[B! PyTorch] misshikiã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

misshiki id:misshiki

PyTorchã«é–¢ã™ã‚‹misshikiã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (498)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

Kubernetesã¨PyTorch Lightningã«ã‚ˆã‚‹åŒ»ç™‚AIé–‹ç™ºç’°å¢ƒã¨ãã®Tips - ã‚¨ãƒ ã‚¹ãƒªãƒ¼ãƒ†ãƒƒã‚¯ãƒ–ãƒã‚°
ã“ã¡ã‚‰ã¯ã‚¨ãƒ ã‚¹ãƒªãƒ¼ Advent Calendar 2024 5æ—¥ç›®ã®è¨˜äº‹ã§ã™ã€‚AIãƒ»æ©Ÿæ¢°å¦ç¿’ãƒãƒ¼ãƒ ã®ä¸‰æµ¦ (@mamo3gr) ãŒã€æ·±å±¤å¦ç¿’ã«åŸºã¥ãåŒ»ç™‚AIã®é–‹ç™ºç’°å¢ƒã¨Tipsã«ã¤ã„ã¦ãŠé€ã‚Šã—ã¾ã™ã€‚ DALL-E 3ãŒç”Ÿæˆã—ãŸã€Œå·¨å¤§ã‚¯ãƒ©ã‚¹ã‚¿ã«ã‚ˆã‚‹AIé–‹ç™ºã€ã®ã‚¤ãƒ¡ãƒ¼ã‚¸å›³ Kubernetesã‚¯ãƒ©ã‚¹ã‚¿ã‚ˆã‚‹åŒ»ç™‚AIã®é–‹ç™º ãƒ¢ãƒ‡ãƒ«ã®æ€§èƒ½æ”¹å–„ã«æ³¨åŠ›ã§ãã‚‹ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯PyTorch Lightning ãƒ‡ãƒã‚¤ã‚¹ç‰¹æœ‰ã®ã‚³ãƒ¼ãƒ‰ã‚’æ›¸ã‹ãªãã¦ã‚ˆã„ å¦ç¿’ãƒã‚°ã‚„ãƒã‚§ãƒƒã‚¯ãƒã‚¤ãƒ³ãƒˆã®ä¿å˜æ©Ÿèƒ½ãŒå……å®Ÿã—ã¦ã„ã‚‹ ç´°ã‹ãªãƒ¦ãƒ¼ãƒ†ã‚£ãƒªãƒ†ã‚£é–¢æ•°ã‚„ã‚¯ãƒ©ã‚¹ã®å®Ÿè£…ãŒã‚ã‚‹ ãƒãƒžã‚Šãƒã‚¤ãƒ³ãƒˆã¨ãƒ¯ãƒ¼ã‚¯ã‚¢ãƒ©ã‚¦ãƒ³ãƒ‰ å¦ç¿’ãƒã‚°ãƒ•ã‚¡ã‚¤ãƒ«ã‚’æ£ã—ãä¿å˜ã™ã‚‹ å¦ç¿’ã‚’ç¶™ç¶šå®Ÿè¡Œã™ã‚‹ ã¾ã¨ã‚ We are hiring !! ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢æŽ¡ç”¨ãƒšãƒ¼ã‚¸ã¯ã“ã¡ã‚‰ ã‚«ã‚¸ãƒ¥ã‚¢ãƒ«é¢è«‡ã‚‚ãŠæ°—è»½ã«ã©ã†ãž ã‚¤ãƒ³ã‚¿ãƒ¼ãƒ³ã‚‚å¸¸æ™‚å‹Ÿé›†ã—ã¦ã„ã¾ã™ Kubernetesã‚¯ãƒ©ã‚¹ã‚¿ã‚ˆã‚‹åŒ»ç™‚AIã®
misshiki 2024/12/06
â€œPyTroch Lightningã¯æ·±å±¤å¦ç¿’ã®ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã§PyTorchã®ãƒ©ãƒƒãƒ‘ãƒ¼ã§ã™ã€‚ è¤‡é›‘ã§æ‰‹é–“ãŒã‹ã‹ã‚Šãƒã‚°ã‚’åŸ‹ã‚è¾¼ã¿ãŒã¡ãªå®Ÿè£…ã‚’æ‰•ã„å‡ºã›ã‚‹ãŸã‚ã€æ©Ÿæ¢°å¦ç¿’ãƒ¢ãƒ‡ãƒ«ã®ãƒ‡ã‚¶ã‚¤ãƒ³ã‚„æ”¹å–„ã«æ³¨åŠ›ã§ãã¾ã™ã€‚ â€ ãã‚“ãªã«è‰¯ã„ã®ã‹ã€‚

