[B! llm] ktakeda47ã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

ktakeda47 id:ktakeda47

llmã«é–¢ã™ã‚‹ktakeda47ã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (12)

${{author_name}}$

{{{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}}

æ—¥æœ¬èªžã«ç‰¹åŒ–ã—ãŸOCRã€æ–‡æ›¸ç”»åƒè§£æžPythonãƒ‘ãƒƒã‚±ãƒ¼ã‚¸ã€ŒYomiTokuã€ã‚’å…¬é–‹ã—ã¾ã—ãŸï½œKotaro.Kinoshita
ã¯ã˜ã‚ã«æœ€è¿‘ã€LLMã¸ã®RAGã‚’ç”¨ã„ãŸæ–‡æ›¸ãƒ‡ãƒ¼ã‚¿ã®é€£æºç‰ã‚’ç›®çš„ã«æµ·å¤–ã‚’ä¸å¿ƒã«OCRã‚„æ–‡æ›¸ç”»åƒè§£æžæŠ€è¡“ã«é–¢é€£ã™ã‚‹æ–°ã—ã„ã‚µãƒ¼ãƒ“ã‚¹ãŒæ´»ç™ºã«ãƒªãƒªãƒ¼ã‚¹ã•ã‚Œã¦ã„ã¾ã™ã€‚ ã—ã‹ã—ã€ãã®å¤šãã¯æ—¥æœ¬èªžã‚’ãƒ¡ã‚¤ãƒ³ã‚¿ãƒ¼ã‚²ãƒƒãƒˆã«é–‹ç™ºã•ã‚Œã¦ã„ã‚‹ã‚ã‘ã§ã¯ã‚ã‚Šã¾ã›ã‚“ã€‚æ—¥æœ¬èªžæ–‡æ›¸ã¯ã€è‹±æ•°å—ã«åŠ ãˆã¦ã€ã²ã‚‰ãŒãªã€æ¼¢å—ã€è¨˜å·ãªã©æ•°åƒç¨®é¡žã®æ–‡å—ã‚’è˜åˆ¥ã™ã‚‹å¿…è¦ãŒã‚ã£ãŸã‚Šã€ç¸¦æ›¸ããªã©æ—¥æœ¬èªžãƒ‰ã‚ãƒ¥ãƒ¡ãƒ³ãƒˆç‰¹æœ‰ã®ãƒ¬ã‚¤ã‚¢ã‚¦ãƒˆã«å¯¾å‡¦ã™ã‚‹å¿…è¦ãŒã‚ã£ãŸã‚Šã¨æ—¥æœ¬èªžç‰¹æœ‰ã®é›£ã—ã•ãŒã‚ã‚Šã¾ã™ã€‚ ã§ã™ãŒã€ä»Šå¾Œã€æµ·å¤–ã®é–‹ç™ºè€…ãŒã“ã‚Œã‚‰ã®èª²é¡Œã«å¯¾å‡¦ã™ã‚‹ãŸã‚ã€æ—¥æœ¬ã®ãƒ‰ã‚ãƒ¥ãƒ¡ãƒ³ãƒˆç”»åƒè§£æžã«ç‰¹åŒ–ã—ãŸã‚‚ã®ã‚’ãƒªãƒªãƒ¼ã‚¹ã™ã‚‹å¯èƒ½æ€§ã¯ä½Žãã€ã‚„ã¯ã‚Šè‡ªå›½ã®è¨€èªžå‘ã‘ã®ã‚µãƒ¼ãƒ“ã‚¹ã¯è‡ªå›½ã®ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ãŒé–‹ç™ºã™ã¹ãã ã¨ç†è€…ã¯è€ƒãˆã¦ã„ã¾ã™ã€‚ ã‚‚ã¡ã‚ã‚“ã€Azure Document Intelligenceã‚’ã¯ã˜ã‚ã¨ã—ãŸã€ã‚¯ãƒ©ã‚¦ãƒ‰ã‚µãƒ¼ãƒ“ã‚¹ã®ãƒ‰ã‚ãƒ¥ãƒ¡ãƒ³ãƒˆè§£æžã‚µãƒ¼ãƒ“ã‚¹ã¯ã‚ã‚Šã¾ã™ãŒã€ã‚¯ãƒ©ã‚¦ãƒ‰ã‚’åˆ©ç”¨ã§ããªã„ãƒ¦
ktakeda47 2024/11/27
llm

