class="wp-block-paragraph">MSYS2 MinGW UCRT x64 : こちらを用いて、Siril を構築していきます。<br>Universal C Runtime（UCRT）を使用して64Bit Windowsアプリケーションをgccでビルドします。</p>



<p class="wp-block-paragraph">パッケージデータベースとコアシステムパッケージを入力して更新します。</p>



<pre class="wp-block-code"><code>pacman -Syu</code></pre>



<p class="wp-block-paragraph">依存関係をインストールする</p>



<pre class="wp-block-code"><code>curl -s -o siril-deps.sh https://gitlab.com/free-astro/siril/-/raw/master/build/windows/native-gitlab-ci/siril-deps.sh

bash siril-deps.sh</code></pre>



<p class="wp-block-paragraph">curl が動かない。。。再インストールしよう。</p>



<pre class="wp-block-code"><code>pacman -S mingw-w64-ucrt-x86_64-curl</code></pre>



<p class="wp-block-paragraph">113のパッケージがインストールされました。</p>



<p class="wp-block-paragraph">Siril ソースコードをダウンロードします。</p>



<pre class="wp-block-code"><code>git clone https://gitlab.com/free-astro/siril.git
cd siril
# ヘッダーファイルなどをダウンロード
git submodule update --init</code></pre>



<p class="wp-block-paragraph">ビルドシステムを生成し、コードをコンパイルします。</p>



<pre class="wp-block-code"><code>meson setup _build --buildtype release -Dgtk=true</code></pre>



<p class="wp-block-paragraph">若しくは、</p>



<pre class="wp-block-code"><code>meson setup --buildtype release _build -Dgtk=true -Dcpp_args="-march=native" -Dc_args="-march=native"

meson setup --prefix /usr/local --buildtype release _build -Dgtk=true -Dcpp_args="-march=native" -Dc_args="-march=native"</code></pre>



<p class="wp-block-paragraph">あれ、arch_nativeが有効にならないな～。。。</p>



<p class="wp-block-paragraph">ビルド、インストールします。</p>



<pre class="wp-block-code"><code>ninja -C _build install</code></pre>



<p class="wp-block-paragraph">Python仮想環境の準備や、モジュールインストールのために、Windows環境からSirilを起動させます。</p>



<p class="wp-block-paragraph">エクスプローラーから、<br>C:\msys64\ucrt64\binにあるsirilを起動させます。</p>



<pre class="wp-block-code"><code>log: ローカルのSPCCデータベースリポジトリは最新です！
Entering wrapping_idle for function 00007ff7d8326f50
log: python仮想環境を準備しています: C:\Users\USERNAME\AppData\Local\siril\venv.
log: pythonモジュールが最新か確認しています...
log: バックグラウンドでpythonモジュールをインストール/更新します。これは数秒掛かるかもしれません...
log: Pythonモジュールは最新です</code></pre>



<p class="wp-block-paragraph">Siril 1.5 Python仮想環境である、C:\Users\USERNAME\AppData\Local\siril<br>このディレクトリで、仮想環境をアクティブにして、pipを用います。</p>



<h3 class="wp-block-heading">Siril 1.5 Python仮想環境</h3>



<p class="wp-block-paragraph">Siril 1.5 Python仮想環境を動かします。「アクティブ」にする。</p>



<p class="wp-block-paragraph">PowerShell :</p>



<pre class="wp-block-code"><code>. C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\Activate.ps1</code></pre>



<p class="wp-block-paragraph">コマンドプロンプト（cmd）:</p>



<pre class="wp-block-code"><code>C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\activate.bat</code></pre>



<p class="wp-block-paragraph">PowerShellでのエラー表示</p>



<pre class="wp-block-code"><code>このシステムではスクリプトの実行が無効になっているため、ファイル C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\A
ctivate.ps1を読み込むことができません。詳細については、「about_Execution_Policies」(https://go.microsoft.com/fwlink/?L
inkID=135170) を参照してください。
発生場所 行:1 文字:3
+ . C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\Activate.ps1
+   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    + CategoryInfo          : セキュリティ エラー: (: ) &#91;]、PSSecurityException
    + FullyQualifiedErrorId : UnauthorizedAccess</code></pre>



<p class="wp-block-paragraph">PowerShellのスクリプト実行許可レベルを変更する。</p>



<pre class="wp-block-code"><code>Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope Process</code></pre>



<ul class="wp-block-list">
<li>Set-ExecutionPolicy: PowerShellのスクリプト実行許可レベルを変更</li>



<li>-ExecutionPolicy Bypass: 警告や確認を出さず、すべてのスクリプトの実行を許可</li>



<li>-Scope Process: 変更の影響範囲を現在実行中のPowerShellプロセス（ウィンドウ）だけに限定</li>
</ul>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">例 6: 現在の PowerShell セッションの実行ポリシーを設定する<br><a rel="noopener" target="_blank" href="https://learn.microsoft.com/ja-jp/powershell/module/microsoft.powershell.security/set-executionpolicy?view=powershell-7.6#6-powershell">https://learn.microsoft.com/ja-jp/powershell/module/microsoft.powershell.security/set-executionpolicy?view=powershell-7.6#6-powershell</a></p>
</blockquote>



<h3 class="wp-block-heading">PyTorch インストール</h3>



<p class="wp-block-paragraph">Siril 1.5 仮想環境をアクティブにする。</p>



<pre class="wp-block-code"><code>Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope Process

. C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\Activate.ps1</code></pre>



<pre class="wp-block-code"><code>(venv) PS　#仮想環境アクティブ表示</code></pre>



<p class="wp-block-paragraph">コマンドを用いて、Siril Python仮想環境下へ、PyTorch, onnxruntime-gpuをインストールする。</p>



<pre class="wp-block-code"><code>pip 自身をインストール、最新版にアップデートする
python -m pip install --upgrade pip</code></pre>



<pre class="wp-block-code"><code>Downloading pip-26.2.1-py3-none-any.whl (1.8 MB)
   ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.8/1.8 MB 9.9 MB/s eta 0:00:00
Installing collected packages: pip
  Attempting uninstall: pip
    Found existing installation: pip 25.0.1
    Uninstalling pip-25.0.1:
      Successfully uninstalled pip-25.0.1
Successfully installed pip-26.2.1</code></pre>



<pre class="wp-block-code"><code>pip3 --version
pip 26.2.1 from C:\Users\USERNAME\AppData\Local\siril\venv\Lib\site-packages\pip (python 3.12)</code></pre>



<p class="wp-block-paragraph">PyTorchをインストール</p>



<pre class="wp-block-code"><code>python -m pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128</code></pre>



<pre class="wp-block-code"><code>Installing collected packages: torchaudio, mpmath, typing-extensions, sympy, setuptools, pillow, networkx, MarkupSafe, fsspec, filelock, jinja2, torch, torchvision
Successfully installed MarkupSafe-3.0.3 filelock-3.32.3 fsspec-2026.7.0 jinja2-3.1.6 mpmath-1.3.0 networkx-3.6.1 pillow-12.3.0 setuptools-78.1.0 sympy-1.14.0 torch-2.11.0+cu128 torchaudio-2.11.0+cu128 torchvision-0.26.0+cu128 typing-extensions-4.16.0</code></pre>



