{"id":32039,"date":"2025-07-02T21:14:11","date_gmt":"2025-07-02T13:14:11","guid":{"rendered":"https:\/\/www.kdjingpai.com\/?p=32039"},"modified":"2025-07-02T21:14:11","modified_gmt":"2025-07-02T13:14:11","slug":"glm-41v-thinking-2","status":"publish","type":"post","link":"https:\/\/www.kdjingpai.com\/pt\/glm-41v-thinking-2\/","title":{"rendered":"GLM-4.1V-Thinking\uff1a\u5f00\u6e90\u89c6\u89c9\u63a8\u7406\u6a21\u578b\uff0c\u652f\u6301\u591a\u6a21\u6001\u590d\u6742\u4efb\u52a1"},"content":{"rendered":"<p>GLM-4.1V-Thinking \u662f\u4e00\u4e2a\u7531\u6e05\u534e\u5927\u5b66 KEG \u5b9e\u9a8c\u5ba4\uff08THUDM\uff09\u5f00\u53d1\u7684\u5f00\u6e90\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u591a\u6a21\u6001\u63a8\u7406\u80fd\u529b\u3002\u57fa\u4e8e GLM-4-9B-0414 \u57fa\u7840\u6a21\u578b\uff0cGLM-4.1V-Thinking \u901a\u8fc7\u5f3a\u5316\u5b66\u4e60\u548c\u201c\u601d\u7ef4\u94fe\u201d\u63a8\u7406\u673a\u5236\uff0c\u663e\u8457\u63d0\u5347\u4e86\u590d\u6742\u4efb\u52a1\u7684\u5904\u7406\u80fd\u529b\u3002\u5b83\u652f\u6301 64k \u8d85\u957f\u4e0a\u4e0b\u6587\u30014K \u9ad8\u5206\u8fa8\u7387\u56fe\u50cf\u5904\u7406\uff0c\u5e76\u517c\u5bb9\u4efb\u610f\u56fe\u50cf\u5bbd\u9ad8\u6bd4\uff0c\u540c\u65f6\u652f\u6301\u4e2d\u82f1\u6587\u53cc\u8bed\u3002\u8be5\u6a21\u578b\u5728\u6570\u5b66\u3001\u4ee3\u7801\u3001\u957f\u6587\u6863\u7406\u89e3\u548c\u89c6\u9891\u63a8\u7406\u7b49\u4efb\u52a1\u4e2d\u8868\u73b0\u51fa\u8272\uff0c\u90e8\u5206\u8bc4\u6d4b\u751a\u81f3\u8d85\u8d8a\u4e86 GPT-4o\u3002\u4ee3\u7801\u548c\u6a21\u578b\u5df2\u5728 GitHub \u4e0a\u5f00\u653e\uff0c\u91c7\u7528 MIT \u8bb8\u53ef\u8bc1\uff0c\u5141\u8bb8\u514d\u8d39\u5546\u7528\uff0c\u9002\u5408\u5f00\u53d1\u8005\u3001\u7814\u7a76\u4eba\u5458\u548c\u4f01\u4e1a\u4f7f\u7528\u3002<\/p>\n<p><img loading=\"lazy\" decoding=\"async\" class=\"size-full wp-image-32025 aligncenter\" src=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/07\/47019d0656d1265.jpeg\" alt=\"\" width=\"688\" height=\"500\" srcset=\"https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/07\/47019d0656d1265.jpeg 688w, https:\/\/www.kdjingpai.com\/wp-content\/uploads\/2025\/07\/47019d0656d1265-18x12.jpeg 18w\" sizes=\"auto, (max-width: 688px) 100vw, 688px\" \/><\/p>\n<p>&nbsp;<\/p>\n<h2>\u529f\u80fd\u5217\u8868<\/h2>\n<ul>\n<li>\u652f\u6301 