library

python

ai

development

technology
ãƒªãƒ³ã‚¯
ã€Œç¾åœ¨ã®LLMã«çœŸã®æŽ¨è«–ã¯å›°é›£ã€â”€â”€Appleã®ç ”ç©¶è€…ã‚‰ãŒè«–æ–‡ç™ºè¡¨
ç±³Appleã®AIç ”ç©¶è€…ã‚‰ã¯10æœˆ7æ—¥ï¼ˆç¾åœ°æ™‚é–“ï¼‰ã€ã€ŒGSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Modelsã€ï¼ˆLLMã«ãŠã‘ã‚‹æ•°å¦çš„æŽ¨è«–ã®é™ç•Œã‚’ç†è§£ã™ã‚‹ï¼‰ã¨ã„ã†è«–æ–‡ã‚’ç™ºè¡¨ã—ãŸã€‚ ã“ã®è«–æ–‡ã¯ã€LLMï¼ˆå¤§è¦æ¨¡è¨€èªžãƒ¢ãƒ‡ãƒ«ï¼‰ãŒã€æœ¬å½“ã«äººé–“ã®ã‚ˆã†ã«è«–ç†çš„ã«è€ƒãˆã¦å•é¡Œã‚’è§£ã‘ã‚‹ã®ã‹ã€ã¨ã„ã†ç–‘å•ã‚’æ¤œè¨¼ã—ã¦ã„ã‚‹ã€‚çµè«–ã¨ã—ã¦ã¯ã€LLMã¯ä»Šã®ã¨ã“ã‚ã€è¡¨é¢çš„ãªãƒ‘ã‚¿ãƒ¼ãƒ³ã‚’çœŸä¼¼ã¦ç”ãˆã‚’å‡ºã—ã¦ã„ã‚‹ã ã‘ã§ã€çœŸã®æŽ¨è«–èƒ½åŠ›ã¯æŒã£ã¦ã„ãªã„ã¨ä¸»å¼µã—ã¦ã„ã‚‹ã€‚ ç ”ç©¶è€…ã‚‰ã¯ã€ã“ã‚Œã‚‰ã®å•é¡Œç‚¹ã‚’æ¤œè¨¼ã™ã‚‹ãŸã‚ã«ã€ã€ŒGSM-Symbolicã€ã¨ã„ã†æ–°ã—ã„ãƒ†ã‚¹ãƒˆæ–¹æ³•ã‚’é–‹ç™ºã—ãŸã€‚ã“ã‚Œã¯ã€LLMã®æ•°å¦çš„æŽ¨è«–èƒ½åŠ›ã‚’è©•ä¾¡ã™ã‚‹ãŸã‚ã®ãƒ™ãƒ³ãƒãƒžãƒ¼ã‚¯ãƒ‡ãƒ¼ã‚¿ã‚»ãƒƒãƒˆã€ŒGSM8Kã€ã‚’æ”¹è‰¯ã—ã€å•é¡Œã®è¡¨ç¾ã‚„æ•°å—ã‚’æŸ”è»Ÿã«å¤‰ãˆã‚‰ã‚Œã‚‹ã‚ˆã†ã«ã—ãŸ
ktakeda47 2024/10/13
ã‚ªãƒ¬ã˜ã‚ƒã‚“ã€‚ "è¡¨é¢çš„ãªãƒ‘ã‚¿ãƒ¼ãƒ³ã‚’çœŸä¼¼ã¦ç”ãˆã‚’å‡ºã—ã¦ã„ã‚‹ã ã‘"