PyTorch
ãƒªãƒ³ã‚¯
GitHub - kotaro-kinoshita/yomitoku: Yomitoku is an AI-powered document image analysis package designed specifically for the Japanese language.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert
misshiki 2024/11/27
â€œYomiToku ã¯æ—¥æœ¬èªžã«ç‰¹åŒ–ã—ãŸ AI æ–‡ç« ç”»åƒè§£æžã‚¨ãƒ³ã‚¸ãƒ³(Document AI)ã§ã™ã€‚ç”»åƒå†…ã®æ–‡å—ã®å…¨æ–‡ OCR ãŠã‚ˆã³ãƒ¬ã‚¤ã‚¢ã‚¦ãƒˆè§£æžæ©Ÿèƒ½ã‚’æœ‰ã—ã¦ãŠã‚Šã€ç”»åƒå†…ã®æ–‡å—æƒ…å ±ã‚„å›³è¡¨ã‚’èªè˜ã€æŠ½å‡ºã€å¤‰æ›ã—ã¾ã™ã€‚â€

OCR

Python

PyTorch
ãƒªãƒ³ã‚¯
Rustã¨PyTorchã§ä½œã‚‹æŽ¨è«–ã‚µãƒ¼ãƒãƒ¼
UV Study : Rust LTä¼šã§ç™ºè¡¨ã—ãŸå†…å®¹ã«ãªã‚Šã¾ã™ https://uniquevision.connpass.com/event/335781/
misshiki 2024/11/21
å…¨11ãƒšãƒ¼ã‚¸ã®ã‚¹ãƒ©ã‚¤ãƒ‰è³‡æ–™ã€‚Rustã§PyTorchã®ãƒ¢ãƒ‡ãƒ«ã‚’ãƒãƒ¼ãƒ‰ã—ã¦æŽ¨è«–ã™ã‚‹æ–¹æ³•ã‚’ç´¹ä»‹ã€‚

Rust

PyTorch
ãƒªãƒ³ã‚¯
PyTorch Deprecation of Conda Nightly Builds
Please see: [Announcement] Deprecating PyTorchâ€™s official Anaconda channel Â· Issue #138506 Â· pytorch/pytorch Â· GitHub PyTorch will stop publishing Anaconda packages that depend on Anacondaâ€™s default packages due to the high maintenance costs for conda builds which are not justifiable with the ROI we observe today (as seen in download discrepancies between PyPI vs. conda). We are directing users to
misshiki 2024/11/18
â€œPyTorch ã¯ã€Anaconda ã®ãƒ‡ãƒ•ã‚©ãƒ«ãƒˆ ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã«ä¾å˜ã™ã‚‹ Anaconda ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã®å…¬é–‹ã‚’åœæ¢ã—ã¾ã™ã€‚â€

Python

PyTorch
ãƒªãƒ³ã‚¯
[Announcement] Deprecating PyTorchâ€™s official Anaconda channel Â· Issue #138506 Â· pytorch/pytorch
misshiki 2024/11/18
â€œPyTorch ã®å…¬å¼ Anaconda ãƒãƒ£ãƒ³ãƒãƒ«ã®å»ƒæ¢â€ ãã‚‚ãã‚‚ç·ãƒ€ã‚¦ãƒ³ãƒãƒ¼ãƒ‰æ•°ã®5ï¼…æœªæº€ã ã£ãŸã¨ã€‚ä»Šå¾Œã¯PyPIï¼ˆpipï¼‰ã‚’ä½¿ã†ã‹ã€condaã‚’ä½¿ã„ç¶šã‘ã‚‹ãªã‚‰conda-forgeã‚ˆã†ã«ã¨ã€‚