<p class="wp-block-paragraph">動作確認</p>



<pre class="wp-block-code"><code>python -c "import torch; print('PyTorch Version:', torch.__version__); print('CUDA Available:', torch.cuda.is_available()); print('GPU Device:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"

PyTorch Version: 2.11.0+cu128
CUDA Available: True
GPU Device: NVIDIA GeForce RTX 5060 Laptop GPU</code></pre>



<h3 class="wp-block-heading">onnxruntime-gpu インストール</h3>



<div class="wp-block-cocoon-blocks-icon-box common-icon-box block-box information-box">
<p class="wp-block-paragraph">CUDA 12.xでは、ONNX Runtime: onnxruntime-gpu 1.28.0 を用いても、<br>GPU (RTX5060) は、動きません。</p>



<p class="wp-block-paragraph">ONNX Runtime: 1.26.x-1.21.x , CUDA: 12.8 , cuDNN: 9.x<br>この3つの要素を組み合わせます。</p>



<p class="wp-block-paragraph">onnxruntime-gpu==1.26.0, version 1.26をインストールします。</p>
</div>



<pre class="wp-block-code"><code>python -m pip install onnxruntime-gpu==1.26.0</code></pre>



<p class="wp-block-paragraph">動作確認</p>



<pre class="wp-block-code"><code> python -c "import onnxruntime as ort; print(ort.get_available_providers())"
&#91;'TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']</code></pre>



<p class="wp-block-paragraph">CUDAとCPUが有効になっています。</p>



<pre class="wp-block-code"><code>Installing collected packages: flatbuffers, protobuf, onnxruntime-gpu
Successfully installed flatbuffers-25.12.19 onnxruntime-gpu-1.26.0 protobuf-7.36.0</code></pre>



<p class="wp-block-paragraph">動作確認</p>



<pre class="wp-block-code"><code>python -c "import torch; print('CUDA利用可能:', torch.cuda.is_available()); print('Detected GPU:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'None')"</code></pre>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph">Siril Python スクリプトを使っても、PyTorch, onnxruntime-gpu をインストールできるでしょう。GPU Framework Manager を使います。<br><br>コマンドラインからのインストールが、動作やログが表示されるので、こちらではコマンド操作をしています。</p>
</div>



<h3 class="wp-block-heading">Cosmic Clarity Suite で RTX5060 を使う</h3>



<p class="wp-block-paragraph">ここから試行錯誤します。</p>



<h4 class="wp-block-heading">エラーログを読む</h4>



<p class="wp-block-paragraph">Siril 1.5仮想環境をアクティブにして、</p>



<pre class="wp-block-code"><code>(venv) PS C:\local\CosmicClaritySuite_Windows&gt; .\setiastrocosmicclarity.exe</code></pre>



<p class="wp-block-paragraph">若しくは、Siril 1.5 Pythonスクリプト、Cosmic Clarity Suite Sharpen を用いる。</p>



<p class="wp-block-paragraph">以下、エラーログ</p>



<pre class="wp-block-code"><code>Using device: cuda (CUDA)
Loaded TIFF image with dtype: float32
Final bit depth set to: 32-bit floating point
Sharpening Image: result_drizzle_19620s.fit.tif
Total Chunks: 225
Stellar Sharpening:
Traceback (most recent call last):
  File "Python Code\SetiAstroCosmicClarity.py", line 1391, in &lt;module&gt;
  File "Python Code\SetiAstroCosmicClarity.py", line 1216, in process_images
  File "Python Code\SetiAstroCosmicClarity.py", line 930, in sharpen_image
RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1
Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.</code></pre>



<pre class="wp-block-code"><code> Traceback (most recent call last):
   File "Python Code\SetiAstroCosmicClarity.py", line 1391, in &lt;module&gt;
   File "Python Code\SetiAstroCosmicClarity.py", line 1216, in process_images
   File "Python Code\SetiAstroCosmicClarity.py", line 930, in sharpen_image
 RuntimeError: CUDA error: no kernel image is available for execution on the device
 CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
 For debugging consider passing CUDA_LAUNCH_BLOCKING=1
 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.
 &#91;PYI-18620:ERROR] Failed to execute script 'SetiAstroCosmicClarity' due to unhandled exception!

  *#              Cosmic Clarity Suite - Sharpen V6.5 AI3.5s          # 
  *#                                                                  #
  *#                         SetiAstro                                #
  *#                    Copyright © 2025                              #
  *#                                                                  #
 Using device: cuda (CUDA)
 Loaded TIFF image with dtype: float32
 Final bit depth set to: 32-bit floating point
 Sharpening Image: NGC2685_drizzle_19620s.fit.tif
 Total Chunks: 225
 Stellar Sharpening:
</code></pre>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">エラー表示内容：<br>Using device: cuda (CUDA) ：<br>　Cosmic Clarityが、「GPUを使って処理します。CUDA経由でNVIDIA GPUを使用します」<br>Stellar Sharpening：星像のシャープ化処理を開始<br><br>no kernel image is available&#8230;：<br>CUDAでは、GPU上で実行する小さなプログラムを CUDA kernel（カーネル） と呼びます。</p>



<p class="wp-block-paragraph">イメージとしては、<br>Cosmic Clarity→PyTorchなど→CUDA kernel→RTX 5060<br>という流れです。<br><br>RTX 5060の場合は、<br>Compute Capability→12.0→sm_120（CUDA表記）<br></p>



<p class="wp-block-paragraph">ところが、Cosmic Clarityに含まれているCUDA関連コードが、例えば古いGPU向けにしか用意されていないらしく、</p>



<p class="wp-block-paragraph">Cosmic Clarity→「このCUDA kernelを実行して」<br>↓<br>RTX 5060<br>↓<br>「このkernelは私（sm_120）用のコードを持っていない」<br>↓<br>no kernel image is available</p>



<p class="wp-block-paragraph">となります。</p>
</div>



<p class="wp-block-paragraph">どうやら、Cosmic Clarity の .exe ファイルは、パッケージ（PyInstaller）によって古いバージョンの PyTorch が同梱されているらしい。<br>Cosmic Clarityの中に同梱されているCUDA対応ライブラリやPyTorchなどが、RTX 5060（sm_120）用のkernelを持っていないらしい。</p>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">こちらの環境では、</p>



<p class="wp-block-paragraph">Windows 11<br>│<br>├── NVIDIA Driver 616.56<br>│ └── RTX 5060を正常に認識<br>│<br>├── CUDA Toolkit 12.8.2<br>│<br>└── Cosmic Clarity<br>　　　│<br>　　　├── setiastrocosmicclarity.exe<br>　　　│<br>　　　└── _internal<br>　　　　　　│<br>　　　　　　├── PyTorch<br>　　　　　　├── CUDA関連DLL<br>　　　　　　└── CUDA kernel<br>　　　　　　　　　│<br>　　　　　　　　　↓<br>　　　　　　　　　RTX 5060<br>　　　　　　　　　sm_120<br>　　　　　　　　　│<br>　　　　　　　　　X<br>　　　　　　　　　対応kernelがない？<br>　　　　　　　　　│<br>　　　　　　　　　↓<br>　　　　　　　　　no kernel image is available</p>
</div>



