๐Ÿ—ฃ๏ธ Large Language Model (LLM)

๐Ÿ—ฃ๏ธ Large Language Model (LLM)

[LLM] LORA: LOW-RANK ADAPTATION OF LARGE LANGUAGE MODELS ๋ฆฌ๋ทฐ

์ด๋ฒˆ์—๋Š” "LORA: LOW-RANK ADAPTATION OF LARGE LANGUAGE MODELS" ๋…ผ๋ฌธ์„ ํ•œ๋ฒˆ ๋ฆฌ๋ทฐํ•ด ๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.๋…ผ๋ฌธ ๋งํฌ LoRA: Low-Rank Adaptation of Large Language ModelsAn important paradigm of natural language processing consists of large-scale pre-training on general domain data and adaptation to particular tasks or domains. As we pre-train larger models, full fine-tuning, which retrains all model parameters, becomes learxiv.orgAb..

๐Ÿ—ฃ๏ธ Large Language Model (LLM)

[LLM] Parameter-Efficient Transfer Learning for NLP ๋ฆฌ๋ทฐ

์ด๋ฒˆ์—๋Š” "Parameter-Efficient Transfer Learning for NLP" ๋…ผ๋ฌธ์„ ํ•œ๋ฒˆ ๋ฆฌ๋ทฐํ•ด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.๋…ผ๋ฌธ ๋งํฌ Parameter-Efficient Transfer Learning for NLPFine-tuning large pre-trained models is an effective transfer mechanism in NLP. However, in the presence of many downstream tasks, fine-tuning is parameter inefficient: an entire new model is required for every task. As an alternative, we propose transferarxiv.orgAbstract๋Œ€๊ทœ๋ชจ ์‚ฌ์ „..

๐Ÿ—ฃ๏ธ Large Language Model (LLM)

[LLM] Training language models to follow instructions with human feedback (Instruct GPT / RLHF) Review

์Šคํ„ฐ๋”” ์ค€๋น„๋ฅผ ํ•˜๋‹ค๊ฐ€ ์ •๋ฆฌํ•œ ๋‚ด์šฉ์„ ํ•œ๋ฒˆ Review๋ฅผ ํ•ด๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.Training language models to follow instructions with human feedbackAbstract์–ธ์–ด ๋ชจ๋ธ์˜ ํฌ๊ธฐ๋ฅผ ํ‚ค์šฐ๋Š” ๊ฒƒ๋งŒ์œผ๋กœ๋Š” ์‚ฌ์šฉ์ž ์˜๋„์— ๋” ์ž˜ ๋ถ€ํ•ฉํ•˜๋„๋ก ๋งŒ๋“œ๋Š” ๊ฒƒ์ด ์•„๋‹™๋‹ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ์€ ์ง„์‹คํ•˜์ง€ ์•Š๊ฑฐ๋‚˜, ์œ ํ•ดํ•˜๊ฑฐ๋‚˜, ์‚ฌ์šฉ์ž์—๊ฒŒ ๋„์›€์ด ๋˜์ง€ ์•Š๋Š” ์ถœ๋ ฅ์„ ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ฆ‰, ์ด๋Ÿฌํ•œ ๋ชจ๋ธ์€ ์‚ฌ์šฉ์ž์˜ ์˜๋„์— ๋งž์ถฐ์ ธ ์žˆ์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ์—ฌ๊ธฐ์„œ๋Š” ์ธ๊ฐ„์˜ ํ”ผ๋“œ๋ฐฑ์œผ๋กœ ๋ฏธ์„ธ ์กฐ์ •ํ•˜์—ฌ ๋‹ค์–‘ํ•œ ์ž‘์—…์—์„œ ์‚ฌ์šฉ์ž ์˜๋„์™€ ์ผ์น˜ํ•˜๋Š” ์–ธ์–ด ๋ชจ๋ธ์„ ๊ฐœ๋ฐœํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ์ œ์‹œํ•ฉ๋‹ˆ๋‹ค.OpenAI API๋ฅผ ํ†ตํ•ด ์ˆ˜์ง‘๋œ ๋ ˆ์ด๋ธ”๋Ÿฌ๊ฐ€ ์ž‘์„ฑํ•œ ํ”„๋กฌํ”„ํŠธ์™€ ๋ฐ๋ชจ ๋ฐ์ดํ„ฐ๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ ๋ชจ๋ธ์ด ์›ํ•˜๋Š” ๋™์ž‘์„ ์‹œ์—ฐํ•˜๋Š” ๋ฐ..