PyTorch

Python
ãƒªãƒ³ã‚¯
0ã‹ã‚‰ä½œã‚‹LLMãƒ¼Llama
æœ¬è¨˜äº‹ã®å¯¾è±¡èªè€…ï¼š LLMï¼ˆå¤§è¦æ¨¡è¨€èªžãƒ¢ãƒ‡ãƒ«ï¼‰ã®è¤‡é›‘ãªæ§‹é€ ã‚„éšŽå±¤ã‚’ç†è§£ã—ã¦ã„ã‚‹ãŒã€ãã‚Œã‚’ã©ã®ã‚ˆã†ã«çµ„ã¿åˆã‚ã›ã‚‹ã‹ãŒåˆ†ã‹ã‚‰ãªã„äºº LlaMaãƒ¢ãƒ‡ãƒ«ã«é–¢ã™ã‚‹ã™ã¹ã¦ã®ã‚ªãƒšãƒ¬ãƒ¼ã‚¿ã¨ã‚¢ãƒ¼ã‚ãƒ†ã‚¯ãƒãƒ£ï¼ˆRMSNormã€ROPEã€SwiGLUã®å®Ÿè£…ã‚’å«ã‚€ï¼‰ã‚’ä¸€è¡Œãšã¤åˆ†è§£ã—ã¾ã™ã€‚ æœ¬è¨˜äº‹ã§ã¯huggingfaceã®ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã‚’ä½¿ç”¨ã—ã¦ãŠã‚‰ãšã€ã™ã¹ã¦pytorchã§å®Ÿè£…ã—ã¦ã„ã¾ã™ã€‚ã¾ãŸã€äº‹å‰å¦ç¿’æ¸ˆã¿ãƒ¢ãƒ‡ãƒ«ã‚‚ä½¿ç”¨ã—ã¦ã„ã¾ã›ã‚“ã€‚ ã‚¹ã‚¿ãƒ¼ãƒˆåœ°ç‚¹ã¯ã€Žæºæ°ç‰©èªžã€ã®åŽŸæ–‡ã§ã‚ã‚Šã€ã‚´ãƒ¼ãƒ«åœ°ç‚¹ã¯ã‚ãªãŸè‡ªèº«ãŒãƒˆãƒ¬ãƒ¼ãƒ‹ãƒ³ã‚°ã—ãŸå¤§è¦æ¨¡ãƒ¢ãƒ‡ãƒ«ã§ã™ã€‚ pytorchã‚’æº–å‚™ã—ã¦ãã ã•ã„ã€‚GPUãŒãªãã¦ã‚‚å¤§ä¸ˆå¤«ã§ã™ã€‚é‡è¦ãªã®ã¯LLMã®åŽŸç†ã‚’å¦ã¶ã“ã¨ã§ã‚ã‚Šã€ã“ã®æ–‡ç« ã‚’èªã‚“ã ã ã‘ã§æ–°ã—ã„å¤§è¦æ¨¡ãƒ¢ãƒ‡ãƒ«ã®ã‚¢ãƒ¼ã‚ãƒ†ã‚¯ãƒãƒ£ã‚’ä½œã‚Œã‚‹ã‚ã‘ã§ã¯ã‚ã‚Šã¾ã›ã‚“ã€‚ æœ¬è¨˜äº‹ã§ã¯ã€ã§ãã‚‹é™ã‚Šå¹³æ˜“ãªè¨€è‘‰ã‚’ä½¿ã£ã¦åŽŸç†ã‚’è§£èª¬ã—ã¦ã„ãã¾ã™ã€‚ åºæ–‡ æœ¬è¨˜äº‹ã®ã™ã¹ã¦ã®ã‚³ãƒ¼ãƒ‰ã¯Goo
misshiki 2024/10/29
â€œLlaMaã‚’æ§‹ç¯‰ã™ã‚‹å‰ã«ã€ã¾ãšã¯ã‚·ãƒ³ãƒ—ãƒ«ãªSeq2seqãƒ¢ãƒ‡ãƒ«ã‚’ä½œæˆã—ã€ãã®å¾Œã€ã“ã®åŸºæœ¬çš„ãªSeq2seqãƒ¢ãƒ‡ãƒ«ã«LlaMaã®ã‚ªãƒšãƒ¬ãƒ¼ã‚¿ã§ã‚ã‚‹RMSã€Ropeã€SwiGLUã‚’é †æ¬¡è¿½åŠ ã—ã¦ã„ãã¾ã™ã€‚æœ€çµ‚çš„ã«å®Œå…¨ãªLlaMaãƒ¢ãƒ‡ãƒ«ã‚’æ§‹ç¯‰ã—ã¾ã™ã€‚â€ PyTorchä½¿ç”¨ã€‚

è‡ªç„¶è¨€èªžå‡¦ç†

PyTorch
ãƒªãƒ³ã‚¯
GitHub - pytorch/ao: PyTorch native quantization and sparsity for training and inference
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert
misshiki 2024/09/27
â€œtorchao: ã‚«ã‚¹ã‚¿ãƒ ãƒ‡ãƒ¼ã‚¿åž‹ã¨æœ€é©åŒ–ã®ãŸã‚ã® PyTorch ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã€‚æŽ¨è«–ã¨ãƒˆãƒ¬ãƒ¼ãƒ‹ãƒ³ã‚°ã®ãŸã‚ã«é‡ã¿ã€å‹¾é…ã€ã‚ªãƒ—ãƒ†ã‚£ãƒžã‚¤ã‚¶ãƒ¼ã€ã‚¢ã‚¯ãƒ†ã‚£ãƒ™ãƒ¼ã‚·ãƒ§ãƒ³ã‚’é‡ååŒ–ãŠã‚ˆã³ã‚¹ãƒ‘ãƒ¼ã‚¹åŒ–ã—ã¾ã™ã€‚â€

PyTorch
ãƒªãƒ³ã‚¯
GitHub - mks0601/ExAvatar_RELEASE: Official PyTorch implementation of "Expressive Whole-Body 3D Gaussian Avatar", ECCV 2024.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert
misshiki 2024/09/24
â€œExpressive Whole-Body 3D Gaussian Avatar (ECCV 2024)ã®å†å®Ÿè£…â€