<p class="wp-block-paragraph">Cosmic Clarityが使用しているCUDA関連の実行コードのどこかが、RTX 5060（sm_120）に対応していない可能性が高いらしい。</p>



<pre class="wp-block-code"><code>setiastrocosmicclarity.exe にあるPyTorch は、RTX 30シリーズや40シリーズまでに対応か？
　古いPyTorchは、RTX 5060（Blackwell世代）の計算命令（sm_120）が無いので動かせない。

CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect：
 CUDAのエラーは非同期（時間差）で報告されることがあるため、以下のスタックトレース（エラーが発生した場所の記録）は正確ではない可能性があります

RuntimeError: CUDA error: no kernel image is available for execution on the device：
 このGPUデバイス上で実行できる、カーネルイメージ（コンパイル済みの命令データ）が見つからない。

GPUは大量の計算を「裏でまとめて」処理しています。そのため、プログラムが「これを計算して！」とGPUに命令を投げた瞬間ではなく、少し後になってGPU側から「さっきの命令、僕の言葉じゃなくて読めなかったよ」とエラーが返ってくる。そのため、プログラムがエラーで止まった一行手前や、別の場所が本当の引き金かもしれないよ、という開発者向けの注意書き。
　環境変数 CUDA_LAUNCH_BLOCKING=1 を設定して実行すると良いかもと言っている。
　GPUの不具合検知を有効にするには、TORCH_USE_CUDA_DSA を付けてコンパイルしてね、とも</code></pre>



<pre class="wp-block-code"><code>python "Python Code\SetiAstroCosmicClarity.py"
C:\Users\minmin\AppData\Local\Programs\Python\Python312\python.exe: can't open file 'C:\\local\\CosmicClaritySuite_Windows\\Python Code\\SetiAstroCosmicClarity.py': &#91;Errno 2] No such file or directory
### 動かない</code></pre>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">[PYI-13972:ERROR] Failed to execute script &#8216;SetiAstroCosmicClarity&#8217;<br>　という表示から、setiastrocosmicclarity.exe は PyInstallerで固められたアプリかな？<br><br>構造としては、おそらく<br>CosmicClaritySuite_Windows<br>│<br>├─ setiastrocosmicclarity.exe<br>│<br>└─ _internal<br>　　├─ torch<br>　　├─ torchvision<br>　　├─ torchaudio<br>　　├─ numpy<br>　　├─ scipy<br>　　├─ &#8230;<br>　　└─ その他DLL<br>と思われる。（Frankに聞けば良い&#8230;）</p>
</div>



<p class="wp-block-paragraph">C:\local\CosmicClaritySuite_Windows_internal\torchにある、<br><code>version.py</code><br>を見てみると、</p>



<pre class="wp-block-code"><code>Get-Content .\_internal\torch\version.py
from typing import Optional

__all__ = &#91;'__version__', 'debug', 'cuda', 'git_version', 'hip']
__version__ = '2.5.1+cu121'
debug = False
cuda: Optional&#91;str] = '12.1'
git_version = 'a8d6afb511a69687bbb2b7e88a3cf67917e1697e'
hip: Optional&#91;str] = None</code></pre>



<p class="wp-block-paragraph">Cosmic Clarity Windowsに同梱されているPyTorchは、<br>PyTorch 2.5.1+cu121<br>CUDA 12.1</p>



<p class="wp-block-paragraph">Cosmic Clarity内蔵のPyTorch 2.5.1+cu121を、RTX 5060 / Blackwell対応の新しいPyTorchに交換すると、他の部分との互換性が保てない。<br></p>



<pre class="wp-block-code"><code>(Get-Item .\setiastrocosmicclarity.exe).VersionInfo | Format-List *
FileVersionRaw     : 0.0.0.0
ProductVersionRaw  : 0.0.0.0
Comments           :
CompanyName        :
FileBuildPart      : 0
FileDescription    :
FileMajorPart      : 0
FileMinorPart      : 0
FileName           : C:\local\CosmicClaritySuite_Windows\setiastrocosmicclarity.exe
FilePrivatePart    : 0
FileVersion        :
InternalName       :
IsDebug            : False
IsPatched          : False
IsPrivateBuild     : False
IsPreRelease       : False
IsSpecialBuild     : False
Language           :
LegalCopyright     :
LegalTrademarks    :
OriginalFilename   :
PrivateBuild       :
ProductBuildPart   : 0
ProductMajorPart   : 0
ProductMinorPart   : 0
ProductName        :
ProductPrivatePart : 0
ProductVersion     :
SpecialBuild       :</code></pre>



<p class="wp-block-paragraph">アップデートを使ってみます。<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity/releases/tag/Windows">https://github.com/setiastro/cosmicclarity/releases/tag/Windows</a><br>darkstar_*.onnx<br>darkstar_*.pth<br>deep_*.onnx<br>deep_*.pth<br>satellite*.onnx<br>satellite*.pth<br>setiastro*.exe<br>superres_*.onnx<br>superres_*.pth<br>各ファイルを入れ替えます。<br></p>



<pre class="wp-block-code"><code>エラー表示
 .\setiastrocosmicclarity.exe
Traceback (most recent call last):
  File "Python Code\SetiAstroCosmicClarity.py", line 5, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\__init__.py", line 2108, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\functional.py", line 7, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\nn\__init__.py", line 8, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\nn\modules\__init__.py", line 1, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\nn\modules\module.py", line 29, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\utils\__init__.py", line 8, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\utils\data\__init__.py", line 1, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\utils\data\dataloader.py", line 20, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\distributed\__init__.py", line 122, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\distributed\device_mesh.py", line 40, in &lt;module&gt;
  File "&lt;frozen importlib._bootstrap&gt;", line 1360, in _find_and_load
  File "&lt;frozen importlib._bootstrap&gt;", line 1331, in _find_and_load_unlocked
  File "&lt;frozen importlib._bootstrap&gt;", line 935, in _load_unlocked
  File "PyInstaller\loader\pyimod02_importers.py", line 378, in exec_module
  File "torch\distributed\distributed_c10d.py", line 234, in &lt;module&gt;
  File "torch\distributed\distributed_c10d.py", line 285, in Backend
AttributeError: type object 'torch._C._distributed_c10d.BackendType' has no attribute 'XCCL'. Did you mean: 'NCCL'?
&#91;PYI-3680:ERROR] Failed to execute script 'SetiAstroCosmicClarity' due to unhandled exception!</code></pre>