web

society

llm

apple

ai

development

science
ãƒªãƒ³ã‚¯
GitHub Modelsã®ã”ç´¹ä»‹ï¼šGitHubä¸Šã«æ–°ä¸–ä»£AIã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ã‚’
Author Thomas Dohmke ç§ãŸã¡ã¯ã€é©åˆ‡ãªãƒ„ãƒ¼ãƒ«ã¨ãƒˆãƒ¬ãƒ¼ãƒ‹ãƒ³ã‚°ãŒã‚ã‚Œã°ã€ã™ã¹ã¦ã®é–‹ç™ºè€…ãŒAI ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ã«ãªã‚Œã‚‹ã¨ä¿¡ã˜ã¦ã„ã¾ã™ã€‚GitHub Modelsã¯ã€GitHub Codespacesã§ãƒ¢ãƒ‡ãƒ«ã‚’ä½¿ã£ãŸãƒ—ãƒ¬ã‚¤ã‚°ãƒ©ã‚¦ãƒ³ãƒ‰ã‹ã‚‰ã‚³ãƒ¼ãƒ‰ç”Ÿæˆã€AzureçµŒç”±ã§ã®ãƒ‡ãƒ—ãƒã‚¤ã¾ã§ã€ã„ã‹ã«ã‚·ãƒ³ãƒ—ãƒ«ã«è¡Œãˆã‚‹ã‹ã‚’ãŠè¦‹ã›ã—ã¾ã™ã€‚é™å®šãƒ‘ãƒ–ãƒªãƒƒã‚¯ãƒ™ãƒ¼ã‚¿ã¸ã®ã‚µã‚¤ãƒ³ã‚¢ãƒƒãƒ—ã¯ã“ã¡ã‚‰ã‹ã‚‰ã€‚ å®¶åºç”¨ã‚³ãƒ³ãƒ”ãƒ¥ãƒ¼ã‚¿ã®é»Žæ˜ŽæœŸã‹ã‚‰ã€é–‹ç™ºè€…ã®å‰µä½œã®ä¸»æµã¯ã‚³ãƒ¼ãƒ‰ã‚’ä½¿ã£ã¦ã‚½ãƒ•ãƒˆã‚¦ã‚§ã‚¢ã‚’ãƒ“ãƒ«ãƒ‰ã€ã‚«ã‚¹ã‚¿ãƒžã‚¤ã‚ºã€ãƒ‡ãƒ—ãƒã‚¤ã™ã‚‹ã“ã¨ã§ã—ãŸã€‚AIã®æ™‚ä»£ã¨ãªã£ãŸä»Šæ—¥ã€æ©Ÿæ¢°å¦ç¿’ãƒ¢ãƒ‡ãƒ«ã‚’æ´»ç”¨ã™ã‚‹èƒ½åŠ›ã¨ã„ã†ã€äºŒæ¬¡çš„ã‹ã¤åŒæ§˜ã«é‡è¦ãªå‰µé€ æ§˜å¼ãŒæ€¥é€Ÿã«å°é ã—ã¦ã„ã¾ã™ã€‚ãƒãƒƒã‚¯ã‚¨ãƒ³ãƒ‰ã¨ãƒ•ãƒãƒ³ãƒˆã‚¨ãƒ³ãƒ‰ã®ã‚³ãƒ¼ãƒ‰ã¨1ã¤ä»¥ä¸Šã®ãƒ¢ãƒ‡ãƒ«ã‚’å«ã‚€ãƒ•ãƒ«ã‚¹ã‚¿ãƒƒã‚¯ã§ã€ç”ŸæˆAIã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ã‚’æ§‹ç¯‰ã™ã‚‹ã“ã¨ãŒå¢—ãˆã¦ã„ãªãŒã‚‰ã€ã‚½ãƒ•ãƒˆã‚¦ã‚§ã‚¢é–‹ç™ºè€…ã®å¤šãã¯ã‚ªãƒ¼ãƒ—ãƒ³ãƒ¢ãƒ‡ãƒ«ã‚„ã‚¯ãƒãƒ¼ã‚º
ktakeda47 2024/08/03
github