64k \u8d85\u957f\u4e0a\u4e0b\u6587\uff0c\u5904\u7406\u957f\u6587\u6863\u6216\u590d\u6742\u5bf9\u8bdd\u3002<\/li>\n<li>\u5904\u7406 4K \u9ad8\u5206\u8fa8\u7387\u56fe\u50cf\uff0c\u652f\u6301\u4efb\u610f\u5bbd\u9ad8\u6bd4\u3002<\/li>\n<li>\u63d0\u4f9b\u4e2d\u82f1\u6587\u53cc\u8bed\u652f\u6301\uff0c\u9002\u5408\u591a\u8bed\u8a00\u573a\u666f\u3002<\/li>\n<li>\u96c6\u6210\u201c\u601d\u7ef4\u94fe\u201d\u63a8\u7406\u673a\u5236\uff0c\u63d0\u5347\u6570\u5b66\u3001\u4ee3\u7801\u548c\u903b\u8f91\u4efb\u52a1\u7684\u51c6\u786e\u6027\u3002<\/li>\n<li>\u652f\u6301\u89c6\u9891\u63a8\u7406\uff0c\u53ef\u5206\u6790\u89c6\u9891\u5185\u5bb9\u5e76\u56de\u7b54\u76f8\u5173\u95ee\u9898\u3002<\/li>\n<li>\u5f00\u6e90\u4ee3\u7801\u548c\u6a21\u578b\uff0c\u57fa\u4e8e MIT \u8bb8\u53ef\u8bc1\uff0c\u5141\u8bb8\u514d\u8d39\u5546\u7528\u3002<\/li>\n<li>\u63d0\u4f9b Hugging Face \u548c ModelScope \u5728\u7ebf\u6f14\u793a\uff0c\u5feb\u901f\u4f53\u9a8c\u6a21\u578b\u80fd\u529b\u3002<\/li>\n<li>\u652f\u6301\u5355\u5f20 3090 \u663e\u5361\u8fd0\u884c\uff0c\u9002\u5408\u8d44\u6e90\u6709\u9650\u7684\u5f00\u53d1\u73af\u5883\u3002<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u4f7f\u7528\u5e2e\u52a9<\/h2>\n<h3>\u5b89\u88c5\u4e0e\u90e8\u7f72<\/h3>\n<p>GLM-4.1V-Thinking \u63d0\u4f9b\u5b8c\u6574\u7684\u4ee3\u7801\u548c\u6a21\u578b\u6587\u4ef6\uff0c\u90e8\u7f72\u8fc7\u7a0b\u7b80\u5355\uff0c\u9002\u5408\u5f00\u53d1\u8005\u5728\u672c\u5730\u6216\u670d\u52a1\u5668\u4e0a\u8fd0\u884c\u3002\u4ee5\u4e0b\u662f\u8be6\u7ec6\u7684\u5b89\u88c5\u548c\u4f7f\u7528\u6b65\u9aa4\uff1a<\/p>\n<h4>1. \u73af\u5883\u51c6\u5907<\/h4>\n<p>\u9700\u8981\u5728\u652f\u6301 GPU \u7684\u73af\u5883\u4e2d\u8fd0\u884c\uff0c\u63a8\u8350\u4f7f\u7528 NVIDIA \u663e\u5361\uff08\u5982 RTX 3090\uff09\u3002\u786e\u4fdd\u5df2\u5b89\u88c5 Python 3.8 \u6216\u4ee5\u4e0a\u7248\u672c\uff0c\u4ee5\u53ca PyTorch\u3002\u4ee5\u4e0b\u662f\u5b89\u88c5\u4f9d\u8d56\u7684\u6b65\u9aa4\uff1a<\/p>\n<pre><code>pip install git+https:\/\/github.com\/huggingface\/transformers.git\r\npip install torch torchvision torchaudio\r\npip install -r requirements.txt\r\n<\/code><\/pre>\n<p>\u5982\u679c\u9700\u8981\u8fdb\u884c\u6a21\u578b\u5fae\u8c03\uff0c\u53ef\u53c2\u8003\u00a0<code>finetune\/README.md<\/code> \u6587\u4ef6\uff0c\u4f7f\u7528 