{"id":14964,"date":"2024-11-19T20:13:41","date_gmt":"2024-11-19T12:13:41","guid":{"rendered":"https:\/\/www.aisharenet.com\/?p=14964"},"modified":"2024-12-08T20:17:23","modified_gmt":"2024-12-08T12:17:23","slug":"aigongchengxueyuan3fi","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/en\/aigongchengxueyuan3fi\/","title":{"rendered":"AI\u5de5\u7a0b\u5b66\u9662\uff1a3Fine-tuning\uff08\u5927\u8bed\u8a00\u6a21\u578b\u5fae\u8c03\uff09"},"content":{"rendered":"

\ud83d\udcda \u8d44\u6599\u5e93\u7ed3\u6784<\/h2>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
\u6a21\u578b\/\u76ee\u5f55<\/th>\n\u63cf\u8ff0\u4e0e\u5185\u5bb9<\/th>\n<\/tr>\n<\/thead>\n
Axolotl<\/strong><\/a><\/td>\n\u7528\u4e8e\u5fae\u8c03\u8bed\u8a00\u6a21\u578b\u7684\u6846\u67b6<\/td>\n<\/tr>\n
Gemma<\/strong><\/a><\/td>\nGoogle \u6700\u65b0\u7684\u5927\u8bed\u8a00\u6a21\u578b\u5b9e\u73b0<\/td>\n<\/tr>\n
–\u00a0finetune-gemma.ipynb<\/code>\u00a0–\u00a0gemma-sft.py<\/code>\u00a0–\u00a0Gemma_finetuning_notebook.ipynb<\/code><\/td>\n\u5fae\u8c03\u7b14\u8bb0\u672c\u548c\u811a\u672c<\/td>\n<\/tr>\n
LLama2<\/strong><\/a><\/td>\nMeta \u7684\u5f00\u6e90\u5927\u8bed\u8a00\u6a21\u578b<\/td>\n<\/tr>\n
–\u00a0generate_response_stream.py<\/code>\u00a0–\u00a0Llama2_finetuning_notebook.ipynb<\/code>\u00a0–\u00a0Llama_2_Fine_Tuning_using_QLora.ipynb<\/code><\/td>\n\u5b9e\u73b0\u548c\u5fae\u8c03\u6307\u5357<\/td>\n<\/tr>\n
Llama3<\/strong><\/a><\/td>\n\u5373\u5c06\u63a8\u51fa\u7684 Meta \u5927\u8bed\u8a00\u6a21\u578b\u5b9e\u9a8c<\/td>\n<\/tr>\n
–\u00a0Llama3_finetuning_notebook.ipynb<\/code><\/td>\n\u521d\u59cb\u5fae\u8c03\u5b9e\u9a8c<\/td>\n<\/tr>\n
LlamaFactory<\/strong><\/a><\/td>\n\u5927\u8bed\u8a00\u6a21\u578b\u7684\u8bad\u7ec3\u548c\u90e8\u7f72\u6846\u67b6<\/td>\n<\/tr>\n
LLMArchitecture\/ParameterCount<\/strong><\/a><\/td>\n\u6a21\u578b\u67b6\u6784\u7684\u6280\u672f\u7ec6\u8282<\/td>\n<\/tr>\n
Mistral-7b<\/strong><\/a><\/td>\nMistral AI<\/a> \u7684 70 \u4ebf\u53c2\u6570\u6a21\u578b<\/td>\n<\/tr>\n
–\u00a0LLM_evaluation_harness_for_Arc_Easy_and_SST.ipynb<\/code>\u00a0–\u00a0Mistral_Colab_Finetune_ipynb_Colab_Final.ipynb<\/code>\u00a0–\u00a0notebooks_chatml_inference.ipynb<\/code>\u00a0–\u00a0notebooks_DPO_fine_tuning.ipynb<\/code>\u00a0–\u00a0notebooks_SFTTrainer TRL.ipynb<\/code>\u00a0–\u00a0SFT.py<\/code><\/td>\n\u7528\u4e8e\u8bc4\u4f30\u3001\u5fae\u8c03\u548c\u63a8\u7406\u7684\u7efc\u5408\u7b14\u8bb0\u672c<\/td>\n<\/tr>\n
Mixtral<\/strong><\/a><\/td>\nMixtral \u7684\u4e13\u5bb6\u6df7\u5408\u6a21\u578b<\/td>\n<\/tr>\n
–\u00a0Mixtral_fine_tuning.ipynb<\/code><\/td>\n\u5fae\u8c03\u5b9e\u73b0<\/td>\n<\/tr>\n
VLM<\/strong><\/a><\/td>\n\u89c6\u89c9\u8bed\u8a00\u6a21\u578b<\/td>\n<\/tr>\n
–\u00a0Florence2_finetuning_notebook.ipynb<\/code>\u00a0–\u00a0PaliGemma_finetuning_notebook.ipynb<\/code><\/td>\n\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\u7684\u5b9e\u73b0<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