๐Ÿ—ฃ๏ธ Large Language Model (LLM)

[LLM] Improving Language Understanding by Generative Pre-Training (GPT-1 ๋…ผ๋ฌธ Review)

์ด๋ฒˆ์—” GPT-1 Model์— ๋ฐํ•˜์—ฌ ๋…ผ๋ฌธ์„ ์ฝ๊ณ  ๊ณต๋ถ€ํ•œ ๋‚ด์šฉ์„ ๋ฆฌ๋ทฐํ•ด ๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.์ด ๋…ผ๋ฌธ์—์„œ๋Š” Transformer ๋ชจ๋ธ์— ๋ฐํ•œ ์„ค๋ช…์ด ํฌํ•จ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ๋ž˜์„œ, ์–ด๋Š์ •๋„ Transformer Model์— ๋ฐํ•œ ์ง€์‹์ด ์ž‡์–ด์•ผ ์ดํ•ดํ•˜์‹ค์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ํ•œ๋ฒˆ ์ฝ๊ณ  ์™€์ฃผ์„ธ์š”! [NLP] Transformer Model - ํŠธ๋žœ์Šคํฌ๋จธ ๋ชจ๋ธ ์•Œ์•„๋ณด๊ธฐ์ด๋ฒˆ ๊ธ€์—์„œ๋Š” Transformer ๋ชจ๋ธ์˜ ์ „๋ฐ˜์ ์ธ Architecture ๋ฐ ๊ตฌ์„ฑ์— ๋ฐํ•˜์—ฌ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค. Transformer: Attention is All You Need Transformer ๋ชจ๋ธ์€ 2017๋…„์— "Attention is All You Need"๋ผ๋Š” ๋…ผ๋ฌธ์„ ํ†ตํ•ด์„œ ์†Œ๊ฐœ๋˜์—ˆ์Šตdaehyun-bigbread.tistory.comAbstract์ž์—ฐ์–ด ์ด..

๐Ÿ—ฃ๏ธ Large Language Model (LLM)

[LLM] Retrieve Augmented Generation (RAG)

์ด๋ฒˆ ๊ธ€์—์„  Retrieve Augmented Generation (RAG)์— ๋ฐํ•˜์—ฌ ํ•œ๋ฒˆ ์•Œ์•„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.Retrieve Augmented Generation (RAG) RAG(Retrieve-Augmented Generation)์€ ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(LLM)์˜ ํ•œ๊ณ„์ ์„ ๊ทน๋ณตํ•˜๊ณ  ๋” ์ •ํ™•ํ•˜๊ณ  ํ’๋ถ€ํ•œ ์‘๋‹ต์„ ์ œ๊ณตํ•˜๊ธฐ ์œ„ํ•ด ๊ณ ์•ˆ๋œ ์•„ํ‚คํ…์ฒ˜์ž…๋‹ˆ๋‹ค.RAG๋Š” ํŠน์ • ์งˆ์˜์— ๋Œ€ํ•ด ์™ธ๋ถ€ ์ง€์‹ ๋ฒ ์ด์Šค์—์„œ ๊ด€๋ จ ์ •๋ณด๋ฅผ ๊ฒ€์ƒ‰ํ•œ ํ›„, ์ด ์ •๋ณด๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ…์ŠคํŠธ๋ฅผ ์ƒ์„ฑํ•˜๋Š” ๋ฐฉ์‹์œผ๋กœ ์ž‘๋™ํ•ฉ๋‹ˆ๋‹ค.ํŠนํžˆ ์ •๋ณด๊ฐ€ ํ’๋ถ€ํ•œ ์‘๋‹ต์„ ์ƒ์„ฑํ•˜๋Š” ๋ฐ ์œ ๋ฆฌํ•ฉ๋‹ˆ๋‹ค.RAG ArchitectureRAG(Retrieve Augmented Generation)์€ ํฌ๊ฒŒ ์„ธ ๊ฐ€์ง€ ๋‹จ๊ณ„๋กœ ๊ตฌ์„ฑ๋ฉ๋‹ˆ๋‹ค.Retrieve ๋‹จ๊ณ„, Augmented ๋‹จ๊ณ„, ๊ทธ๋ฆฌ๊ณ  Gene..