ai

gpt

llm

programming

development
ãƒªãƒ³ã‚¯
æœ€è¿‘ãƒãƒ¼ã‚«ãƒ«LLMãŒã‚¢ãƒ„ã„ã‚‰ã—ã„
æœ€è¿‘ã€ãƒãƒ¼ã‚«ãƒ«LLMãŒã‚¢ãƒ„ããªã£ã¦ã„ã‚‹ã¨ã„ã†è©±ã‚’twitterã§ãƒãƒ©ãƒ›ãƒ©è¦‹ã‹ã‘ã‚‹ã€‚ ãƒãƒ¼ã‚«ãƒ«LLMã£ã¦ä½•ã˜ã‚ƒï¼Ÿã¨ã„ã†ã¨ã€ã‚ªãƒ¼ãƒ—ãƒ³ã«å…¬é–‹ã•ã‚Œã¦ã„ã‚‹ãƒ¢ãƒ‡ãƒ«ã®ã‚¦ã‚¨ã‚¤ãƒˆã‚’DLã—ã¦ãã¦æ‰‹å…ƒã®PCä¸Šã§æŽ¨è«–ã•ã›ã‚‹äº‹ã§ã‚ã‚‹ã€‚ ã‚ªãƒ¼ãƒ—ãƒ³ãªAIãƒ¢ãƒ‡ãƒ«ã¨ã¯é€†ã®å˜åœ¨ã¨ã—ã¦ã€ãƒ¢ãƒ‡ãƒ«ãŒDLã§ããªã„ã‚¯ãƒãƒ¼ã‚ºãªAIãƒ¢ãƒ‡ãƒ«ã‚‚ã‚ã‚‹ã€‚ OpenAIã‚„Anthropicã®ã‚ˆã†ãªæœ€å…ˆç«¯AIä¼æ¥ã®ã‚¯ãƒãƒ¼ã‚ºãªAIãƒ¢ãƒ‡ãƒ«ã«æ¯”ã¹ã¦ã€ã‚ªãƒ¼ãƒ—ãƒ³ã«å…¬é–‹ã•ã‚Œã¦ã„ã‚‹ã‚ªãƒ¼ãƒ—ãƒ³ãªAIãƒ¢ãƒ‡ãƒ«ã®æ€§èƒ½ã¯ä»Šã§ã‚‚ã‹ãªã‚Šå¾Œã‚Œã‚’å–ã£ã¦ã„ã‚‹ã€‚ ã ã‹ã‚‰åŽ»å¹´ã®é–“ã¯ã‚ãã¾ã§AIã®æœ¬å‘½ã¯ChatGPTã®ã‚ˆã†ãªã‚¯ãƒãƒ¼ã‚ºãƒ¢ãƒ‡ãƒ«ã§ã‚ã£ã¦ã€ã‚ªãƒ¼ãƒ—ãƒ³ãªAIãƒ¢ãƒ‡ãƒ«ãªã‚“ã¦çœ¼ä¸ã«ã‚ã‚Šã¾ã›ã‚“ã¿ãŸã„ãªé¢¨æ½®ãŒç„¡ãã¯ãªã‹ã£ãŸã€‚ã ãŒæœ€è¿‘ã¯é¢¨å‘ããŒå°‘ã€…å¤‰ã‚ã£ã¦ãã¦ã„ã‚‹ã€‚ GPTã®APIé«˜ã„å•é¡Œ ï¼† OpenAIãŒAIãƒ™ãƒ³ãƒãƒ£ãƒ¼çš†æ®ºã—ã«ã—ã¦ã—ã¾ã†å•é¡Œ ã¾ãšã€Œçµå±€GPTã®APIã‚’å©ã„ã¦ã‚µãƒ¼ãƒ“ã‚¹é‹å–¶ã—ã¦æˆåŠŸã—
ktakeda47 2024/05/21
"ãã‚‚ãã‚‚ã€ã‚¿ãƒ€ã§æŽ¨è«–ã—ãŸã„ã ã‘ãªã‚‰å¤§æŠµã®äººã«ã¨ã£ã¦ã¯Command R+ã‚„Gemini1.5Flashã®ç„¡æ–™APIæž ã‚’å©ã„ã¨ã‘ã°ååˆ†ãªçŠ¶æ³ã‹ã‚‚ã—ã‚Œãªã„ã€‚ç¾çŠ¶ã§ã¯ãƒãƒ¼ã‚«ãƒ«LLMã‚ˆã‚Šã‚‚Gemini1.5Flashã®æ–¹ãŒå¤šåˆ†æ€§èƒ½é«˜ã„ã®ã ã‹ã‚‰ã€‚"

llm

mac

chatgpt

ai

web

society

programming

openai

google

gemini
ãƒªãƒ³ã‚¯
ãªãœæˆ‘ã€…ã¯ GitHub Copilot Enterprise ã®å°Žå…¥ã‚’è¦‹é€ã£ãŸã®ã‹ - ä¸€ä¼‘.com Developers Blog
CTO å®¤ã®æ©ç”°ã§ã™ã€‚ ä»Šå›žã¯ GitHub Copilot Enterprise ã‚’è©•ä¾¡ã—ã¦ã¿ã¦ã€ç¾æ™‚ç‚¹ã§ã¯ã¾ã æŽ¡ç”¨ã—ãªã„ã“ã¨ã‚’æ±ºã‚ãŸã€ã¨ã„ã†ãŠè©±ã‚’ã”ç´¹ä»‹ã—ãŸã„ã¨æ€ã„ã¾ã™ã€‚ ãã£ã‹ã‘ ã¨ã‚ã‚‹ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ãŒ Slack ã§è‡ªèº«ã® times ãƒãƒ£ãƒãƒ«ã«æ™‚é›¨å ‚ã•ã‚“ã® GitHub Copilot Enterprise ã®ã‚¹ã‚¹ãƒ¡ã¨ã„ã†è¨˜äº‹ã‚’æŠ•ç¨¿ã—ãŸã“ã¨ãŒç™ºç«¯ã§ã—ãŸã€‚ç‰¹ã«æ„Ÿæƒ³ã¯ãªã URL ã« ðŸ‘€ ã ã‘ãŒæ·»ãˆã‚‰ã‚Œã¦ã„ãŸã®ã§ã€å¾Œã§è¦‹ã‚‹ãã‚‰ã„ã®ãƒ¡ãƒ¢ã ã£ãŸã‚“ã ã¨æ€ã„ã¾ã™ã€‚ ãã‚Œã‚’è¦‹ã‹ã‘ãŸåˆ¥ã®ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ãŒæŠ€è¡“é›‘è«‡ãƒãƒ£ãƒãƒ«ã«ãã®æŠ•ç¨¿ã‚’å…±æœ‰ã—ã¦ã€ã“ã‚Œã¯å‡„ãã†ã¨è©±é¡Œã‚’å‘ã‘ãŸã¨ã“ã‚ã€CTO ã®ã€Œè©•ä¾¡ã—ã¦ã¿ã‚‹ï¼Ÿã€ã®ä¸€è¨€ã§ã€æœ‰å¿—ãŒé›†ã£ã¦è©•ä¾¡ãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆãŒå§‹ã¾ã‚Šã¾ã—ãŸã€‚ é›‘è«‡ãƒãƒ£ãƒãƒ«ã§ãã£ã‹ã‘ã¨ãªã‚‹æŠ•ç¨¿ãŒå…±æœ‰ã•ã‚Œã¦ã‹ã‚‰ã€30åˆ†è¶³ã‚‰ãšã®å‡ºæ¥äº‹ã§ã—ãŸï¼ˆç¬‘ï¼‰ã€‚ ã“ã®è©±é¡ŒãŒå‡ºãŸã®ã¯é‡‘æ›œæ—¥ã§ã—ãŸãŒã€é€±æ˜Žã‘æ—©ã€…ã«ç¨Ÿè°ã‚’çµ‚ãˆã€ç«æ›œ
ktakeda47 2024/04/16
"ã©ã®ã‚ˆã†ãªãƒ•ã‚¡ã‚¤ãƒ«ã«ã©ã®ã‚ˆã†ãªå¤‰æ›´ã‚’ãŠã“ãªã£ãŸã‹ã¨ã„ã† what ã®æƒ…å ±ã¯ã†ã¾ãè¦ç´„ã—ã¦ãã‚Œã¾ã™ã€‚ ã—ã‹ã—ã€ãã® PR ã®å¤‰æ›´ãŒå¿…è¦ã¨ãªã£ãŸèƒŒæ™¯ã‚„å¤‰æ›´ã®æ„å›³ã¨ã„ã£ãŸ why ã®æƒ…å ±ã¯æœŸå¾…ã—ãŸã»ã©ã«ã¯ç››ã‚Šè¾¼ã¾ã‚Œã¾ã›ã‚“"

