{"id":53779,"date":"2025-02-16T09:51:54","date_gmt":"2025-02-16T01:51:54","guid":{"rendered":"https:\/\/fwq.ai\/blog\/53779\/"},"modified":"2025-02-16T09:51:54","modified_gmt":"2025-02-16T01:51:54","slug":"deepseek-r1-671b%e6%9c%ac%e5%9c%b0%e8%bf%90%e8%a1%8c%e6%8c%87%e5%8d%97","status":"publish","type":"post","link":"https:\/\/fwq.ai\/blog\/53779\/","title":{"rendered":"DeepSeek-R1 671B\u672c\u5730\u8fd0\u884c\u6307\u5357"},"content":{"rendered":"<p>\u539f\u59cb\u7684 DeepSeek R1 \u662f\u4e00\u4e2a 6710 \u4ebf\u53c2\u6570\u7684\u8bed\u8a00\u6a21\u578b\uff0c\u7531 Unsloth AI \u56e2\u961f\u8fdb\u884c\u4e86\u52a8\u6001\u91cf\u5316\uff0c\u5927\u5c0f\u51cf\u5c11\u4e86 80%\uff08\u4ece 720 GB \u51cf\u5c11\u5230 131 GB\uff09\uff0c\u540c\u65f6\u4fdd\u6301\u4e86\u5f3a\u5927\u7684\u6027\u80fd\u3002\u5f53\u6dfb\u52a0\u6a21\u578b\u5378\u8f7d\u65f6\uff0c\u8be5\u6a21\u578b\u53ef\u4ee5\u5728 24GB VRAM \u4e0b\u4ee5\u4f4e\u4ee4\u724c\/\u79d2\u7684\u63a8\u7406\u901f\u5ea6\u8fd0\u884c\u3002<\/p>\n<blockquote><p>\n  \u4e3a\u4ec0\u4e48\u5927\u5c0f\u5bf9\u5927\u578b\u8bed\u8a00\u6a21\u578b\u5f88\u91cd\u8981\n<\/p><\/blockquote>\n<p>\u5927\u578b\u8bed\u8a00\u6a21\u578b\u672c\u8d28\u4e0a\u9700\u8981\u5927\u91cf\u5b58\u50a8\u548c\u8ba1\u7b97\u8d44\u6e90\u3002\u4e3a\u6240\u6709\u53c2\u6570\u4fdd\u6301\u5168\u7cbe\u5ea6\u8868\u793a\uff08\u901a\u5e38\u4e3a FP16 \u6216 FP32\uff09\u53d8\u5f97\u4e0d\u5207\u5b9e\u9645\uff0c\u5c24\u5176\u662f\u5bf9\u4e8e\u5c40\u90e8\u63a8\u7406\uff0c\u56e0\u4e3a\u5b83\u9700\u8981\u5185\u5b58\u3002\u91cf\u5316\uff08\u51cf\u5c11\u6743\u91cd\u8868\u793a\u7684\u4f4d\u5bbd\uff09\u901a\u8fc7\u5927\u5e45\u51cf\u5c11\u6a21\u578b\u5927\u5c0f\u548c\u5185\u5b58\u5360\u7528\u63d0\u4f9b\u4e86\u4e00\u79cd\u89e3\u51b3\u65b9\u6848\u3002\u7136\u800c\uff0c\u6574\u4e2a\u7f51\u7edc\u7684\u7b80\u5355\u3001\u7edf\u4e00\u7684\u91cf\u5316\u53ef\u80fd\u4f1a\u5bfc\u81f4\u4e25\u91cd\u7684\u6027\u80fd\u4e0b\u964d\uff0c\u8868\u73b0\u4e3a\u4e0d\u7a33\u5b9a\u7684\u8f93\u51fa\u6216\u91cd\u590d\u7684 token \u751f\u6210\u3002<\/p>\n<blockquote><p>\n  \u52a8\u6001\u91cf\u5316\uff1a\u4e00\u79cd\u91cf\u8eab\u5b9a\u5236\u7684\u65b9\u6cd5\n<\/p><\/blockquote>\n<p>Unsloth