PyTorch

ã‚³ãƒ³ãƒ”ãƒ¥ãƒ¼ã‚¿ãƒ“ã‚¸ãƒ§ãƒ³
ãƒªãƒ³ã‚¯
GitHub - pytorch-labs/LeanRL: LeanRL is a fork of CleanRL, where selected PyTorch scripts optimized for performance using compile and cudagraphs.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert
misshiki 2024/09/24
â€œLeanRL ã¯ã€ä¸€èˆ¬çš„ãªå¼·åŒ–å¦ç¿’ (RL) ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã®å˜ä¸€ãƒ•ã‚¡ã‚¤ãƒ«ã® PyTorch ãƒ™ãƒ¼ã‚¹ã®å®Ÿè£…ã§æ§‹æˆã•ã‚Œã‚‹è»½é‡ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã§ã™ã€‚â€

PyTorch

æ°—ã«ãªã‚‹
ãƒªãƒ³ã‚¯
ã€TorchSharpã€‘C#ã§æ©Ÿæ¢°å¦ç¿’ã€€â‘ æ‰‹æ›¸ãæ–‡å—èªè˜ã‚¢ãƒ—ãƒªã®å®Ÿè£… - Qiita
0.ã€€ã¯ã˜ã‚ã« æ©Ÿæ¢°å¦ç¿’ã«é–¢ã™ã‚‹ãƒ—ãƒã‚°ãƒ©ãƒ ã¯ï¼ŒåŸºæœ¬çš„ã«pythonã«ã‚ˆã‚Šå®Ÿè£…ã•ã‚Œã¾ã™ï¼Žã—ã‹ã—C#(.Net)ã§ä½œæˆæ¸ˆã¿ã®ãƒ—ãƒã‚°ãƒ©ãƒ ã«æ©Ÿæ¢°å¦ç¿’ã®æ©Ÿèƒ½ã‚’è¿½åŠ ã—ãŸã„å ´åˆã‚„ï¼ŒGUIã‚‚ä½œæˆã—ãŸã„å ´åˆãªã©ï¼ŒC#(.Net)ã§æ©Ÿæ¢°å¦ç¿’ã®å‡¦ç†ã‚’å®Ÿè£…ã—ãªã‘ã‚Œã°ãªã‚‰ãªã„çŠ¶æ³ã‚‚ã‚ã‚‹ã¨æ€ã„ã¾ã™ï¼ŽC#ç”¨ã®æ©Ÿæ¢°å¦ç¿’ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã¯ã„ãã¤ã‹ã‚ã‚Šã¾ã™ãŒï¼Œ.NET Foundationã«çµ„ã¿è¾¼ã¾ã‚Œã¦ã„ã‚‹TorchSharpãŒç„¡é›£ãªé¸æŠžè‚¢ã ã¨æ€ã„ã¾ã™ï¼ŽPytorchãƒ™ãƒ¼ã‚¹ãªã®ã§ï¼ŒPytorchã«ç²¾é€šã—ã¦ã„ã‚‹æ–¹ã¯TorchSharpã‚‚ä½¿ã„ã“ãªã›ã‚‹ã¨æ€ã„ã¾ã™ï¼Žã—ã‹ã—ï¼ŒTorchSharpã«é–¢ã™ã‚‹æƒ…å ±ãŒå°‘ãªã‹ã£ãŸãŸã‚ï¼Œæœ¬ç¨¿ã§ã¯TorchSharp + C#(.Net 8.0)ã«ã‚ˆã‚‹æ•™å¸«ã‚ã‚Šå¦ç¿’ã®å®Ÿè£…ä¾‹ã¨ã—ã¦ï¼Œæ‰‹æ›¸ãæ•°å—ã®ã‚¯ãƒ©ã‚¹åˆ†é¡žã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ã‚’GUIä»˜ã§ä½œæˆã™ã‚‹ä¾‹ã‚’ç´¹ä»‹ã—ã¾ã™ï¼Žã€€æ•™å¸«ãƒ‡ãƒ¼ã‚¿ã®å¦ç¿’ã«åŠ ãˆã¦ï¼Œä»¥ä¸‹ã®ã‚ˆã†ã«GUIã«ã‚ãƒ£ãƒ³
misshiki 2024/09/03
â€œTorchSharp + C#(.Net 8.0)ã«ã‚ˆã‚‹æ•™å¸«ã‚ã‚Šå¦ç¿’ã®å®Ÿè£…ä¾‹ã¨ã—ã¦ï¼Œæ‰‹æ›¸ãæ•°å—ã®ã‚¯ãƒ©ã‚¹åˆ†é¡žã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ã‚’GUIä»˜ã§ä½œæˆã—ã¾ã™ï¼Žâ€