๐Ÿ—ฃ๏ธ Large Language Model (LLM)

[LLM] Prompt Engineering (ํ”„๋กฌํฌํŠธ ์—”์ง€๋‹ˆ์–ด๋ง)

์•ž์— LLM ์†Œ๊ฐœ ๊ธ€์—์„œ Prompt Engineering์— ๊ด€ํ•œ ์†Œ๊ฐœ๋ฅผ ๊ฐ„๋‹จํžˆ ํ–ˆ์Šต๋‹ˆ๋‹ค. ์ด๋ฒˆ๊ธ€์—์„œ ๋” ์ž์„ธํžˆ ๋‹ค๋ค„๋ณด๊ฒ ์Šต๋‹ˆ๋‹ค.Prompt EngineeringPrompt Engineering์ด๋ž€ ์ƒ์„ฑํ˜• ์ธ๊ณต์ง€๋Šฅ(AI), ํŠนํžˆ LLM์„ ํ™œ์šฉํ•˜์—ฌ ์›ํ•˜๋Š” ๊ฒฐ๊ณผ๋ฅผ ์–ป์„ ์ˆ˜ ์žˆ๋„๋ก ํšจ๊ณผ์ ์ธ ํ”„๋กฌํ”„ํŠธ๋ฅผ ์„ค๊ณ„ํ•˜๊ณ  ๊ตฌ์„ฑํ•˜๋Š” ๊ณผ์ • ์„ ๋งํ•ฉ๋‹ˆ๋‹ค. ์ด๋Š” AI์˜ ๋Šฅ๋ ฅ์„ ์ตœ๋Œ€ํ•œ ํ™œ์šฉํ•˜์—ฌ ๋” ๋‚˜์€ ๊ฒฐ๊ณผ๋ฅผ ์–ป๋Š” ๋ฐฉ๋ฒ•์„ ์—ฐ๊ตฌํ•˜๊ณ  ์ ์šฉํ•˜๋Š” ๊ธฐ์ˆ  ์ž…๋‹ˆ๋‹ค.์˜ˆ๋ฅผ ๋“ค์–ด, ํŠน์ •ํ•œ ์งˆ๋ฌธ์„ ํ–ˆ์„ ๋•Œ, ๋‹ต๋ณ€์˜ ์ •ํ™•์„ฑ๊ณผ ํ’ˆ์งˆ์„ ๋†’์ด๊ธฐ ์œ„ํ•ด ํ”„๋กฌํ”„ํŠธ๋ฅผ ๋” ๊ตฌ์ฒด์ ์ด๊ณ  ๋ช…ํ™•ํ•˜๊ฒŒ ๋งŒ๋“œ๋Š” ๊ฒƒ์ด ๋ฐ”๋กœ Prompt Engineering์ž…๋‹ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ๋ชจ๋ธ์ด ๋” ๋‚˜์€ ์„ฑ๋Šฅ์„ ๋ฐœํœ˜ํ•˜๊ณ , ์‚ฌ์šฉ์ž๋Š” ์›ํ•˜๋Š” ๊ฒฐ๊ณผ๋ฅผ ์ •ํ™•ํ•˜๊ฒŒ ์–ป์„ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.ํšจ์œจ์ ์ธ ํ”„๋กฌํ”„ํŠธ ์ž‘์„ฑ:..