AI \u56e2\u961f\u7684\u65b9\u6cd5\u6d89\u53ca\u52a8\u6001\u91cf\u5316\uff0c\u5176\u4e2d\u6839\u636e\u4e0d\u540c\u7f51\u7edc\u7ec4\u4ef6\u7684\u654f\u611f\u5ea6\u5206\u914d\u53ef\u53d8\u4f4d\u7cbe\u5ea6\u3002\u5173\u952e\u6280\u672f\u89c1\u89e3\u5305\u62ec\uff1a<\/p>\n<ul>\n<li>\u9009\u62e9\u6027\u7cbe\u5ea6\u5206\u914d\uff1a\u521d\u59cb\u5bc6\u96c6\u5c42\u548c\u5411\u4e0b\u6295\u5f71\uff08down_proj\uff09\u77e9\u9635\u5bf9\u4e8e\u8bbe\u7f6e\u7a33\u5b9a\u7684\u8868\u793a\u548c\u7ba1\u7406 SwiGLU \u6fc0\u6d3b\u4e2d\u7684\u7f29\u653e\u5c5e\u6027\u81f3\u5173\u91cd\u8981\uff0c\u5b83\u4eec\u4fdd\u6301\u5728\u66f4\u9ad8\u7684\u7cbe\u5ea6\uff084 \u4f4d\u6216 6 \u4f4d\uff09\u3002\u76f8\u53cd\uff0c\u5927\u90e8\u5206\u53c2\u6570\uff08\u4e3b\u8981\u662f\u6df7\u5408\u4e13\u5bb6 (MoE) \u5c42\u4e2d\u7684\u53c2\u6570\uff0c\u7ea6\u5360\u6a21\u578b\u7684 88%\uff09\u88ab\u79ef\u6781\u91cf\u5316\u4e3a 1.5-2 \u4f4d\u3002<\/li>\n<li>\u91cd\u8981\u6027\u77e9\u9635\u6821\u51c6\uff1a\u5728\u91cf\u5316\u8fc7\u7a0b\u4e2d\u52a0\u5165\u91cd\u8981\u6027\u77e9\u9635\u5141\u8bb8\u8be5\u65b9\u6cd5\u52a8\u6001\u8c03\u6574\u6bcf\u5c42\u7684\u7cbe\u5ea6\u6c34\u5e73\u3002\u6b64\u6821\u51c6\u53ef\u9632\u6b62\u5e38\u89c1\u7684\u9677\u9631\uff0c\u4f8b\u5982\u901a\u5e38\u7531\u5747\u5300\u91cf\u5316\u5f15\u8d77\u7684\u65e0\u9650\u5faa\u73af\u6216\u65e0\u610f\u4e49\u7684\u8f93\u51fa\u3002<\/li>\n<li>\u5c42\u7279\u5b9a\u7684\u654f\u611f\u6027\u5206\u6790\uff1a\u6280\u672f\u8bc4\u4f30\u8868\u660e\uff0c\u867d\u7136 MoE \u5c42\u53ef\u4ee5\u5bb9\u5fcd\u8f83\u4f4e\u7684\u7cbe\u5ea6\uff0c\u4f46\u6ce8\u610f\u673a\u5236\u3001\u5d4c\u5165\u5c42\u548c\u6700\u7ec8\u8f93\u51fa\u5934\u7b49\u7ec4\u4ef6\u9700\u8981\u66f4\u591a\u4f4d\u6765\u4fdd\u7559\u6fc0\u6d3b\u5206\u5e03\u3002\u8fd9\u79cd\u7ec6\u81f4\u5165\u5fae\u7684\u7b56\u7565\u53ef\u786e\u4fdd\u8ba1\u7b97\u56fe\u4e2d\u7684\u5173\u952e\u8def\u5f84\u4fdd\u6301\u8db3\u591f\u7684\u4fdd\u771f\u5ea6\u3002<\/li>\n<\/ul>\n<h2>1\u3001\u91cf\u5316\u6a21\u578b\u53d8\u4f53\u548c\u6027\u80fd<\/h2>\n<p>Unsloth AI \u53d1\u5e03\u4e86\u591a\u4e2a\u52a8\u6001\u91cf\u5316\u53d8\u4f53\uff0c\u6bcf\u4e2a\u53d8\u4f53\u90fd\u5e73\u8861\u4e86\u6a21\u578b\u5927\u5c0f\u548c\u8f93\u51fa\u8d28\u91cf\uff1a<\/p>\n<p>\u4f8b\u5982\uff0c\u5728\u53d7\u63a7\u6d4b\u8bd5\u4e2d\uff0c\u6a21\u578b\u7684\u4efb\u52a1\u662f\u751f\u6210 Flappy Bird \u6e38\u620f\u7684 Python \u5b9e\u73b0\uff0c\u5373\u4f7f\u662f\u6700\u5c0f\u7684 1.58 \u4f4d\u53d8\u4f53\u4e5f\u80fd\u4fdd\u6301\u5b9e\u8d28\u6027\u7684\u529f\u80fd\u3002\u76f8\u6bd4\u4e4b\u4e0b\uff0c\u6240\u6709\u5c42\u7684\u7edf\u4e00\u91cf\u5316\u4f1a\u5bfc\u81f4\u91cd\u590d\u8f93\u51fa\u6216\u5b8c\u5168\u65e0\u6cd5\u751f\u6210\u8fde\u8d2f\u7684\u4ee3\u7801\u3002<\/p>\n<h2>2\u3001\u672c\u5730\u90e8\u7f72 DeepSeek R1<\/h2>\n<p>\u52a8\u6001\u91cf\u5316\u6a21\u578b\u65e8\u5728\u5728\u5e38\u89c1\u7684\u63a8\u7406\u5f15\u64ce\uff08\u5982 llama.cpp\uff09\u4e0a\u8fd0\u884c\uff0c\u5b83\u652f\u6301 Unsloth AI \u7248\u672c\u4f7f\u7528\u7684 GGUF \u6587\u4ef6\u683c\u5f0f\u3002\u4ee5\u4e0b\u662f\u90e8\u7f72\u8fc7\u7a0b\u7684\u6982\u8ff0\uff1a<\/p>\n<h3>2.1 \u6784\u5efa\u63a8\u7406\u5f15\u64ce<\/h3>\n<p>\u5728\u542f\u7528 GPU \u652f\u6301\u7684\u60c5\u51b5\u4e0b\u514b\u9686\u548c\u7f16\u8bd1 llama.cpp\uff1a<\/p>\n<pre><code>git clone https:\/\/github.com\/ggerganov\/llama.cpp\ncd llama.cpp\ncmake . -B build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON -DLLAMA_CURL=ON\ncmake --build build --config Release -j --clean-first --target llama-quantize llama-cli llama-gguf-split<\/code><\/pre>\n<h3>2.2 \u4e0b\u8f7d\u6a21\u578b<\/h3>\n<p>\u4f7f\u7528 Hugging Face Hub \u68c0\u7d22\u6240\u9700\u7684\u6a21\u578b\u53d8\u4f53\uff1a<\/p>\n<pre><code>from huggingface_hub import snapshot_download\n\nsnapshot_download(\n    repo_id=\"unsloth\/DeepSeek-R1-GGUF\",\n    local_dir=\"DeepSeek-R1-GGUF\",\n    allow_patterns=[\"*UD-IQ1_S*\"],  # For the 1.58-bit version\n)<\/code><\/pre>\n<h3>2.3 GPU \u5378\u8f7d\u6ce8\u610f\u4e8b\u9879<\/h3>\n<p>\u6839\u636e\u53ef\u7528\u7684 VRAM\uff0c\u4f7f\u7528\u4ee5\u4e0b\u547d\u4ee4\u786e\u5b9a\u8981\u5378\u8f7d\u5230 GPU \u7684\u5c42\u6570\uff1a<\/p>\n<pre><code>n_offload = floor((GPU_VRAM_GB \/ Model_FileSize_GB) * (Total_Layers - 4))<\/code><\/pre>\n<p>\u4f7f\u7528\u7c7b\u4f3c\u4ee5\u4e0b\u547d\u4ee4\u6267\u884c\u6a21\u578b\u8fd0\u884c\u63a8\u7406\uff1a<\/p>\n<pre><code>.