llm

github

ai

programming

development

business
ãƒªãƒ³ã‚¯
Googleã€ã‚ªãƒ¼ãƒ—ãƒ³ãªç”ŸæˆAIãƒ¢ãƒ‡ãƒ«ã€ŒGemmaã€å…¬é–‹ã€€å•†ç”¨åˆ©ç”¨OKã€Geminiã¨åŒã˜æŠ€è¡“ã®è»½é‡LLM
ã¾ãŸã€Gemmaã®äº‹å‰ãƒˆãƒ¬ãƒ¼ãƒ‹ãƒ³ã‚°æ¸ˆã¿ãƒ¢ãƒ‡ãƒ«ã§ã¯ã€å¦ç¿’ãƒ‡ãƒ¼ã‚¿ã‹ã‚‰ç‰¹å®šã®å€‹äººæƒ…å ±ã‚„ãã®ä»–ã®æ©Ÿå¯†ãƒ‡ãƒ¼ã‚¿ã‚’é™¤å¤–ã—ã¦ã„ã‚‹ã¨å®‰å…¨æ€§ã‚‚ã‚¢ãƒ”ãƒ¼ãƒ«ã€‚é–‹ç™ºè€…ã‚„ç ”ç©¶è€…å‘ã‘ã«ã€å®‰å…¨ã§è²¬ä»»ã‚ã‚‹AIã‚¢ãƒ—ãƒªã‚±ãƒ¼ã‚·ãƒ§ãƒ³ã‚’æ§‹ç¯‰ã§ãã‚‹ã¨ã„ã†ãƒ„ãƒ¼ãƒ«ã‚ãƒƒãƒˆã€ŒResponsible Generative AI Toolkitã€ã‚‚ä½µã›ã¦å…¬é–‹ã—ã¦ã„ã‚‹ã€‚ é–¢é€£è¨˜äº‹ Googleã€â€œç¾è¡Œæœ€å¼·â€ã®ç”ŸæˆAIç™ºè¡¨ã€€æœˆ2900å††ã§åˆ©ç”¨å¯ã€€ãƒãƒ£ãƒƒãƒˆAIã‚µãƒ¼ãƒ“ã‚¹ã¯Bardâ†’Geminiã«åˆ·æ–° ç±³Googleã¯2æœˆ8æ—¥ï¼ˆç¾åœ°æ™‚é–“ï¼‰ã€ã€Œç¾è¡Œæœ€å¼·ã€ã‚’ã†ãŸã†ç”ŸæˆAIã€ŒGemini Advancedã€ã‚’ç™ºè¡¨ã—ãŸã€‚ã™ã§ã«ã‚µãƒ¼ãƒ“ã‚¹ã‚’æä¾›é–‹å§‹ã—ã¦ãŠã‚Šã€æœˆé¡2900å††ã§åˆ©ç”¨å¯èƒ½ã€‚2ã‚«æœˆé–“ã®ç„¡æ–™è©¦ç”¨æœŸé–“ã‚‚ç”¨æ„ã™ã‚‹ã€‚ Googleã€ã€ŒGemini 1.5 Proã€é™å®šãƒªãƒªãƒ¼ã‚¹ã€€ã‚³ãƒ³ãƒ†ã‚ã‚¹ãƒˆã‚¦ã‚£ãƒ³ãƒ‰ã‚¦ã¯100ä¸‡ãƒˆãƒ¼ã‚¯ãƒ³ Googleã¯ã€ç”ŸæˆAIã®æ¬¡ä¸–ä»£ãƒ¢ãƒ‡ãƒ«ã€ŒGemini
ktakeda47 2024/02/22
google