C#

PyTorch
ãƒªãƒ³ã‚¯
OpenVINOã§å„ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ï¼ˆONNXã€TensorFlowã€TFLiteã€PaddlePaddleï¼‰ã®é‡ã¿ã‚’èªã¿è¾¼ã‚“ã§æŽ¨è«–
ãŸã ã—ã€PyTorchã•ã‚“ã€ã‚ãªãŸã¯ãƒ€ãƒ¡ã§ã™ã€‚ç›´æŽ¥èªã¿è¾¼ã‚ã¾ã›ã‚“ðŸ‘» ä»Šå›žã®ã‚µãƒ³ãƒ—ãƒ«ã¯ã€Google Colaboratoryä¸Šã§ãŠè©¦ã—ã—ã¾ã™ã€‚ ã„ã¤ã®é–“ã«ã‹ã€Colaboratoryä¸Šã§ã®OpenVINOã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ãƒ»å®Ÿè¡Œã‚‚ç°¡å˜ã«ãªã£ã¦ã¦ã€åŠ©ã‹ã‚Šã¾ã™ã€‚ å€‹äººçš„ã«ã¯ã€PaddlePaddleã®é‡ã¿ã‚’ç›´æŽ¥èªã¿è¾¼ã‚ã‚‹ã®ãŒè‰¯ã„æ„Ÿã˜ã§ã™ã€‚ ãƒŽãƒ¼ãƒˆãƒ–ãƒƒã‚¯ã¯ä»¥ä¸‹ã®ãƒªãƒã‚¸ãƒˆãƒªã§å…¬é–‹ã—ã¦ã„ã¾ã™ã€‚ è©¦ã—ã¦ã¿ãŸã„æ–¹ã¯ã€ŒOpen in Colabã€ã‹ã‚‰ãƒŽãƒ¼ãƒˆãƒ–ãƒƒã‚¯ã‚’é–‹ã„ã¦ã€ä¸Šã‹ã‚‰é †ã«å®Ÿè¡Œã—ã¦ã„ã£ã¦ãã ã•ã„ã€‚ ä»¥é™ã¯å‡¦ç†ã®ç°¡å˜ãªèª¬æ˜Žã§ã™ã€‚ ãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ« OpenVINOã®ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã§ã™ã€‚ ä»Šå›žã®ã‚µãƒ³ãƒ—ãƒ«ã‚’å‹•ã‹ã™ã ã‘ã§ã‚ã‚Œã°ã€pipã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«1è¡Œã§æ¸ˆã¿ã¾ã™ã€‚
misshiki 2024/09/02
TensorFlow

PyTorch

ã‚³ãƒ³ãƒ”ãƒ¥ãƒ¼ã‚¿ãƒ“ã‚¸ãƒ§ãƒ³
ãƒªãƒ³ã‚¯
PyTorchã§CUDAãŒä½¿ãˆãªã„ã¨ãã«ç¢ºèªã™ã¹ãé …ç›®ï½œãŠã‚„ã•ã„
çµè«–ã‹ã‚‰PyTorchã§åˆ©ç”¨ã—ãŸã„CUDAãƒãƒ¼ã‚¸ãƒ§ãƒ³â‰¦CUDA ToolKitã®ãƒãƒ¼ã‚¸ãƒ§ãƒ³â‰¦GPUãƒ‰ãƒ©ã‚¤ãƒãƒ¼ã®å¯¾å¿œCUDAãƒãƒ¼ã‚¸ãƒ§ãƒ³ ã“ã®æ¡ä»¶ã‚’æº€ãŸã—ã¦ã„ãªã„ã¨PyTorchã§CUDAãŒåˆ©ç”¨ã§ãã¾ã›ã‚“ã€‚ ã©ã†ã—ã¦ã‚‚torch.cuda.is_available()ã®çµæžœãŒTrueã«ãªã‚‰ãªã„äººã‚’å¯¾è±¡ã«ã€ä»¥ä¸‹ç¢ºèªã™ã¹ãé …ç›®ã‚’è©³ã—ãèª¬æ˜Žã—ã¾ã™ã€‚ 1. CUDAæœ‰åŠ¹ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã®PyTorchã‚’ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã—ã¾ã—ãŸã‹ï¼Ÿå˜ç´”ã«pip3 install torchã‚’å®Ÿè¡Œã™ã‚‹ã¨CUDAç„¡åŠ¹(CPUæœ‰åŠ¹)ã®PyTorchãŒã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã•ã‚Œã¾ã™ã€‚CUDAæœ‰åŠ¹ã®PyTorchã‚’ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã™ã‚‹ãŸã‚ã«ã¯ã€https://pytorch.org/get-started/locally/Â ã“ã¡ã‚‰ã‹ã‚‰Run this Commandã‚’ã‚³ãƒ”ãƒšã—ã¦ã‚³ãƒžãƒ³ãƒ‰ãƒ©ã‚¤ãƒ³ã«è²¼ã‚Šä»˜ã‘ã¦ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã™ã‚‹å¿…è¦ãŒã‚ã‚Šã¾ã™ã€‚ç‰¹ã«PyCharm
misshiki 2024/08/23
PyTorch
ãƒªãƒ³ã‚¯
PyTorch is dead. Long live JAX. | Neel Gupta
PyTorch is dead. Long live JAX.The future of Deep Learning frameworks July 19, 2024 Â· 24 min read Assumed audience: ML researchers who frequently work with PyTorch, but are interested in trying out JAX or have yet to be convinced. IntroductionUsually, people start these â€˜critiquesâ€™ with a disclaimer that they are not trying to trash the framework, and talk about how itâ€™s a tradeoff. However, this
misshiki 2024/08/19
â€œã“ã®ãƒ–ãƒã‚°ãƒã‚¹ãƒˆã¯ã€PyTorchãŒå®Ÿéš›ã®ç ”ç©¶ãƒ¯ãƒ¼ã‚¯ãƒãƒ¼ãƒ‰ï¼ˆç‰¹ã«GPUä¸Šï¼‰ã«æœ€é©ã§ã‚ã‚‹ã¨ã„ã†ã€ã‚ˆãè¨€ã‚ã‚Œã‚‹ç¥žè©±ã‚’æ£ã™ãŸã‚ã®ã‚‚ã®ã§ã™ã€‚ãã‚Œã¯ã‚‚ã†äº‹å®Ÿã§ã¯ã‚ã‚Šã¾ã›ã‚“ã€‚â€