<div class="wp-block-cocoon-blocks-info-box block-box primary-box">
<p class="wp-block-paragraph">AttributeError: type object &#8216;torch._C._distributed_c10d.BackendType&#8217; has no attribute &#8216;XCCL&#8217;. Did you mean: &#8216;NCCL&#8217;?<br>[PYI-3680:ERROR] Failed to execute script &#8216;SetiAstroCosmicClarity&#8217; due to unhandled exception!<br><br>新しく上書きしたCosmic Clarityのコードが、_internal のPyTorch 2.5.1+cu121 と組み合わさったため、互換性が崩れた。</p>



<p class="wp-block-paragraph">EXEやPythonコードだけ更新して、_internalを残す組み合わせでは動かない。</p>
</div>



<h4 class="wp-block-heading">SetiAstroCosmicClarity.py を使ってみます。</h4>



<p class="wp-block-paragraph">作者のFrankin Marekさんは、プログラムのソースコードを完全に公開しています。有難い事です。</p>



<div class="wp-block-cocoon-blocks-blank-box-1 blank-box block-box">
<p class="wp-block-paragraph">公式の完全なWindows版を入手します。</p>



<p class="wp-block-paragraph">今回のReleaseページを見ると、現在確認できるWindows更新版は、</p>



<p class="wp-block-paragraph">Files to grab to update just various files instead of having to download the full Cosmic Clarity</p>



<p class="wp-block-paragraph">という位置付けです。<br>これは _internal を更新するものではありません。現在のGitHubリポジトリには、Pythonソースも公開されています。</p>
</div>



<p class="wp-block-paragraph">C:\local\CosmicClarityRTX5060ディレクトリを作り、仮想環境下で動くか試します。</p>



<pre class="wp-block-code"><code>
py -3.12 -m venv .venv

Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope Process

.\.venv\Scripts\Activate.ps1

(.venv) PS C:\local\CosmicClarityRTX5060&gt; python --version
Python 3.12.10

(.venv) PS C:\local\CosmicClarityRTX5060&gt; python -c "import sys; print(sys.executable)" C:\local\CosmicClarityRTX5060\.venv\Scripts\python.exe

python -m pip install torch==2.9.0 torchvision==0.24.0 torchaudio==2.9.0 --index-url https://download.pytorch.org/whl/cu128
### 2.9GBをダウンロードします。</code></pre>



<p class="wp-block-paragraph">動作テストをします。</p>



<pre class="wp-block-code"><code>python -c "import torch; print('PyTorch:', torch.__version__); print('Torch CUDA:', torch.version.cuda); print('CUDA available:', torch.cuda.is_available()); print('GPU:', torch.cuda.get_device_name(0)); print('Capability:', torch.cuda.get_device_capability(0))"</code></pre>



<pre class="wp-block-code"><code>PyTorch: 2.9.0+cu128
Torch CUDA: 12.8
CUDA available: True
GPU: NVIDIA GeForce RTX 5060 Laptop GPU
Capability: (12, 0)</code></pre>



<pre class="wp-block-code"><code>python -c "import torch; x=torch.randn(1024,1024,device='cuda'); y=x@x; torch.cuda.synchronize(); print('CUDA TEST: OK', y.shape)"
###import torch; PyTorchを読み込む。
###x=torch.randn(1024,1024,device='cuda'); GPU（CUDA）上に、1024×1024のランダムな数値行列 x を作る。
### y=x@x; GPU上で、1024×1024行列 × 1024×1024行列、計算を実行する。</code></pre>



<pre class="wp-block-code"><code>CUDA TEST: OK torch.Size(&#91;1024, 1024])</code></pre>



<p class="wp-block-paragraph">setiastro /cosmicclarity<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity?utm_source=chatgpt.com">https://github.com/setiastro/cosmicclarity?utm_source=chatgpt.com</a><br>requirements.txtを見ると、<br>PyQt5, torch, numpy, tifffile, astropy, Pillow, torchvision, lz4, zstandard, rawpy</p>



<p class="wp-block-paragraph">これらをインストールして試します。</p>



<pre class="wp-block-code"><code>python -m pip install numpy tifffile astropy Pillow lz4 zstandard rawpy</code></pre>



<pre class="wp-block-code"><code>python -c "import numpy, tifffile, astropy, PIL, lz4, zstandard, rawpy; print('numpy:', numpy.__version__); print('tifffile:', tifffile.__version__); print('astropy:', astropy.__version__); print('Pillow:', PIL.__version__); print('lz4:', lz4.__version__); print('zstandard:', zstandard.__version__); print('rawpy:', rawpy.__version__)"</code></pre>



<pre class="wp-block-code"><code>python -c "import torch, torchvision; print('torch:', torch.__version__); print('torchvision:', torchvision.__version__); print('CUDA:', torch.version.cuda); print('GPU:', torch.cuda.get_device_name(0))"</code></pre>



<p class="wp-block-paragraph">依存関係を見てみると、<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity?utm_source=chatgpt.com">https://github.com/setiastro/cosmicclarity?utm_source=chatgpt.com</a><br>requirements.txt<br>PyQt5, torch, numpy, tifffile, astropy, Pillow, torchvision, lz4, zstandard,</p>



<p class="wp-block-paragraph">PyQt5とありますが、現在のPythonソースはPyQt6をimportしています。</p>



<pre class="wp-block-code"><code>python -m pip install xisf PyQt6 sep</code></pre>



<pre class="wp-block-code"><code>python -m pip install onnxruntime-gpu==1.26.0</code></pre>



<p class="wp-block-paragraph">C:\local\CosmicClarityRTX5060ディレクトリへ [SetiAstroCosmicClarity.py]　を配置します。</p>



<pre class="wp-block-code"><code>setiastro / cosmicclarity (Public)
https:&#47;&#47;github.com/setiastro/cosmicclarity/
SetiAstroCosmicClarity.py</code></pre>



<pre class="wp-block-code"><code>python .\SetiAstroCosmicClarity.py
### GUI起動を確認</code></pre>



<p class="wp-block-paragraph">C:\local\CosmicClarityRTX5060ディレクトリに、input、outputディレクトリを作る。<br>inputディレクトリへfitsファイルを配置する。</p>



<pre class="wp-block-code"><code> python .\SetiAstroCosmicClarity.py --sharpening_mode "Stellar Only"

###エラー表示
stellar_model_radius_1.load_state_dict(torch.load(os.path.join(exe_dir, 'deep_sharp_stellar_cnn_AI3_5s.pth'), map_location=device))</code></pre>



<p class="wp-block-paragraph">次のファイルを同じディレクトリへ配置すると、SetiAstroCosmicClarity.pyは動いた。<br><br>deep_nonstellar_sharp_cnn_radius_1AI3_5s.pth<br>deep_nonstellar_sharp_cnn_radius_2AI3_5s.pth<br>deep_nonstellar_sharp_cnn_radius_4AI3_5s.pth<br>deep_nonstellar_sharp_cnn_radius_8AI3_5s.pth<br>deep_sharp_stellar_cnn_AI3_5s.pth</p>