\/build\/bin\/llama-cli \\\n    --model DeepSeek-R1-GGUF\/DeepSeek-R1-UD-IQ1_S\/DeepSeek-R1-UD-IQ1_S-00001-of-00003.gguf \\\n    --cache-type-k q4_0 \\\n    --threads 16 \\\n    --prio 2 \\\n    --temp 0.6 \\\n    --ctx-size 8192 \\\n    --seed 3407 \\\n    --n-gpu-layers 7 \\\n    -no-cnv \\\n    --prompt \"&lt;|User|&gt;Create a Flappy Bird game in Python.&lt;|Assistant|&gt;\"<\/code><\/pre>\n<h2>3\u3001\u6211\u7684\u7ecf\u9a8c<\/h2>\n<p>\u6211\u60f3\u4eb2\u81ea\u5c1d\u8bd5\u4e00\u4e0b\u8fd9\u79cd\u52a8\u6001\u91cf\u5316\u3002\u4e3a\u4e86\u6d4b\u8bd5\u8be5\u6a21\u578b\uff0c\u6211\u5728 VastAI \u4e0a\u79df\u7528\u4e86\u4e00\u4e2a 80 GB \u7684 GPU\uff0c\u6bcf\u5c0f\u65f6\u4ec5\u9700 2.7 \u7f8e\u5143\u3002\u8003\u8651\u5230\u539f\u59cb\u6a21\u578b\u7684\u5e9e\u5927\u89c4\u6a21\uff0c\u6211\u5bf9\u91cf\u5316\u7248\u672c\u7684\u6027\u80fd\u548c\u6548\u7387\u975e\u5e38\u6ee1\u610f\u3002\u4ee5\u4e0b\u662f\u6211\u8fdb\u884c\u7684\u4e00\u4e9b\u793a\u4f8b\u6d4b\u8bd5\uff1a<\/p>\n<h3>3.1 Flappy Bird \u6e38\u620f\u751f\u6210<\/h3>\n<p>\u8be5\u6a21\u578b\u6210\u529f\u751f\u6210\u4e86\u7ecf\u5178 Flappy Bird \u6e38\u620f\u7684 Python \u5b9e\u73b0\u3002\u5c3d\u7ba1\u6fc0\u8fdb\u91cf\u5316\u5b58\u5728\u4e00\u4e9b\u5178\u578b\u7684\u5c0f\u95ee\u9898\uff0c\u4f46\u6838\u5fc3\u529f\u80fd\u5b8c\u597d\uff0c\u4ee3\u7801\u53ea\u9700\u8fdb\u884c\u5c11\u91cf\u4fee\u6539\u5373\u53ef\u8fd0\u884c\u3002<\/p>\n<p>  \u6b64\u6f14\u793a\u6765\u81ea unsloth.ai\uff0c\u4f46\u6211\u4f7f\u7528 1.58bit \u7248\u672c\u83b7\u5f97\u4e86\u7c7b\u4f3c\u7684\u7ed3\u679c <\/p>\n<h3>3.2 \u8fd0\u52a8\u68c0\u6d4b<\/h3>\n<h3>4\u3001\u5e38\u89c1\u9677\u9631\u548c\u6280\u672f\u8003\u8651<\/h3>\n<ul>\n<li>\u6807\u8bb0\u5316\u7ec6\u5fae\u5dee\u522b\uff1a<\/li>\n<\/ul>\n<p>\u6ce8\u610f\u7279\u6b8a\u6807\u8bb0\uff08\u4f8b\u5982\uff0c&lt;|User|&gt;\u3001&lt;|Assistant|&gt;\u3001&lt;|begin_of_sentence|&gt;\u3001&lt;|end_of_sentence|&gt;\uff09\u3002\u5904\u7406\u4e0d\u5f53\u53ef\u80fd\u4f1a\u5bfc\u81f4 