PyTorch

JAX
ãƒªãƒ³ã‚¯
FlexAttention: The Flexibility of PyTorch with the Performance of FlashAttention
In theory, Attention is All You Need. In practice, however, we also need optimized attention implementations like FlashAttention. Although these fused attention implementations have substantially improved performance and enabled long contexts, this efficiency has come with a loss of flexibility. You can no longer try out a new attention variant by writing a few PyTorch operators - you often need t
misshiki 2024/08/17
â€œFlexAttention: PyTorch ã®æŸ”è»Ÿæ€§ã¨ FlashAttention ã®ãƒ‘ãƒ•ã‚©ãƒ¼ãƒžãƒ³ã‚¹ã‚’èžåˆâ€

PyTorch
ãƒªãƒ³ã‚¯
GitHub - pytorch/torchchat: Run PyTorch LLMs locally on servers, desktop and mobile
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert
misshiki 2024/08/02
å…¬å¼ã€‚â€œtorchchat ã¯ã€LLM ã‚’ã‚·ãƒ¼ãƒ ãƒ¬ã‚¹ã«å®Ÿè¡Œã™ã‚‹æ©Ÿèƒ½ã‚’ç¤ºã™å°ã•ãªã‚³ãƒ¼ãƒ‰ãƒ™ãƒ¼ã‚¹ã§ã™ã€‚ç‹¬è‡ªã® (C/C++) ã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ (ãƒ‡ã‚¹ã‚¯ãƒˆãƒƒãƒ—ã¾ãŸã¯ã‚µãƒ¼ãƒãƒ¼) å†…ã€ãŠã‚ˆã³ iOS ã¨ Android ä¸Šã§ã€Python ã‚’ä½¿ç”¨ã—ã¦ LLM ã‚’å®Ÿè¡Œã§ãã¾ã™ã€‚â€

PyTorch

è‡ªç„¶è¨€èªžå‡¦ç†
ãƒªãƒ³ã‚¯
PyTorch 2.4 Release Blog
misshiki 2024/07/31
â€œPyTorch 2.4ã§ã¯ã€torch.compileã«Pythonã®æœ€æ–°ãƒãƒ¼ã‚¸ãƒ§ãƒ³(3.12)ã®ã‚µãƒãƒ¼ãƒˆãŒè¿½åŠ ã•ã‚Œã¾ã—ãŸã€‚ â€