ai

llm

gemmini

gemma

gcp
ãƒªãƒ³ã‚¯
LlamaIndexã‚’ä½¿ã£ã¦ãƒãƒ¼ã‚«ãƒ«ç’°å¢ƒã§RAGã‚’å®Ÿè¡Œã™ã‚‹æ–¹æ³• - é›»é€šç·ç ” ãƒ†ãƒƒã‚¯ãƒ–ãƒã‚°
ã“ã‚“ã«ã¡ã¯ã€‚é›»é€šç·ç ”ã‚³ãƒ¼ãƒãƒ¬ãƒ¼ãƒˆæœ¬éƒ¨ã‚·ã‚¹ãƒ†ãƒ æŽ¨é€²éƒ¨ã®å±±ä¸‹ã§ã™ã€‚ æœ€è¿‘ã¯ChatGPTãªã©ã®Large Language Model(LLM)ã‚’åˆ©ç”¨ã—ãŸAIãŒè©±é¡Œã§ã™ãã€‚ ãã®LLMã‚’å¿œç”¨ã—ãŸRetrieval-Augmented Generation(RAG)ã¨ã„ã†æŠ€è¡“ãŒã‚ã‚Šã¾ã™ã€‚ ã“ã‚Œã¯ã€LLMã«æ–‡æ›¸æ¤œç´¢ãªã©ã‚’ç”¨ã„ã¦é–¢é€£ã™ã‚‹æƒ…å ±ã‚’ä¸Žãˆã¦å›žç”ã•ã›ã‚‹ã“ã¨ã§ã€ LLMãŒçŸ¥è˜ã¨ã—ã¦æŒã£ã¦ã„ãªã„å†…å®¹ã‚’å›žç”ã•ã›ãŸã‚Šèª¤ã£ãŸæƒ…å ±ã‚’ç”ãˆã¦ã—ã¾ã†ãƒãƒ«ã‚·ãƒãƒ¼ã‚·ãƒ§ãƒ³ã‚’æŠ‘æ¢ã™ã‚‹æŠ€è¡“ã§ã™ã€‚ ä»Šå›žã¯ã“ã®RAGã‚’LlamaIndexã¨ã„ã†ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã‚’ä½¿ã£ã¦ãƒãƒ¼ã‚«ãƒ«ç’°å¢ƒã§å®Ÿè£…ã™ã‚‹æ–¹æ³•ã«ã¤ã„ã¦ç´¹ä»‹ã—ã¾ã™ã€‚ ãªãœãƒãƒ¼ã‚«ãƒ«ç’°å¢ƒã§LLMã‚’åˆ©ç”¨ã—ãŸã„ã®ã‹ å¤§å¤‰ä¾¿åˆ©ãªãƒ„ãƒ¼ãƒ«ã®LLMã§ã™ãŒã€åˆ©ç”¨ãŒé›£ã—ã„ã“ã¨ã‚‚ã‚ã‚Šã¾ã™ã€‚ ä¾‹ãˆã°ã€æ©Ÿå¯†æƒ…å ±ã‚’å–æ‰±ã„ãŸã„ã€å¤–éƒ¨ã‚¤ãƒ³ã‚¿ãƒ¼ãƒãƒƒãƒˆã¸ã®æŽ¥ç¶šã«åˆ¶é™ãŒæŽ›ã‹ã£ã¦ã„ã‚‹ã¨ã„ã£ãŸå ´åˆã§ã™ã€‚ æœ€çµ‚çš„ã«OpenAI
ktakeda47 2024/01/23
technology