๐Ÿ—ฃ๏ธ Large Language Model (LLM)

[LLM] What is Large Language Model (LLM)?

LLM ํŒŒํŠธ ๋ถ€๋ถ„์— ๋Œ€ํ•œ ๊ธ€์„ ์ฒ˜์Œ์จ๋ณด๋„ค์š”. ์•ž์œผ๋กœ ๋‹น๋ถ„๊ฐ„์€ LLM์— ๊ด€ํ•œ ๊ธ€๋งŒ ์˜ฌ๋ฆด๊ฑฐ ๊ฐ™์•„์š”..๊ณต๋ถ€๋ฅผ ๊ทธ์ชฝ์œผ๋กœ ํ•˜๊ณ  ์žˆ์–ด์„œ. ํ•œ๋ฒˆ ์‹œ์ž‘ํ•ด๋ณผ๊ป˜์š”.Large Language Model (LLM)LLM์œผ๋กœ ๋“ค์–ด๊ฐ€๊ธฐ ์ „์—, ์ผ๋‹จ AI๊ฐ€ ์–ด๋– ํ•œ ๊ฐœ๋…์ธ์ง€ ํ•œ๋ฒˆ ์งš๊ณ  ๋„˜์–ด๊ฐ€์•ผ ํ•  ํ•„์š”๊ฐ€ ์žˆ์Šต๋‹ˆ๋‹ค.AI๋Š” ์ธ๊ฐ„์˜ ํ•™์Šต ๋Šฅ๋ ฅ, ์ถ”๋ก  ๋Šฅ๋ ฅ, ์ง€๊ฐ ๋Šฅ๋ ฅ์„ ์ธ๊ณต์ ์œผ๋กœ ๊ตฌํ˜„ํ•˜๋ ค๋Š”, ์ปดํ“จํ„ฐ ๊ณผํ•™์˜ ์„ธ๋ถ€ ๋ถ„์•ผ์ค‘ ํ•˜๋‚˜์ž…๋‹ˆ๋‹ค.์ฆ‰, AI๋Š” ์ธ๊ฐ„์˜ ์ธ์ง€ ๋Šฅ๋ ฅ์„ ๋ชจ๋ฐฉํ•˜์—ฌ ๋ฐ์ดํ„ฐ๋ฅผ ์ฒ˜๋ฆฌํ•˜๊ณ  ๋ถ„์„ํ•˜๋Š” ๊ธฐ์ˆ ์ž…๋‹ˆ๋‹ค. ํ•œ๋ฒˆ ๋‹ค์ด์–ด๊ทธ๋žจ์„ ๋ณด๊ณ  ์–ด๋– ํ•œ Flow๋กœ ๊ตฌ์„ฑ๋˜๋Š”์ง€ ์„ค๋ช…์„ ๋“œ๋ฆฌ๊ฒ ์Šต๋‹ˆ๋‹ค.INPUT (์ž…๋ ฅ): AI ์‹œ์Šคํ…œ์— ์ฃผ์–ด์ง€๋Š” ๋ฐ์ดํ„ฐ ๋˜๋Š” ์ •๋ณด์ž…๋‹ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, ํ…์ŠคํŠธ, ์ด๋ฏธ์ง€, ์Œ์„ฑ ๋“ฑ์ด ์ž…๋ ฅ์œผ๋กœ ๋“ค์–ด๊ฐˆ ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.ํ•™์Šต, ์ถ”๋ก , ์ง€๊ฐ: ..

Bigbread1129
'๐Ÿ—ฃ๏ธ Large Language Model (LLM)' ์นดํ…Œ๊ณ ๋ฆฌ์˜ ๊ธ€ ๋ชฉ๋ก