PyTorch

Python
ãƒªãƒ³ã‚¯
Pythonå‘ã‘æ©Ÿæ¢°å¦ç¿’ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã€ŒPyTorch 2.4ã€ãŒãƒªãƒªãƒ¼ã‚¹
PyTorch 2.4ã«ãŠã‘ã‚‹ãƒ™ãƒ¼ã‚¿ç‰ˆã¨ã—ã¦ã¯ã€torch.compile()ã«ãŠã‘ã‚‹Python 3.12ã‚’ä½¿ç”¨ã—ãŸãƒ¢ãƒ‡ãƒ«æœ€é©åŒ–ã¸ã®å¯¾å¿œã‚„ã€CPUã§AOTInductorã‚’ä½¿ç”¨ã™ã‚‹éš›ã«ãƒ•ãƒªãƒ¼ã‚ºãƒ•ãƒ©ã‚°ã‚’ã‚ªãƒ³ã«ã§ãã‚‹ã‚ˆã†ã«ãªã‚Šã€AOTInductorã§Inductor CPPãƒãƒƒã‚¯ã‚¨ãƒ³ãƒ‰ã¨åŒã˜ã‚ªãƒšãƒ¬ãƒ¼ã‚·ãƒ§ãƒ³ã‚·ãƒŠãƒªã‚ªã‚»ãƒƒãƒˆã‚’ã‚«ãƒãƒ¼ã—ã¦ã€åŒç‰ã®ãƒ‘ãƒ•ã‚©ãƒ¼ãƒžãƒ³ã‚¹ã®å®Ÿç¾ã‚’å¯èƒ½ã«ã™ã‚‹æ©Ÿèƒ½ãŒè¿½åŠ ã•ã‚Œã¦ã„ã‚‹ã€‚ ã‚ã‚ã›ã¦ã€Pythonã®çµ„ã¿è¾¼ã¿æ¼”ç®—åã®ã‚ˆã†ã«å‹•ä½œã™ã‚‹ã‚«ã‚¹ã‚¿ãƒ æ¼”ç®—åã‚’ä½¿ç”¨ã—ã¦ã€PyTorchã‚’æ‹¡å¼µã™ã‚‹ã“ã¨ãŒå®¹æ˜“ã«ãªã‚‹é«˜ãƒ¬ãƒ™ãƒ«Pythonã‚«ã‚¹ã‚¿ãƒ æ¼”ç®—åAPIã®è¿½åŠ ã‚„ã€åˆæœŸåŒ–æ™‚é–“ã‚’å¤§å¹…ã«çŸç¸®ã—ã¦ã‚¹ã‚±ãƒ¼ãƒ©ãƒ“ãƒªãƒ†ã‚£ã‚’å‘ä¸Šã™ã‚‹ã€TCPStoreç”¨ã®æ–°ãŸãªãƒ‡ãƒ•ã‚©ãƒ«ãƒˆã‚µãƒ¼ãƒãƒãƒƒã‚¯ã‚¨ãƒ³ãƒ‰ã¨ãªã‚‹libuvã®å°Žå…¥ãŒè¡Œã‚ã‚ŒãŸã€‚ ãƒ—ãƒãƒˆã‚¿ã‚¤ãƒ—æ©Ÿèƒ½ã¨ã—ã¦ã¯ã€dim-0ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã”ã¨ã®ã‚·ãƒ£ãƒ¼ãƒ‡ã‚£ãƒ³ã‚°ã‚’ä½¿ç”¨ã—ã¦FSDP1ã®ãƒ•
misshiki 2024/07/31
â€œPythonå‘ã‘ã®æ©Ÿæ¢°å¦ç¿’ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã§ã‚ã‚‹PyTorchã®é–‹ç™ºãƒãƒ¼ãƒ ã¯ã€æœ€æ–°ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã§ã‚ã‚‹ã€ŒPyTorch 2.4ã€ã‚’7æœˆ24æ—¥ï¼ˆç¾åœ°æ™‚é–“ï¼‰ã«ãƒªãƒªãƒ¼ã‚¹ã—ãŸã€‚â€

PyTorch
ãƒªãƒ³ã‚¯
GitHub - SylphAI-Inc/LightRAG: The PyTorch Library for LLM Applications.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert
misshiki 2024/07/09
â€œå¤§è¦æ¨¡è¨€èªžãƒ¢ãƒ‡ãƒ«ã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³å‘ã‘ PyTorch ãƒ©ã‚¤ãƒ–ãƒ©ãƒª âš¡ LightRAGã¯ã€Retriever-Agent-Generator (RAG)ãƒ‘ã‚¤ãƒ—ãƒ©ã‚¤ãƒ³ã®æ§‹ç¯‰ã¨æœ€é©åŒ–ã®ä¸¡æ–¹ã«ãŠã„ã¦é–‹ç™ºè€…ã‚’æ”¯æ´â€