programming

gpt

llm

chatgpt

ai
ãƒªãƒ³ã‚¯
ç”»åƒã‚’è¦‹ã¦è³ªå•ã«ç”ãˆã‚‰ã‚Œã‚‹ã‚ªãƒ¼ãƒ—ãƒ³ã‚½ãƒ¼ã‚¹ãªGPT-4ãƒ¬ãƒ™ãƒ«ã®AIã€ŒLLaVA-1.5ã€ã‚’GCPä¸Šã§å‹•ä½œã•ã›ã¦ã¿ãŸ
Microsoftã‚„ã‚¦ã‚£ã‚¹ã‚³ãƒ³ã‚·ãƒ³å¤§å¦ãƒžãƒ‡ã‚£ã‚½ãƒ³æ ¡ãªã©ã®ç ”ç©¶ãƒãƒ¼ãƒ ãŒé–‹ç™ºã—ã€2023å¹´4æœˆ17æ—¥ã«å…¬é–‹ã—ãŸã€ŒLLaVAã€ã¯ã€Œè¦–è¦šã€ã‚’æŒã¤AIã§ã€ç”»åƒã‚’å…¥åŠ›ã™ã‚‹ã¨ãã®ç”»åƒã«åŸºã¥ã„ã¦è¿”ç”ã‚’è¡Œã†ã“ã¨ãŒã§ãã¾ã™ã€‚2023å¹´10æœˆ5æ—¥ã«ç™»å ´ã—ãŸLLaVA-1.5ã¯ã•ã‚‰ã«ã‚¯ã‚ªãƒªãƒ†ã‚£ãŒå‘ä¸Šã—ã¦ã„ã‚‹ã¨ã®ã“ã¨ãªã®ã§ã€å®Ÿéš›ã«Googleã®ã‚¯ãƒ©ã‚¦ãƒ‰ã‚³ãƒ³ãƒ”ãƒ¥ãƒ¼ãƒ†ã‚£ãƒ³ã‚°ã‚µãƒ¼ãƒ“ã‚¹ã€ŒGoogle Cloud Platform(GCP)ã€ä¸Šã§å‹•ä½œã•ã›ã¦ã¿ã¾ã—ãŸã€‚ LLaVA/pyproject.toml at main Â· haotian-liu/LLaVA https://github.com/haotian-liu/LLaVA 2023å¹´4æœˆã«ãƒªãƒªãƒ¼ã‚¹ã•ã‚ŒãŸæ—§ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã®æ€§èƒ½ã‚„ã€ãƒ‡ãƒ¢ã‚µã‚¤ãƒˆã®ä½¿ã„æ–¹ã«ã¤ã„ã¦ã¯ä¸‹è¨˜ã®è¨˜äº‹ã§ç¢ºèªã§ãã¾ã™ã€‚ ç”»åƒã‚’èªè˜ã—ã¦å¹´é½¢æŽ¨æ¸¬å¯èƒ½ï¼†äººåã‚¯ã‚¤ã‚ºã«ã‚‚æ£ç”ã§ãã‚‹ç„¡æ–™ã®é«˜æ€§èƒ½ãƒãƒ£ãƒƒãƒˆAIã€ŒLLa
ktakeda47 2024/01/04
llm

ai

gpt

gcp

technology
ãƒªãƒ³ã‚¯
å•†ç”¨åˆ©ç”¨å¯èƒ½ã§130å„„ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ãƒ¼ã®æ—¥æœ¬èªžLLMã€ŒELYZA-japanese-Llama-2-13bã€ãŒå…¬é–‹ï¼ãƒ‡ãƒ¢ã‚‚å…¬é–‹ã•ã‚Œã¦ãŠã‚Šã€å®Ÿéš›ã«èƒ½åŠ›ã‚’è©¦ã™ã“ã¨ãŒå¯èƒ½
ktakeda47 2023/12/29
llm