<pre class="wp-block-code"><code>python .\SetiAstroCosmicClarity.py</code></pre>



<p class="wp-block-paragraph">動作ログ、スタンドアローンはこれで動きました。</p>



<pre class="wp-block-code"><code>Using device: cuda (CUDA)
Loaded TIFF image with dtype: float32
Final bit depth set to: 32-bit floating point
Sharpening Image: result_drizzle_19620s.fit.tif
Total Chunks: 225
Stellar Sharpening:
Progress: 100.00% (225/225 chunks processed)
Stellar sharpening complete. Shape: (2731, 2731)
Updating luminance for non-stellar sharpening...
Non-Stellar Sharpening:
Progress: 100.00% (225/225 chunks processed)</code></pre>



<p class="wp-block-paragraph">以上を踏まえて、CosmicClaritySuite_Windowsディレクトリに変更を行います。</p>



<h4 class="wp-block-heading">CosmicClaritySuite_Windows へ &#8220;Windows Files Excluding _internal&#8221;を導入</h4>



<p class="wp-block-paragraph">Windows Files Excluding _internal にある各ファイルに入れ替えます。<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity/releases/tag/Windows">https://github.com/setiastro/cosmicclarity/releases/tag/Windows</a></p>



<p class="wp-block-paragraph">darkstar_*.onnx<br>darkstar_*.pth<br>deep_*.onnx<br>deep_*.pth<br>satellite*.onnx<br>satellite*.pth<br>setiastro*.exe<br>superres_*.onnx<br>superres_*.pth<br>各ファイルを入れ替えます。</p>



<h4 class="wp-block-heading">CosmicClaritySuite_Windows へ SetiAstroCosmicClarity.py を導入</h4>



<p class="wp-block-paragraph">cosmicclarityにある、SetiAstroCosmicClarity.py、SetiAstroCosmicClarity_denoise.pyをダウンロードし配置します。<br><a rel="noopener" target="_blank" href="https://github.com/setiastro/cosmicclarity">https://github.com/setiastro/cosmicclarity</a></p>



<p class="wp-block-paragraph">不足するパッケージをインストールします。</p>



<pre class="wp-block-code"><code>python -m pip install xisf astropy sep opencv-python</code></pre>



<h4 class="wp-block-heading">CosmicClaritySuite_Windows へ バッチファイルを導入</h4>



<p class="wp-block-paragraph">[SetiAstroCosmicClarity.py]　へSiril から諸設定などの情報を受け取れるバッチファイルを作ります。<br>&#8220;run_SetiAstroCosmicClarity.bat&#8221; とします。</p>



<pre class="wp-block-code"><code>@echo off
call C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\activate.bat
python "%~dp0SetiAstroCosmicClarity.py" %*</code></pre>



<p class="wp-block-paragraph">[SetiAstroCosmicClarity_denoise.py]　へSiril から諸設定などの情報を受け取れるバッチファイルを作ります。<br>&#8220;run_SetiAstroCosmicClarity_denoise.bat&#8221; とします。</p>



<pre class="wp-block-code"><code>@echo off
call C:\Users\USERNAME\AppData\Local\siril\venv\Scripts\activate.bat
python "%~dp0SetiAstroCosmicClarity_denoise.py" %*</code></pre>



<h4 class="wp-block-heading">%~dp0</h4>



<ul class="wp-block-list">
<li>%0：実行されているバッチファイル自身のファイル名（絶対パス）</li>



<li>~：前後の不要なダブルクォーテーション（&#8221;）を取り除く</li>



<li>d：ドライブ名（例：C:）だけを抽出</li>



<li>p：フォルダの経路（パス：\Users\USERNAME\AppData\Local\siril\venv\Scripts\）を抽出</li>
</ul>



<p class="wp-block-paragraph">Siril などの外部ソフトからこのバッチファイルが呼び出されたとき、パソコンの内部（カレントディレクトリ）は「Sirilのプログラムがある場所」になってしまっています。もし単に python SetiAstroCosmicClarity.py とだけ書いておくと、システムは Siril のフォルダから SetiAstroCosmicClarity.py を探してしまい、「ファイルが見つかりません」というエラーになります。</p>



<p class="wp-block-paragraph">%~dp0 を付けて &#8220;%~dp0SetiAstroCosmicClarity.py&#8221; と書くことで、「このバッチファイルと同じフォルダ内にある SetiAstroCosmicClarity.py を実行」 とSirilに指示を出せます。</p>



<div class="wp-block-cocoon-blocks-sticky-box blank-box block-box sticky">
<p class="wp-block-paragraph">最新のGPUを使うには、みんなどうやっているのかな？<br>きっと、もっと簡単なのだろう。。。</p>
</div>



<p class="wp-block-paragraph">&#8220;SetiAstroCosmicClarity_denoise.py&#8221; も同様に設定します。<br>SetiAstroCosmicClarity, Siril Pythonスクリプトの動作設定は、ここまでです。</p>



<h3 class="wp-block-heading">SyQon Prizm, Starless</h3>



<p class="wp-block-paragraph">C:\Users\USERNAME\AppData\Local\siril<br>以下に、次のディレクトリがあり、そこへそれぞれのファイルを配置します。<br>syqon_prism　→　prism_mini.pt<br>syqon_starless　→　zenith.pt</p>



<hr class="wp-block-separator has-alpha-channel-opacity"/>



<h3 class="wp-block-heading">Build Tools for Visual Studio が必要になるのは、</h3>



<p class="wp-block-paragraph">Siril での画像処理では、開発を行っていないならば、C++をコンパイル、ビルドする必要は、まず、ありません。</p>



<p class="wp-block-paragraph">そのために、Build Tools for Visual Studioは、通常は、不必要です。</p>



<p class="wp-block-paragraph">Siril 1.5 を使い、天体画像処理をPythonスクリプトなどを用いるならば、numpy, PyQt6, astropyなどのスクリプト実行に必要なPythonライブラリや、PyTorchなどのAIフレームワークは、Windows用にコンパイル済みのパッケージ（Wheel形式）が、スクリプト実行時にインストールされます。</p>



<p class="wp-block-paragraph">Windows版 CUDA Toolkitインストーラーを実行すると、途中で「No supported version of Visual Studio was found…」という警告が表示されることがあります。これは「Visual Studio用のプラグイン（開発者向けの統合環境）が入れられない」という警告なので、そのまま進めると、ランタイムやドライバーは正常にインストールされます。</p>



<p class="wp-block-paragraph">もしも、事前にビルドされていないPythonパッケージを pip install する必要が出た場合などは、画面に 「Microsoft Visual C++ 14.0 or greater is required 」とエラー表示されます。</p>



<p class="wp-block-paragraph">この場合は、「Build Tools for Visual Studio」 を開き、「C++ によるデスクトップ開発」にチェックを入れてインストールします。</p>



<p class="wp-block-paragraph">GUIにてBuild Tools for Visual Studioをインストール<br>「C++ によるデスクトップ開発」をクリックする。画面右側で「ATL」と「MFC」にチェックを入れ、「インストール」をクリックする。</p>