LLaMA-Factory \u5de5\u5177\u5305\u3002\u5fae\u8c03\u65f6\u5efa\u8bae\u4f7f\u7528 Zero3 \u7b56\u7565\u4ee5\u786e\u4fdd\u8bad\u7ec3\u7a33\u5b9a\u6027\uff0c\u907f\u514d Zero2 \u53ef\u80fd\u5bfc\u81f4\u7684\u96f6\u635f\u5931\u95ee\u9898\u3002<\/p>\n<h4>2. \u4e0b\u8f7d\u6a21\u578b<\/h4>\n<p>GLM-4.1V-Thinking \u6a21\u578b\u53ef\u4ece Hugging Face \u6216 GitHub \u4ed3\u5e93\u4e0b\u8f7d\u3002\u8fd0\u884c\u4ee5\u4e0b\u4ee3\u7801\u52a0\u8f7d\u6a21\u578b\uff1a<\/p>\n<pre><code>from transformers import AutoProcessor, Glm4vForConditionalGeneration\r\nimport torch\r\nMODEL_PATH = \"THUDM\/GLM-4.1V-9B-Thinking\"\r\nprocessor = AutoProcessor.from_pretrained(MODEL_PATH, use_fast=True)\r\nmodel = Glm4vForConditionalGeneration.from_pretrained(\r\npretrained_model_name_or_path=MODEL_PATH,\r\ntorch_dtype=torch.bfloat16,\r\ndevice_map=\"auto\"\r\n)\r\n<\/code><\/pre>\n<p>\u6a21\u578b\u652f\u6301\u00a0<code>bfloat16<\/code> \u683c\u5f0f\uff0c\u964d\u4f4e\u5185\u5b58\u5360\u7528\uff0c\u9002\u5408\u5355 GPU \u8fd0\u884c\u3002<\/p>\n<h4>3. \u5355\u5f20\u56fe\u50cf\u63a8\u7406<\/h4>\n<p>GLM-4.1V-Thinking \u652f\u6301\u56fe\u50cf\u8f93\u5165\u7684\u63a8\u7406\u4efb\u52a1\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u56fe\u50cf\u63cf\u8ff0\u793a\u4f8b\uff1a<\/p>\n<pre><code>messages = [\r\n{\r\n\"role\": \"user\",\r\n\"content\": [\r\n{\"type\": \"image\", \"url\": \"https:\/\/example.com\/sample_image.png\"},\r\n{\"type\": \"text\", \"text\": \"\u63cf\u8ff0\u8fd9\u5f20\u56fe\u7247\"}\r\n]\r\n}\r\n]\r\ninputs = processor.apply_chat_template(\r\nmessages, tokenize=True, add_generation_prompt=True, return_dict=True, return_tensors=\"pt\"\r\n).to(model.device)\r\ngenerated_ids = model.generate(**inputs, max_new_tokens=8192)\r\noutput_text = processor.decode(generated_ids[0][inputs[\"input_ids\"].shape[1]:], skip_special_tokens=False)\r\nprint(output_text)\r\n<\/code><\/pre>\n<p>\u5c06\u00a0<code>sample_image.png<\/code> \u66ff\u6362\u4e3a\u5b9e\u9645\u56fe\u50cf URL \u6216\u672c\u5730\u8def\u5f84\u3002\u6a21\u578b\u4f1a\u5206\u6790\u56fe\u50cf\u5e76\u751f\u6210\u8be6\u7ec6\u63cf\u8ff0\u3002<\/p>\n<h4>4. \u89c6\u9891\u63a8\u7406<\/h4>\n<p>GLM-4.1V-Thinking \u652f\u6301\u89c6\u9891\u5185\u5bb9\u5206\u6790\u3002\u7528\u6237\u53ef\u901a\u8fc7 