BOS \u6807\u8bb0\u91cd\u590d\u6216 EOS \u5c4f\u853d\u9519\u8bef\u7b49\u95ee\u9898\u3002<\/p>\n<ul>\n<li>\u53c2\u6570\u654f\u611f\u6027\uff1a<\/li>\n<\/ul>\n<p>\u6709\u65f6\uff0c\u52a8\u6001\u91cf\u5316\u53ef\u80fd\u4f1a\u5728\u957f\u5e8f\u5217\u4e2d\u4ea7\u751f\u5b64\u7acb\u7684\u9519\u8bef\u6807\u8bb0\u3002\u8c03\u6574\u63a8\u7406\u53c2\u6570\uff08\u4f8b\u5982 min_p\uff09\uff08\u4f8b\u5982\uff0c\u8c03\u6574\u4e3a 0.1 \u6216 0.05\uff09\u53ef\u4ee5\u5e2e\u52a9\u7f13\u89e3\u8fd9\u4e9b\u7ec6\u5fae\u5dee\u5f02\u3002<\/p>\n<h2>5\u3001\u7ed3\u675f\u8bed<\/h2>\n<p>Unsloth AI \u5bf9 DeepSeek R1 \u7684\u52a8\u6001\u91cf\u5316\u4f53\u73b0\u4e86\u6a21\u578b\u538b\u7f29\u6280\u672f\u7684\u91cd\u5927\u8fdb\u6b65\u3002\u901a\u8fc7\u5728\u5404\u4e2a\u7f51\u7edc\u5c42\u4e4b\u95f4\u667a\u80fd\u5730\u5206\u914d\u4f4d\u7cbe\u5ea6\uff0c\u6b64\u65b9\u6cd5\u4fdd\u7559\u4e86\u57fa\u672c\u7684\u8ba1\u7b97\u4fdd\u771f\u5ea6\uff0c\u540c\u65f6\u5c06\u6a21\u578b\u7684\u5b58\u50a8\u7a7a\u95f4\u51cf\u5c11\u4e86\u591a\u8fbe 80%\u3002\u8fd9\u610f\u5473\u7740\u6700\u5148\u8fdb\u7684\u5927\u578b\u8bed\u8a00\u6a21\u578b\u73b0\u5728\u66f4\u5bb9\u6613\u8bbf\u95ee\uff0c\u4ece\u800c\u80fd\u591f\u5728\u4ee5\u524d\u4e0d\u8db3\u7684\u786c\u4ef6\u4e0a\u8fdb\u884c\u5b9e\u9a8c\u548c\u90e8\u7f72\u3002<\/p>\n<p>\u5982\u679c\u4f60\u6709\u5174\u8da3\u8fdb\u4e00\u6b65\u63a2\u7d22\u8fd9\u4e00\u70b9\uff0c\u6211\u9f13\u52b1\u4f60\u67e5\u770b\u548c \u3002\u4e5f\u8bf7\u67e5\u770b\u4ed6\u4eec\u7684\u539f\u59cb\u6587\u7ae0\uff1a<\/p>\n<hr>\n","protected":false},"excerpt":{"rendered":"<p>\u539f\u59cb\u7684 DeepSeek R1 \u662f\u4e00\u4e2a 6710 \u4ebf\u53c2\u6570\u7684\u8bed\u8a00\u6a21\u578b\uff0c\u7531 Unsloth AI \u56e2\u961f\u8fdb\u884c\u4e86\u52a8\u6001\u91cf\u5316\uff0c\u5927\u5c0f\u51cf\u5c11\u4e86 80%\uff08\u4ece 720 GB \u51cf\u5c11\u5230 131 GB\uff09\uff0c\u540c\u65f6\u4fdd\u6301\u4e86\u5f3a\u5927\u7684\u6027\u80fd\u3002\u5f53\u6dfb\u52a0\u6a21\u578b\u5378\u8f7d\u65f6\uff0c\u8be5\u6a21\u578b\u53ef\u4ee5\u5728 24GB VRAM \u4e0b\u4ee5\u4f4e\u4ee4\u724c\/\u79d2\u7684\u63a8\u7406\u901f\u5ea6\u8fd0\u884c\u3002 \u4e3a\u4ec0\u4e48\u5927\u5c0f\u5bf9\u5927\u578b\u8bed\u8a00\u6a21\u578b\u5f88\u91cd\u8981 \u5927\u578b\u8bed\u8a00\u6a21\u578b\u672c\u8d28\u4e0a\u9700\u8981\u5927\u91cf\u5b58\u50a8\u548c\u8ba1\u7b97\u8d44\u6e90\u3002\u4e3a\u6240\u6709\u53c2\u6570\u4fdd\u6301\u5168\u7cbe\u5ea6\u8868\u793a\uff08\u901a\u5e38\u4e3a FP16 \u6216 FP32\uff09\u53d8\u5f97\u4e0d\u5207\u5b9e\u9645\uff0c\u5c24\u5176\u662f\u5bf9\u4e8e\u5c40\u90e8\u63a8\u7406\uff0c\u56e0\u4e3a\u5b83\u9700\u8981\u5185\u5b58\u3002\u91cf\u5316\uff08\u51cf\u5c11\u6743\u91cd\u8868\u793a\u7684\u4f4d\u5bbd\uff09\u901a\u8fc7\u5927\u5e45\u51cf\u5c11\u6a21\u578b\u5927\u5c0f\u548c\u5185\u5b58\u5360\u7528\u63d0\u4f9b\u4e86\u4e00\u79cd\u89e3\u51b3\u65b9\u6848\u3002\u7136\u800c\uff0c\u6574\u4e2a\u7f51\u7edc\u7684\u7b80\u5355\u3001\u7edf\u4e00\u7684\u91cf\u5316\u53ef\u80fd\u4f1a\u5bfc\u81f4\u4e25\u91cd\u7684\u6027\u80fd\u4e0b\u964d\uff0c\u8868\u73b0\u4e3a\u4e0d\u7a33\u5b9a\u7684\u8f93\u51fa\u6216\u91cd\u590d\u7684 token \u751f\u6210\u3002 \u52a8\u6001\u91cf\u5316\uff1a\u4e00\u79cd\u91cf\u8eab\u5b9a\u5236\u7684\u65b9\u6cd5 Unsloth AI \u56e2\u961f\u7684\u65b9\u6cd5\u6d89\u53ca\u52a8\u6001\u91cf\u5316\uff0c\u5176\u4e2d\u6839\u636e\u4e0d\u540c\u7f51\u7edc\u7ec4\u4ef6\u7684\u654f\u611f\u5ea6\u5206\u914d\u53ef\u53d8\u4f4d\u7cbe\u5ea6\u3002\u5173\u952e\u6280\u672f\u89c1\u89e3\u5305\u62ec\uff1a \u9009\u62e9\u6027\u7cbe\u5ea6\u5206\u914d\uff1a\u521d\u59cb\u5bc6\u96c6\u5c42\u548c\u5411\u4e0b\u6295\u5f71\uff08down_proj\uff09\u77e9\u9635\u5bf9\u4e8e\u8bbe\u7f6e\u7a33\u5b9a\u7684\u8868\u793a\u548c\u7ba1\u7406 SwiGLU \u6fc0\u6d3b\u4e2d\u7684\u7f29\u653e\u5c5e\u6027\u81f3\u5173\u91cd\u8981\uff0c\u5b83\u4eec\u4fdd\u6301\u5728\u66f4\u9ad8\u7684\u7cbe\u5ea6\uff084 \u4f4d\u6216 6 \u4f4d\uff09\u3002\u76f8\u53cd\uff0c\u5927\u90e8\u5206\u53c2\u6570\uff08\u4e3b\u8981\u662f\u6df7\u5408\u4e13\u5bb6 (MoE) \u5c42\u4e2d\u7684\u53c2\u6570\uff0c\u7ea6\u5360\u6a21\u578b\u7684 88%\uff09\u88ab\u79ef\u6781\u91cf\u5316\u4e3a 1.5-2 \u4f4d\u3002 \u91cd\u8981\u6027\u77e9\u9635\u6821\u51c6\uff1a\u5728\u91cf\u5316\u8fc7\u7a0b\u4e2d\u52a0\u5165\u91cd\u8981\u6027\u77e9\u9635\u5141\u8bb8\u8be5\u65b9\u6cd5\u52a8\u6001\u8c03\u6574\u6bcf\u5c42\u7684\u7cbe\u5ea6\u6c34\u5e73\u3002\u6b64\u6821\u51c6\u53ef\u9632\u6b62\u5e38\u89c1\u7684\u9677\u9631\uff0c\u4f8b\u5982\u901a\u5e38\u7531\u5747\u5300\u91cf\u5316\u5f15\u8d77\u7684\u65e0\u9650\u5faa\u73af\u6216\u65e0\u610f\u4e49\u7684\u8f93\u51fa\u3002 \u5c42\u7279\u5b9a\u7684\u654f\u611f\u6027\u5206\u6790\uff1a\u6280\u672f\u8bc4\u4f30\u8868\u660e\uff0c\u867d\u7136 MoE \u5c42\u53ef\u4ee5\u5bb9\u5fcd\u8f83\u4f4e\u7684\u7cbe\u5ea6\uff0c\u4f46\u6ce8\u610f\u673a\u5236\u3001\u5d4c\u5165\u5c42\u548c\u6700\u7ec8\u8f93\u51fa\u5934\u7b49\u7ec4\u4ef6\u9700\u8981\u66f4\u591a\u4f4d\u6765\u4fdd\u7559\u6fc0\u6d3b\u5206\u5e03\u3002\u8fd9\u79cd\u7ec6\u81f4\u5165\u5fae\u7684\u7b56\u7565\u53ef\u786e\u4fdd\u8ba1\u7b97\u56fe\u4e2d\u7684\u5173\u952e\u8def\u5f84\u4fdd\u6301\u8db3\u591f\u7684\u4fdd\u771f\u5ea6\u3002 1\u3001\u91cf\u5316\u6a21\u578b\u53d8\u4f53\u548c\u6027\u80fd Unsloth AI \u53d1\u5e03\u4e86\u591a\u4e2a\u52a8\u6001\u91cf\u5316\u53d8\u4f53\uff0c\u6bcf\u4e2a\u53d8\u4f53\u90fd\u5e73\u8861\u4e86\u6a21\u578b\u5927\u5c0f\u548c\u8f93\u51fa\u8d28\u91cf\uff1a \u4f8b\u5982\uff0c\u5728\u53d7\u63a7\u6d4b\u8bd5\u4e2d\uff0c\u6a21\u578b\u7684\u4efb\u52a1\u662f\u751f\u6210 Flappy Bird \u6e38\u620f\u7684 Python \u5b9e\u73b0\uff0c\u5373\u4f7f\u662f\u6700\u5c0f\u7684 1.58 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[],"class_list":["post-53779","post","type-post","status-publish","format-standard","hentry","category-ai"],"_links":{"self":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/53779","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/comments?post=53779"}],"version-history":[{"count":0,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/posts\/53779\/revisions"}],"wp:attachment":[{"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/media?parent=53779"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/categories?post=53779"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fwq.ai\/blog\/wp-json\/wp\/v2\/tags?post=53779"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}