<p class="wp-block-paragraph">コマンドを用いてインストールするならば、以下を参考にします。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow"><a rel="noopener" target="_blank" href="https://www.kkaneko.jp/tools/win/cuda.html" title="
NVIDIA ドライバ，NVIDIA CUDA ツールキット 12.8，NVIDIA cuDNN v9.19.0 のインストール（Windows 上）｜金子邦彦研究室" class="blogcard-wrap external-blogcard-wrap a-wrap cf"><div class="blogcard external-blogcard eb-left cf"><div class="blogcard-label external-blogcard-label"><span class="fa"></span></div><figure class="blogcard-thumbnail external-blogcard-thumbnail"><img loading="lazy" decoding="async" src="https://s.wordpress.com/mshots/v1/https%3A%2F%2Fwww.kkaneko.jp%2Ftools%2Fwin%2Fcuda.html?w=320&#038;h=180" alt="" class="blogcard-thumb-image external-blogcard-thumb-image" width="320" height="180" /></figure><div class="blogcard-content external-blogcard-content"><div class="blogcard-title external-blogcard-title">
NVIDIA ドライバ，NVIDIA CUDA ツールキット 12.8，NVIDIA cuDNN v9.19.0 のインストール（Windows 上）｜金子邦彦研究室</div><div class="blogcard-snippet external-blogcard-snippet"></div></div><div class="blogcard-footer external-blogcard-footer cf"><div class="blogcard-site external-blogcard-site"><div class="blogcard-favicon external-blogcard-favicon"><img loading="lazy" decoding="async" src="https://www.google.com/s2/favicons?domain=https://www.kkaneko.jp/tools/win/cuda.html" alt="" class="blogcard-favicon-image external-blogcard-favicon-image" width="16" height="16" /></div><div class="blogcard-domain external-blogcard-domain">www.kkaneko.jp</div></div></div></div></a></blockquote>



<p class="wp-block-paragraph">winget を用いて以下のインストールを行うので、先ずは、wingetを確認します。<br>&#8220;Win&#8221;+&#8221;X&#8221;key → &#8220;i&#8221;key → &#8220;Windows PowerShell</p>



<pre class="wp-block-code"><code>winget
Windows Package Manager v1.29.280</code></pre>



<p class="wp-block-paragraph">Build Tools for Visual Studio 2026 のインストール</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph"><br><a rel="noopener" target="_blank" href="https://learn.microsoft.com/ja-jp/windows/package-manager/winget/install">https://learn.microsoft.com/ja-jp/windows/package-manager/winget/install</a></p>
</blockquote>



<pre class="wp-block-code"><code>winget install --scope machine --id Microsoft.VCRedist.2015+.x64 -e --silent --disable-interactivity --force --accept-source-agreements --accept-package-agreements --override "/quiet /norestart"
</code></pre>



<ul class="wp-block-list">
<li>&#8211;scope: インストーラーでユーザーまたはコンピューターのスコープをターゲットにする必要がある場合に指定できます。</li>



<li>&#8211;id: installをアプリケーションの ID に制限します。</li>



<li>-e, &#8211;exact: 大文字小文字の区別の検査を含め、クエリで正確な文字列を使用します。 部分文字列の既定の動作は使用されません。</li>



<li>&#8211;silent: サイレント モードでインストーラーを実行します。 これにより、すべての UI が抑制されます。 既定のエクスペリエンスでは、インストーラーの進行状況が表示されます。</li>



<li><code>--disable-interactivity</code>: 対話型プロンプトを無効にします。</li>



<li><code>--force</code>: コマンドを直接実行し、セキュリティ関連以外の問題を続行します。</li>



<li><code>--accept-source-agreements</code>:  対話型プロンプトを抑制して、WinGet ソース (リポジトリ) の使用許諾契約書に同意します。 これはパッケージ ライセンスとは別であり、ソース自体の使用条件 ( winget コミュニティ リポジトリなど) が対象となります。</li>



<li><code>--accept-package-agreements</code>: パッケージ インストーラーによって提示された使用許諾契約書または EULA を受け入れ、対話型プロンプトを表示しません。 これはパッケージのライセンス条項にのみ適用されます。インストーラーによって提供されるオプションのコンポーネントやバンドルされたソフトウェアには影響しません。 完全に非対話型の installの場合は、 <strong><code>--silent</code> (<code>-h</code>)</strong> と組み合わせます。</li>



<li><code>--override</code>: インストーラーに直接渡される文字列。</li>



<li><code>"/quiet /norestart"</code></li>
</ul>



<h3 class="wp-block-heading">zlib のインストールは不要</h3>



<p class="wp-block-paragraph">zlibは、別途、インストールする必要はありません。<br>cuDNN 9.1.0のドキュメントに案内があります。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">ZLIB version 1.2.13 is statically linked into the cuDNN Windows dynamic libraries. Changing to the static linkage of ZLIB for other platforms will be addressed in future cuDNN releases.<br><a rel="noopener" target="_blank" href="https://docs.nvidia.com/deeplearning/cudnn/backend/latest/release-notes.html">https://docs.nvidia.com/deeplearning/cudnn/backend/latest/release-notes.html</a><br><br>ZLIB バージョン 1.2.13 は、cuDNN の Windows 向けダイナミックライブラリに静的リンクされています。その他のプラットフォームにおける ZLIB の静的リンクへの変更については、今後の cuDNN リリースで対応される予定です。</p>
</blockquote>



<pre class="wp-block-code"><code><a rel="noopener" target="_blank" href="https://www.winimage.com/zLibDll/index.html">https://www.winimage.com/zLibDll/index.html</a>

There is also a DLL for Windows 64 bits, for <strong>AMD64/Intel EM64T</strong> and for Intel Itanium.</code></pre>



<p class="wp-block-paragraph">展開した中にある <code>zlibwapi.dll</code> を、PATH の通った場所にコピーします。cuDNN の <code>bin</code> フォルダにコピーします。</p>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/setting-up-ubuntu26-and-windows11-for-astronomy/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>プロセッサ 拡張機能と Siril</title>
		<link>https://globe3.ddns.net/wp/ryzen-ai7-350-features-and-siril/</link>
					<comments>https://globe3.ddns.net/wp/ryzen-ai7-350-features-and-siril/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Mon, 13 Jul 2026 05:09:03 +0000</pubDate>
				<category><![CDATA[Siril]]></category>
		<category><![CDATA[Ubuntu]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=3542</guid>