GitHub \u4ed3\u5e93\u4e2d\u7684\u793a\u4f8b\u4ee3\u7801\u6216\u5728\u7ebf\u6f14\u793a\u5e73\u53f0\uff08\u5982 Hugging Face\uff09\u4e0a\u4f20\u89c6\u9891\u6587\u4ef6\uff0c\u6a21\u578b\u5c06\u89e3\u6790\u89c6\u9891\u5e76\u56de\u7b54\u76f8\u5173\u95ee\u9898\u3002\u4f8b\u5982\uff0c\u4e0a\u4f20\u4e00\u6bb5\u4f1a\u8bae\u89c6\u9891\uff0c\u8be2\u95ee\u201c\u89c6\u9891\u4e2d\u8ba8\u8bba\u4e86\u54ea\u4e9b\u4e3b\u9898\u201d\uff0c\u6a21\u578b\u4f1a\u63d0\u53d6\u5173\u952e\u4fe1\u606f\u5e76\u751f\u6210\u51c6\u786e\u56de\u7b54\u3002<\/p>\n<h4>5. \u957f\u6587\u6863\u7406\u89e3<\/h4>\n<p>\u6a21\u578b\u652f\u6301 64k \u8d85\u957f\u4e0a\u4e0b\u6587\uff0c\u9002\u5408\u5904\u7406\u957f\u7bc7\u6587\u6863\u3002\u7528\u6237\u53ef\u5c06\u6587\u672c\u8f93\u5165\u6a21\u578b\uff0c\u8be2\u95ee\u6587\u6863\u4e2d\u7684\u5177\u4f53\u5185\u5bb9\u6216\u603b\u7ed3\u5173\u952e\u70b9\u3002\u4f8b\u5982\uff0c\u8f93\u5165\u4e00\u7bc7 50 \u9875\u7684\u5b66\u672f\u8bba\u6587\uff0c\u8be2\u95ee\u201c\u8bba\u6587\u7684\u4e3b\u8981\u7ed3\u8bba\u662f\u4ec0\u4e48\u201d\uff0c\u6a21\u578b\u4f1a\u5feb\u901f\u63d0\u53d6\u5e76\u603b\u7ed3\u3002<\/p>\n<h4>6. \u5728\u7ebf\u6f14\u793a<\/h4>\n<p>\u65e0\u9700\u672c\u5730\u90e8\u7f72\uff0c\u53ef\u901a\u8fc7 Hugging Face \u6216 ModelScope \u63d0\u4f9b\u7684\u5728\u7ebf\u6f14\u793a\u76f4\u63a5\u4f53\u9a8c\u3002\u8bbf\u95ee\u4ee5\u4e0b\u94fe\u63a5\uff1a<\/p>\n<ul>\n<li>Hugging Face \u6f14\u793a\uff1a<code>https:\/\/huggingface.co\/THUDM\/GLM-4.1V-9B-Thinking<\/code><\/li>\n<li>ModelScope \u6f14\u793a\uff1a<code>https:\/\/modelscope.cn\/models\/THUDM\/GLM-4.1V-9B-Thinking<\/code><br \/>\n\u7528\u6237\u53ef\u4e0a\u4f20\u56fe\u50cf\u3001\u89c6\u9891\u6216\u8f93\u5165\u6587\u672c\uff0c\u5feb\u901f\u6d4b\u8bd5\u6a21\u578b\u7684\u63a8\u7406\u80fd\u529b\u3002<\/li>\n<\/ul>\n<h4>7. \u5fae\u8c03\u6a21\u578b<\/h4>\n<p>\u5f00\u53d1\u8005\u53ef\u4f7f\u7528 LLaMA-Factory \u5de5\u5177\u5305\u5bf9\u6a21\u578b\u8fdb\u884c\u5fae\u8c03\uff0c\u4ee5\u9002\u914d\u7279\u5b9a\u4efb\u52a1\u3002\u5fae\u8c03\u914d\u7f6e\u6587\u4ef6\u4f4d\u4e8e\u00a0<code>configs\/lora.yaml<\/code>\uff0c\u8fd0\u884c\u4ee5\u4e0b\u547d\u4ee4\u5f00\u59cb\u5fae\u8c03\uff1a<\/p>\n<pre><code>cd finetune\r\npython finetune.py data\/YourDataset\/ THUDM\/GLM-4-9B-0414 