gpt

ai
ãƒªãƒ³ã‚¯
ã€ä»¤å’Œæœ€æ–°ç‰ˆã€‘ä½•ã‚‚ã‚ã‹ã‚‰ãªã„äººå‘ã‘ã®ãƒãƒ¼ã‚«ãƒ« LLM å…¥é–€
ã“ã‚“ã«ã¡ã¯ã€Saldraã§ã™ã€‚æ™®æ®µã¯Pictoriaã¨ã„ã†ä¼šç¤¾ã§AIã®ç¾Žå°‘å¥³ã®éŒ¬æˆã«åŠ±ã‚“ã§ã„ã‚‹ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ã§ã™ã€‚ã“ã®è¨˜äº‹ã¯ãƒãƒ¼ã‚«ãƒ«LLMã®æ¦‚è¦ã‚’ã¤ã‹ã‚€ã“ã¨ã‚’ç›®çš„ã¨ã—ã¾ã™ã€‚å¯¾è±¡èªè€…ã¯ä»¥ä¸‹ã§ã™ã€‚ ãªã‚“ã¨ãªã ChatGPT ã¯ä½¿ã£ãŸã“ã¨ã‚ã‚‹äºº ãƒãƒ¼ã‚«ãƒ« LLM ã‚’èžã„ãŸã“ã¨ã‚ã‚‹ã‘ã©ã‚„ã£ãŸã“ã¨ãªã„äºº ãƒãƒ¼ã‚«ãƒ« LLM ã¨ã¯ OpenAIãŒAPIã‚’å…¬é–‹ã—ã¦ã‹ã‚‰ã€å¤§è¦æ¨¡è¨€èªžãƒ¢ãƒ‡ãƒ«ï¼ˆä»¥é™LLMã¨ã—ã¾ã™ï¼‰ã¯å¤§ããé€²åŒ–ã—ã¦ã„ãã¾ã—ãŸã€‚ã“ã®é€²åŒ–ã¯OpenAIã®APIã ã‘ã§ãªãã€ãƒãƒ¼ã‚«ãƒ«LLMã‚‚é€²åŒ–ã‚’ã—ã¦ã„ã¾ã™ã€‚ ãƒãƒ¼ã‚«ãƒ«LLMã¨ã¯ã€Œä¸€èˆ¬å‘ã‘ã«ãƒ•ã‚¡ã‚¤ãƒ«ã¨ã—ã¦å…¬é–‹ã•ã‚ŒãŸãƒ¢ãƒ‡ãƒ«ã€ã§æŽ¨è«–ã•ã›ã‚‹éŠã³ã§ã™ã€‚APIã¯ä¾¿åˆ©ã§ã™ãŒã€ã‚¤ãƒ³ã‚¿ãƒ¼ãƒãƒƒãƒˆã®æŽ¥ç¶šãŒå¿…è¦ã§ã‚ã£ãŸã‚Šã€APIæä¾›å´ã«ä¾å˜ã™ã‚‹å•é¡ŒãŒã‚ã‚Šã¾ã™ã€‚ãƒãƒ¼ã‚«ãƒ«LLMã¯è‡ªå‰ã§é‹ç”¨ãŒã§ãã‚‹ãŸã‚ã€APIã«ã¯ãªã„ãƒ¡ãƒªãƒƒãƒˆã‚„é…åŠ›ãŒã‚ã‚Šã¾ã™ã€‚ä¸€æ–¹ã§ã€ç’°å¢ƒæ§‹ç¯‰ã‚„ãƒžã‚·ãƒ³ã‚¹ãƒšãƒƒã‚¯ç‰ã€å°‘ã—å§‹
ktakeda47 2023/12/25
llm

openai

chatgpt

ai

gpt
ãƒªãƒ³ã‚¯
LINEã®3.6Bè¨€èªžãƒ¢ãƒ‡ãƒ«ã§éŠã‚“ã ã‚‰çˆ†ç¬‘ã—ãŸï½œshi3z
LINEã‹ã‚‰36å„„(3.6B)ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã®å¤§è¦æ¨¡è¨€èªžãƒ¢ãƒ‡ãƒ«(LLM)ãŒå…¬é–‹ã•ã‚ŒãŸã®ã§æ—©é€ŸéŠã‚“ã§ã¿ãŸã€‚æ£ç¢ºã«ã¯éŠã‚“ã ã®ã¯æ˜¨æ—¥ã®ãƒ‡ã‚¤ãƒªãƒ¼AIãƒ‹ãƒ¥ãƒ¼ã‚¹ãªã®ã ãŒã€é¢ç™½ã‹ã£ãŸã®ã§ã“ã¡ã‚‰ã«ã‚‚è»¢è¼‰ã™ã‚‹ã€‚ ç´°ã‹ã„ã‚„ã‚Šæ–¹ã¯å¤§å…ˆç”Ÿã®ãƒšãƒ¼ã‚¸ã‚’å‚ç…§ã®ã“ã¨ã€‚ ä¾‹ã«ã‚ˆã£ã¦ã“ã‚“ãªé–¢æ•°ã‚’æ›¸ã„ãŸ def line(prompt): # æŽ¨è«–ã®å®Ÿè¡Œ input_ids = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt") tokens = model.generate( input_ids.to(device=model.device), min_length=50, max_length=300, temperature=1.0, do_sample=True, pad_token_id=tokenizer.pad_token_i
ktakeda47 2023/08/16
ãªã‚‹ã»ã©ã€Œè¨€èªžã€ãƒ¢ãƒ‡ãƒ«

web

society

lol

technology

llm

line
ãƒªãƒ³ã‚¯
Prompt Engineering Guide â€“ Nextra
Prompt Engineering Guide Prompt engineering is a relatively new discipline for developing and optimizing prompts to efficiently use language models (LMs) for a wide variety of applications and research topics. Prompt engineering skills help to better understand the capabilities and limitations of large language models (LLMs). Researchers use prompt engineering to improve the capacity of LLMs on a
ktakeda47 2023/03/22
chatgpt

llm

prompt

nlp

ai

programming

gpt
ãƒªãƒ³ã‚¯
1