\ud83c\udfaf \u6a21\u5757\u6982\u89c8<\/h2>\n

1. LLM \u67b6\u6784<\/h3>\n
    \n
  • \u63a2\u7d22\u4ee5\u4e0b\u6a21\u578b\u7684\u5b9e\u73b0\uff1a\n
      \n
    • Llama2\uff08Meta \u7684\u5f00\u6e90\u6a21\u578b\uff09<\/li>\n
    • Mistral-7b\uff08\u9ad8\u6548\u7684 70 \u4ebf\u53c2\u6570\u6a21\u578b\uff09<\/li>\n
    • Mixtral\uff08\u4e13\u5bb6\u6df7\u5408\u67b6\u6784\uff09<\/li>\n
    • Gemma\uff08Google \u7684\u6700\u65b0\u8d21\u732e\uff09<\/li>\n
    • Llama3\uff08\u5373\u5c06\u63a8\u51fa\u7684\u5b9e\u9a8c\uff09<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n

      2. \ud83d\udee0\ufe0f \u5fae\u8c03\u6280\u672f<\/h3>\n
        \n
      • \u5b9e\u73b0\u7b56\u7565<\/li>\n
      • LoRA\uff08\u4f4e\u79e9\u9002\u914d\uff09\u65b9\u6cd5<\/li>\n
      • \u9ad8\u7ea7\u4f18\u5316\u65b9\u6cd5<\/li>\n<\/ul>\n

        3. \ud83c\udfd7\ufe0f \u6a21\u578b\u67b6\u6784\u5206\u6790<\/h3>\n
          \n
        • \u6a21\u578b\u7ed3\u6784\u7684\u6df1\u5165\u7814\u7a76<\/li>\n
        • \u53c2\u6570\u8ba1\u7b97\u65b9\u6cd5<\/li>\n
        • \u6269\u5c55\u6027\u8003\u8651<\/li>\n<\/ul>\n

          4. \ud83d\udd27 \u4e13\u4e1a\u5b9e\u73b0<\/h3>\n
            \n
          • Code Llama \u7528\u4e8e\u7f16\u7a0b\u4efb\u52a1<\/li>\n
          • \u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff1a\n
              \n
            • Florence2<\/li>\n
            • PaliGemma<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n

              5. \ud83d\udcbb \u5b9e\u8df5\u5e94\u7528<\/h3>\n
                \n
              • \u7efc\u5408 Jupyter \u7b14\u8bb0\u672c<\/li>\n
              • \u54cd\u5e94\u751f\u6210\u6d41\u6c34\u7ebf<\/li>\n
              • \u63a8\u7406\u5b9e\u73b0\u6307\u5357<\/li>\n<\/ul>\n

                6. \ud83d\ude80 \u9ad8\u7ea7\u4e3b\u9898<\/h3>\n
                  \n
                • DPO\uff08\u76f4\u63a5\u504f\u597d\u4f18\u5316\uff09<\/li>\n
                • SFT\uff08\u76d1\u7763\u5fae\u8c03\uff09<\/li>\n
                • \u8bc4\u4f30\u65b9\u6cd5\u8bba<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"

                  \ud83d\udcda \u8d44\u6599\u5e93\u7ed3\u6784 \u6a21\u578b\/\u76ee\u5f55 \u63cf\u8ff0\u4e0e\u5185\u5bb9 Axolotl \u7528\u4e8e\u5fae\u8c03\u8bed\u8a00\u6a21\u578b\u7684\u6846\u67b6 Gemma Google \u6700\u65b0\u7684\u5927\u8bed\u8a00\u6a21\u578b\u5b9e\u73b0 –\u00a0finetune-gemma.ipynb\u00a0–\u00a0gemma-sft.py\u00a0&#8211…<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[34],"tags":[],"class_list":["post-14964","post","type-post","status-publish","format-standard","hentry","category-knowledge"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/posts\/14964","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/comments?post=14964"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/posts\/14964\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/media?parent=14964"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/categories?post=14964"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/en\/wp-json\/wp\/v2\/tags?post=14964"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}