					<description><![CDATA[これは、Siril をソースコードからインストールする際に、用いるプロセッサの拡張機能を使用するメモです。記載、編集途中です。 meson mesonは、素粒子物理の「中間子」でははくて、ソフトウェア構築システムです。h [&#8230;]]]></description>
										<content:encoded><![CDATA[<a rel="noopener" target="_blank" href="https://siril.readthedocs.io/en/latest/installation/source.html" title="Installation from source code &mdash; Siril 1.5.0 documentation" class="blogcard-wrap external-blogcard-wrap a-wrap cf"><div class="blogcard external-blogcard eb-left cf"><div class="blogcard-label external-blogcard-label"><span class="fa"></span></div><figure class="blogcard-thumbnail external-blogcard-thumbnail"><img loading="lazy" decoding="async" src="https://s.wordpress.com/mshots/v1/https%3A%2F%2Fsiril.readthedocs.io%2Fen%2Flatest%2Finstallation%2Fsource.html?w=320&#038;h=180" alt="" class="blogcard-thumb-image external-blogcard-thumb-image" width="320" height="180" /></figure><div class="blogcard-content external-blogcard-content"><div class="blogcard-title external-blogcard-title">Installation from source code &mdash; Siril 1.5.0 documentation</div><div class="blogcard-snippet external-blogcard-snippet"></div></div><div class="blogcard-footer external-blogcard-footer cf"><div class="blogcard-site external-blogcard-site"><div class="blogcard-favicon external-blogcard-favicon"><img loading="lazy" decoding="async" src="https://www.google.com/s2/favicons?domain=https://siril.readthedocs.io/en/latest/installation/source.html" alt="" class="blogcard-favicon-image external-blogcard-favicon-image" width="16" height="16" /></div><div class="blogcard-domain external-blogcard-domain">siril.readthedocs.io</div></div></div></div></a>


<p class="wp-block-paragraph">これは、<br>Siril をソースコードからインストールする際に、用いるプロセッサの拡張機能を使用するメモです。<br>記載、編集途中です。</p>



<h2 class="wp-block-heading">meson</h2>



<p class="wp-block-paragraph">mesonは、素粒子物理の「中間子」でははくて、ソフトウェア構築システムです。<br><a rel="noopener" target="_blank" href="https://mesonbuild.com/">https://mesonbuild.com/</a><br>陽子と中性子を結びつける素粒子である、Gluon「グルーオン」が候補だったそうですが、既に他で使われていたので、mesonになったそうです。<br><a rel="noopener" target="_blank" href="https://mesonbuild.com/FAQ.html">https://mesonbuild.com/FAQ.html</a></p>



<p class="wp-block-paragraph">meson ビルドオプション<br><a rel="noopener" target="_blank" href="https://mesonbuild.com/Builtin-options.html">https://mesonbuild.com/Builtin-options.html</a></p>



<h2 class="wp-block-heading">AMD Ryzen™ AI 7 350 プロセッサ 拡張機能</h2>



<p class="wp-block-paragraph">ここでは、AMD Ryzen AI 7 350 拡張機能とmeson オプションの関係を記述します。</p>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">対応する拡張機能<br>AES , AMD-V , AVX , AVX512 , AVX2 , FMA3 , MMX-plus , SHA , SSE , SSE2 , SSE3 , SSE4.1 , SSE4.2 , SSE4A , SSSE3 , x86-64<br><a rel="noopener" target="_blank" href="https://www.amd.com/ja/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-7-350.html">https://www.amd.com/ja/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-7-350.html</a></p>
</blockquote>



<ul class="wp-block-list">
<li>AES
<ul class="wp-block-list">
<li>AES-NI (Advanced Encryption Standard New Instructions)<br>マイクロプロセッサのx86命令セットへの拡張機能。AES-NIは、AESによる暗号化および復号の高速化を行う。</li>
</ul>
</li>



<li>AMD-V
<ul class="wp-block-list">
<li>コンピュータの仮想環境を支援する。</li>
</ul>
</li>



<li>AVX512
<ul class="wp-block-list">
<li>AVX-512（Advanced Vector Extensions 512）は、512ビット幅のベクトル演算（一度に複数のデータを処理するSIMD技術）を可能にする命令セットです。<br>高い実行効率を実現するため、システム全体での電力供給や熱設計の改善も施されています。</li>
</ul>
</li>



<li>FMA3
<ul class="wp-block-list">
<li>FMA3（Fused Multiply-Add 3-operand）は、3オペランド形式のFused Multiply-Add と呼ばれる浮動小数点数の積和算命令<br>乗算と加算（例えば、A*B+C）を1クロックで処理する。</li>
</ul>
</li>
</ul>



<p class="wp-block-paragraph"></p>



<h2 class="wp-block-heading">機能拡張を有効</h2>



<p class="wp-block-paragraph"></p>



<p class="wp-block-paragraph">プロセッサの機能拡張を有効にするには、以下のオプションを用います。</p>



<pre class="wp-block-code"><code>meson configure _build -Dcpp_args="-march=native" -Dc_args="-march=native"</code></pre>



<h2 class="wp-block-heading">WindowsでのMsys2を用いた Siril構築</h2>



<p class="wp-block-paragraph">依存関係をインストールするには以下のコマンド</p>



<pre class="wp-block-code"><code>curl -s -o siril-deps.sh https://gitlab.com/free-astro/siril/-/raw/master/build/windows/native-gitlab-ci/siril-deps.sh
bash siril-deps.sh</code></pre>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/ryzen-ai7-350-features-and-siril/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
		<item>
		<title>VeraLux Nox, Silentium と Siril</title>
		<link>https://globe3.ddns.net/wp/using-veralux-with-siril/</link>
					<comments>https://globe3.ddns.net/wp/using-veralux-with-siril/#respond</comments>
		
		<dc:creator><![CDATA[minmin]]></dc:creator>
		<pubDate>Wed, 20 May 2026 07:48:00 +0000</pubDate>
				<category><![CDATA[GNU/Linux]]></category>
		<category><![CDATA[Siril]]></category>
		<category><![CDATA[画像処理]]></category>
		<guid isPermaLink="false">https://globe3.ddns.net/wp/?p=3304</guid>

					<description><![CDATA[Cyril Richardさん と オープンソース Cyril RichardさんとRich Stevensonさんが対談してます。https://www.youtube.com/watch?v=Y6b3VVf-tvk  [&#8230;]]]></description>
										<content:encoded><![CDATA[
<h2 class="wp-block-heading">Cyril Richardさん と オープンソース</h2>



<p class="wp-block-paragraph">Cyril RichardさんとRich Stevensonさんが対談してます。<br><a rel="noopener" target="_blank" href="https://www.youtube.com/watch?v=Y6b3VVf-tvk">https://www.youtube.com/watch?v=Y6b3VVf-tvk</a></p>



<p class="wp-block-paragraph">対談を聞いてみると、</p>



<p class="wp-block-paragraph">Siril開発者のCyril Richardさんは、理論物理学者です。<br>天体物理学分野では分光学、氷の分析、系外惑星の大気を研究してるそうです。<br>彼曰く、それらを行う量子力学者だそうです。</p>



<p class="wp-block-paragraph">彼はLinuxのみで作業をするので、Linuxで動作する、無料で自由な天体写真処理ツールが欲しかったそうです。当時はそれが無いのでSirilを開発したそうです。</p>



<p class="wp-block-paragraph">Sirilの始まりは、Christian Builさんの &#8220;IRIS&#8221;のLinux版としてでした。これが2005年の事でした。<br>一時中断していた開発を、2012年ころから、Cyril Richardさんが、その後を引き継いだようです。</p>