PyTorch

è‡ªç„¶è¨€èªžå‡¦ç†
ãƒªãƒ³ã‚¯
PyTorché–¢æ•°åã®æœ«å°¾ã‚¢ãƒ³ãƒ€ãƒ¼ãƒãƒ¼`_`ã®æ„å‘³ - BioErrorLog Tech Blog
in-placeå‡¦ç†ã§ã‚ã‚‹ã“ã¨ã‚’æ„å‘³ã—ã¾ã™ã€‚ ã¯ã˜ã‚ã« PyTorché–¢æ•°åã®æœ«å°¾ã‚¢ãƒ³ãƒ€ãƒ¼ãƒãƒ¼_ æ„å‘³ å…·ä½“ä¾‹ ãŠã‚ã‚Šã« å‚è€ƒ ã¯ã˜ã‚ã« PyTorchã‚’è§¦ã£ã¦ã„ã‚‹ã¨ã€ã—ã°ã—ã°ã‚¢ãƒ³ãƒ€ãƒ¼ãƒãƒ¼_ã‚’æŽ¥å°¾èªžã¨ã™ã‚‹é–¢æ•°ã‚’è¦‹ã‹ã‘ã¾ã™ã€‚ kaiming_normal_ add_ etc... åˆè¦‹ã§æ„å‘³ã‚’çŸ¥ã‚‰ãªã‹ã£ãŸã®ã§å‚™å¿˜éŒ²ãƒ¡ãƒ¢ã§ã™ã€‚ PyTorché–¢æ•°åã®æœ«å°¾ã‚¢ãƒ³ãƒ€ãƒ¼ãƒãƒ¼_ æ„å‘³ PyTorché–¢æ•°åã®æœ«å°¾ã‚¢ãƒ³ãƒ€ãƒ¼ãƒãƒ¼_ã¯ã€in-placeå‡¦ç†ã‚’æ„å‘³ã—ã¾ã™ã€‚ in-placeå‡¦ç†ã¨ã¯ã€å…ƒã®ãƒ‡ãƒ¼ã‚¿ã‚’ç›´æŽ¥å¤‰æ›´ã™ã‚‹å‡¦ç†ã®ã“ã¨ã§ã™ã€‚ In-place operations Operations that have a _ suffix are in-place. For example: x.copy_(y), x.t_(), will change x. Ref. Tensors â€” PyTorch Tu
misshiki 2024/06/13
â€œPyTorché–¢æ•°åã®æœ«å°¾ã‚¢ãƒ³ãƒ€ãƒ¼ãƒãƒ¼_ã¯ã€in-placeå‡¦ç†ã‚’æ„å‘³ã—ã¾ã™ã€‚â€

PyTorch
ãƒªãƒ³ã‚¯
Ryeã‚’ç”¨ã„ãŸPyTorchãŠã‚ˆã³PyGç’°å¢ƒæ§‹ç¯‰
è¦ç´„ Windowsä¸Šã®CUDAç’°å¢ƒã«ãŠã„ã¦ã€Ryeã‚’ç”¨ã„ã¦PyTorchãŠã‚ˆã³PyG (PyTorch Geometric) ã®ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã‚’ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«ã™ã‚‹ã“ã¨ãŒã§ããŸã€‚pyproject.tomlã«ã‚½ãƒ¼ã‚¹ã‚’è¨å®šã™ã‚‹ã“ã¨ãŒå¿…è¦ã¨ãªã‚‹ã€‚ Ryeã«ã¤ã„ã¦ Ryeã¯Pythonã®ãƒãƒ¼ã‚¸ãƒ§ãƒ³ç®¡ç†ã¨ãƒ©ã‚¤ãƒ–ãƒ©ãƒªç®¡ç†ã®ä¸¡æ–¹ã‚’1ã¤ã§è¡Œãˆã‚‹ãƒ„ãƒ¼ãƒ«ã€‚Rustã§å†…éƒ¨å®Ÿè£…ã•ã‚Œã¦ã„ã‚‹ã€‚ã“ã“ã§ã¯ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«æ–¹æ³•ã«ã¯è§¦ã‚Œãªã„ã€‚ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ«æ¸ˆã¿ã§ã‚ã‚‹ã¨ã—ã¦é€²ã‚ã‚‹ã€‚ Rye CUDAç’°å¢ƒã®æ§‹ç¯‰ ä»¥ä¸‹ãŒå¿…è¦ã¨ãªã‚‹ã€‚ NVIDIAãƒ‡ã‚£ã‚¹ãƒ—ãƒ¬ã‚¤ãƒ‰ãƒ©ã‚¤ãƒãƒ¼ã®ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ« NVIDIA CUDA Toolkit ã®ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ« NVIDIA cuDNN ã®ã‚¤ãƒ³ã‚¹ãƒˆãƒ¼ãƒ« ã“ã®3ã¤ã¯çµ„ã¿åˆã‚ã›ã®ç›¸æ€§ãŒã‚ã‚Šã€ä»¥ä¸‹ã®ãƒšãƒ¼ã‚¸ã§ã‚µãƒãƒ¼ãƒˆã•ã‚Œã¦ã„ã‚‹çµ„ã¿åˆã‚ã›ãŒè¨˜è¼‰ã•ã‚Œã¦ã„ã‚‹ã€‚ Support Matrix ã“ã“ã§ã¯ã€æœ€æ–°ã®ãƒ‰ãƒ©ã‚¤ãƒãƒ¼ã¨ CUDA
misshiki 2024/06/13
Python

PyTorch
ãƒªãƒ³ã‚¯
1 2 3 4 5 6 7 8 9 10 æ¬¡ã®ãƒšãƒ¼ã‚¸