<p class="wp-block-paragraph">Sirilの名称は、元の&#8221;IRIS&#8221;の文字を後ろから並び替えて、IRIS → iris → siri<br>そこに&#8221;Linux&#8221;の先頭文字の &#8220;l&#8221; をつなげる形にしたようです。&#8221;siri&#8221; + &#8220;l&#8221; =&gt; Siril<br>UNIXやGNUに見られる命名法です。 Cyril Richardさんの名前とは無関係です。</p>



<p class="wp-block-paragraph">量子力学者の彼は、ノイズの多いデータを扱う事には慣れていたので、Sirilの画像処理アルゴリズムも厳密さと信頼性が高くなるように努めているそうです。</p>



<p class="wp-block-paragraph">彼はきちんと動作し、綺麗な画像が得られる、Linux上のプログラムが欲しかったそうです。<br>無料であり、しかもオープンソースのプログラムが欲しかったので、Sirilをその形にしたそうです。</p>



<p class="wp-block-paragraph">オープンソースで開発した事に因り、Sirilは広がりを見せたと思えます。<br>マルチプラットフォームになり、Pythonスクリプトでの処理が可能になり、多くのプログラマー、テスターが参加し、開発がより速くなりました。また、開発に掛かる費用を、寄付で賄えるようになりました。</p>



<p class="wp-block-paragraph">昨今は、AIでプログラム開発が出来ます。<br>ですが、そのプログラムがより良くなる為に必要な事は、何でしょうか？<br>それは、オープンソースプログラムであり、それゆえに多くの人に楽しまれながら、育まれ、より大きく育つことが出来ます。</p>



<p class="wp-block-paragraph">Cyril RichardさんとRich Stevensonさんの話しから、強くそう思いました。</p>



<h2 class="wp-block-heading">VeraLux Nox</h2>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">VeraLux Noxは、Zenith Membrane（ゼニス・メンブレン）と呼ばれる概念を基盤とした、決定論的な背景モデリングエンジンを導入しています。このアルゴリズムは、統計的サンプリングや曲線近似によって勾配を推定するのではなく、離散ポアソン方程式を直接解くことで、空の背景を物理的に弾性を持つ表面として再構築します。<br><a rel="noopener" target="_blank" href="https://www.youtube.com/watch?v=eK9VU1d-VP0">https://www.youtube.com/watch?v=eK9VU1d-VP0</a></p>
</blockquote>



<h2 class="wp-block-heading">VeraLux Silentium</h2>



<blockquote class="wp-block-quote is-layout-flow wp-block-quote-is-layout-flow">
<p class="wp-block-paragraph">VeraLux Silentium Pythonスクリプトは、Sirilのリニアステージ向けに特別に設計された、強力な非AIノイズ除去エンジンを導入しています。</p>



<p class="wp-block-paragraph">AIベースとは異なり、Silentiumはリニアデータに対して直接処理を行うため、微弱な構造や細かなディテールを保持しつつ、制御された方法で、天体画像に適した安全な手法により背景ノイズを低減します。これにより、微細部分の誤認識や過度な平滑化を行うことなく、ストレッチ処理の前にクリーンなデータを得たいユーザーにとって理想的な解決法となります。<br><a rel="noopener" target="_blank" href="https://www.youtube.com/watch?v=N4CUQA2X-cE">https://www.youtube.com/watch?v=N4CUQA2X-cE</a></p>
</blockquote>



<p class="wp-block-paragraph">Silentium を用いるには、&#8221;Defaults&#8221;ボタンを押して、初期状態でノイズ除去を行うと良いでしょう。<br>なぜなら、Silentiumは、ノイズの統計計算を行うからです。</p>



<ul class="wp-block-list">
<li>スペースキーを押し続けると、未処理の画像、オリジナル画像が表示される。</li>



<li>赤枠はプレビュー
<ul class="wp-block-list">
<li>ノイズ除去を確認したい部分をクリックします。</li>



<li>戻すには、画面上のどこかをクリックするか、「1:1ボタン」をクリックします。</li>
</ul>
</li>
</ul>



<ul class="wp-block-list">
<li>Noise Intensity:<br>ノイズ除去強度</li>



<li>Detail Guard:<br>微細構造の保護</li>



<li>Chromeinance:<br>3色、カラー画像に対してノイズ除去を行います。チェックボタンを外すと輝度のみのノイズ除去<br>数値が上げると、カラーノイズ除去を強く行います。</li>



<li>Deep Space Smoothness (Shadows): 陰（かげ）部分の滑らかさ（？）<br>ノイズ除去を行った後に、天体構造がぼやけ始めたならば、この数値を上げていきます。</li>



<li>Use findstar (.lst) for Star Protection:<br>恒星にノイズ除去処理を行わないように、Sirilのfindstarを用います。<br>作業ディレクトリに、listファイルが作られます。内容は、Dynamic PSFの数値です。</li>



<li>Auto Starless Detection:<br>恒星が無い画像にノイズ除去を行う時にも、このチェックを使います。</li>
</ul>



<p class="wp-block-paragraph">プレビュー画像で良好なノイズ除去が得られたならば、&#8221;PROCESS&#8221;ボタンを押して、処理を全画面に適応します。</p>



<h2 class="wp-block-heading">Siril 画像処理の流れ</h2>



<p class="wp-block-paragraph">掲載画像は、画像処理中です。</p>



<p class="wp-block-paragraph">Name: <br>Processing: Siril 1.5 beta, VeraLux, StarNet2, Cosmic Clarity, GraXpert<br>Star Processing Python script: VeraLux Nox, Aberration Remover, Cosmic Clarity Sharpen<br>Galaxy Processing Python script: GraXpert, SyQon Prism</p>



<p class="wp-block-paragraph">Siril 1.5 betaを使って画像処理します。</p>



<ul id="block-38592a4e-9105-41a4-b0a4-204e6907d0bc" class="wp-block-list">
<li>Background Extraction: VeraLux Nox</li>



<li>SPCC (Gaia Spectrum: G2Vなど)</li>



<li>別名でファイル保存
<ul class="wp-block-list">
<li>恒星処理
<ul class="wp-block-list">
<li>Desaturate Stars: 今回は未使用</li>



<li>Aberrations Remover: 必要に応じて</li>



<li>Deconvolution: Cosmic Clarity Sharpening, Stellar Only (Stel_amt=0.24, Nstel_str=3)</li>



<li>Starnet 2</li>
</ul>
</li>



<li>銀河星雲処理
<ul class="wp-block-list">
<li>ノイズ除去 : VeraLux Silentium (Strength=　, 　)</li>



<li>Deconvolution: GraXpert (Strength=0.64)</li>



<li>Starnet 2</li>



<li>ストレッチ（HT,GHS）</li>



<li>Wavelets Transformation</li>



<li>ノイズ除去 : VeraLux Silentium (Strength=　, 　)</li>



<li>色補正: Color saturation 65%, threshold 1.00</li>
</ul>
</li>
</ul>
</li>



<li>Star recomposition</li>
</ul>
]]></content:encoded>
					
					<wfw:commentRss>https://globe3.ddns.net/wp/using-veralux-with-siril/feed/</wfw:commentRss>
			<slash:comments>0</slash:comments>
		
		
			</item>
	</channel>
</rss>