configs\/lora.yaml\r\n<\/code><\/pre>\n<p>\u786e\u4fdd\u6570\u636e\u96c6\u683c\u5f0f\u6b63\u786e\uff0c\u63a8\u8350\u4f7f\u7528 JSON \u683c\u5f0f\u3002\u5fae\u8c03\u540e\uff0c\u6a21\u578b\u53ef\u66f4\u597d\u5730\u9002\u914d\u7279\u5b9a\u9886\u57df\u7684\u4efb\u52a1\uff0c\u5982\u533b\u7597\u56fe\u50cf\u5206\u6790\u6216\u6cd5\u5f8b\u6587\u6863\u5904\u7406\u3002<\/p>\n<h3>\u7279\u8272\u529f\u80fd\u64cd\u4f5c<\/h3>\n<ul>\n<li><strong>\u601d\u7ef4\u94fe\u63a8\u7406<\/strong>\uff1a\u6a21\u578b\u901a\u8fc7\u201c\u601d\u7ef4\u94fe\u201d\u673a\u5236\u5206\u89e3\u590d\u6742\u95ee\u9898\u3002\u4f8b\u5982\uff0c\u5728\u6570\u5b66\u4efb\u52a1\u4e2d\uff0c\u6a21\u578b\u4f1a\u9010\u6b65\u63a8\u5bfc\u7b54\u6848\uff0c\u786e\u4fdd\u7ed3\u679c\u51c6\u786e\u3002\u7528\u6237\u8f93\u5165\u201c\u6c42\u89e3\u4e8c\u6b21\u65b9\u7a0b x\u00b2 + 2x &#8211; 3 = 0\u201d\uff0c\u6a21\u578b\u4f1a\u8f93\u51fa\u8be6\u7ec6\u7684\u89e3\u9898\u6b65\u9aa4\u3002<\/li>\n<li><strong>\u591a\u6a21\u6001\u652f\u6301<\/strong>\uff1a\u7528\u6237\u53ef\u540c\u65f6\u8f93\u5165\u56fe\u50cf\u548c\u6587\u672c\u3002\u4f8b\u5982\uff0c\u4e0a\u4f20\u4e00\u5f20\u7535\u8def\u56fe\u5e76\u8be2\u95ee\u201c\u7535\u8def\u7684\u5de5\u4f5c\u539f\u7406\u662f\u4ec0\u4e48\u201d\uff0c\u6a21\u578b\u4f1a\u7ed3\u5408\u56fe\u50cf\u548c\u95ee\u9898\u751f\u6210\u8be6\u7ec6\u89e3\u91ca\u3002<\/li>\n<li><strong>\u4e2d\u82f1\u6587\u53cc\u8bed<\/strong>\uff1a\u6a21\u578b\u652f\u6301\u4e2d\u82f1\u6587\u6df7\u5408\u8f93\u5165\uff0c\u9002\u5408\u8de8\u8bed\u8a00\u573a\u666f\u3002\u4f8b\u5982\uff0c\u8f93\u5165\u4e2d\u6587\u95ee\u9898\u548c\u82f1\u6587\u56fe\u50cf\u63cf\u8ff0\uff0c\u6a21\u578b\u4f1a\u4ee5\u6307\u5b9a\u8bed\u8a00\u56de\u7b54\u3002<\/li>\n<\/ul>\n<h3>\u6ce8\u610f\u4e8b\u9879<\/h3>\n<ul>\n<li>\u786e\u4fdd GPU \u5185\u5b58\u5145\u8db3\uff0c\u63a8\u8350\u81f3\u5c11 24GB \u663e\u5b58\u3002<\/li>\n<li>\u957f\u4e0a\u4e0b\u6587\u5904\u7406\u65f6\uff0c\u542f\u7528 YaRN \u914d\u7f6e\u4ee5\u4f18\u5316\u6027\u80fd\uff0c\u914d\u7f6e\u6587\u4ef6\u4e3a\u00a0<code>config.json<\/code>\u00a0\u4e2d\u7684\u00a0<code>\"rope_scaling\": {\"type\": \"yarn\", \"factor\": 4.0}<\/code>\u3002<\/li>\n<li>\u6a21\u578b\u63a8\u7406\u901f\u5ea6\u4f9d\u8d56\u786c\u4ef6\uff0c3090 \u663e\u5361\u53ef\u5b9e\u73b0\u5b9e\u65f6\u54cd\u5e94\u3002<\/li>\n<\/ul>\n<p>&nbsp;<\/p>\n<h2>\u5e94\u7528\u573a\u666f<\/h2>\n<ol>\n<li><strong>\u5b66\u672f\u7814\u7a76<\/strong><br \/>\n\u7814\u7a76\u4eba\u5458\u53ef\u4f7f\u7528 GLM-4.1V-Thinking \u5206\u6790\u957f\u7bc7\u5b66\u672f\u8bba\u6587\uff0c\u63d0\u53d6\u5173\u952e\u7ed3\u8bba\u6216\u603b\u7ed3\u5185\u5bb9\u3002\u6a21\u578b\u8fd8\u80fd\u5904\u7406\u5b9e\u9a8c\u56fe\u50cf\uff0c\u8f85\u52a9\u5206\u6790\u6570\u636e\u56fe\u8868\u3002<\/li>\n<li><strong>\u6559\u80b2\u652f\u6301<\/strong><br \/>\n\u5b66\u751f\u53ef\u4e0a\u4f20\u6570\u5b66\u9898\u76ee\u6216\u79d1\u5b66\u5b9e\u9a8c\u56fe\u7247\uff0c\u6a21\u578b\u4f1a\u63d0\u4f9b\u8be6\u7ec6\u89e3\u9898\u6b65\u9aa4\u6216\u5b9e\u9a8c\u89e3\u91ca\uff0c\u9002\u5408\u81ea\u5b66\u6216\u6559\u5b66\u8f85\u52a9\u3002<\/li>\n<li><strong>\u5185\u5bb9\u521b\u4f5c<\/strong><br \/>\n\u521b\u4f5c\u8005\u53ef\u8f93\u5165\u89c6\u9891\u6216\u56fe\u50cf\u7d20\u6750\uff0c\u751f\u6210\u63cf\u8ff0\u6027\u6587\u672c\u6216\u521b\u610f\u811a\u672c\u3002\u4f8b\u5982\uff0c\u8f93\u5165\u65c5\u6e38\u89c6\u9891\uff0c\u751f\u6210\u666f\u70b9\u4ecb\u7ecd\u3002<\/li>\n<li><strong>\u4f01\u4e1a\u5e94\u7528<\/strong><br \/>\n\u4f01\u4e1a\u53ef\u5c06\u6a21\u578b\u7528\u4e8e\u6587\u6863\u81ea\u52a8\u5316\u5904\u7406\uff0c\u5982\u5206\u6790\u5408\u540c\u6761\u6b3e\u6216\u751f\u6210\u62a5\u544a\u3002\u652f\u6301\u4e2d\u82f1\u6587\u53cc\u8bed\uff0c\u9002\u5408\u8de8\u56fd\u4f01\u4e1a\u3002<\/li>\n<\/ol>\n<p>&nbsp;<\/p>\n<h2>QA<\/h2>\n<ol>\n<li><strong>GLM-4.1V-Thinking \u652f\u6301\u54ea\u4e9b\u8f93\u5165\u7c7b\u578b\uff1f<\/strong><br \/>\n\u6a21\u578b\u652f\u6301\u56fe\u50cf\u3001\u89c6\u9891\u3001\u6587\u672c\u8f93\u5165\uff0c\u517c\u5bb9 4K \u56fe\u50cf\u548c 64k \u4e0a\u4e0b\u6587\uff0c\u9002\u5408\u591a\u6a21\u6001\u4efb\u52a1\u3002<\/li>\n<li><strong>\u662f\u5426\u9700\u8981\u9ad8\u6027\u80fd\u786c\u4ef6\uff1f<\/strong><br \/>\n\u5355\u5f20 RTX 3090 \u663e\u5361\u5373\u53ef\u8fd0\u884c\uff0c\u63a8\u8350 24GB \u663e\u5b58\u4ee5\u786e\u4fdd\u6d41\u7545\u63a8\u7406\u3002<\/li>\n<li><strong>\u5982\u4f55\u8fdb\u884c\u6a21\u578b\u5fae\u8c03\uff1f<\/strong><br \/>\n\u4f7f\u7528 LLaMA-Factory \u5de5\u5177\u5305\uff0c\u53c2\u8003 GitHub \u4ed3\u5e93\u4e2d\u7684\u00a0<code>finetune\/README.md<\/code>\u00a0\u6587\u4ef6\uff0c\u914d\u7f6e\u00a0<code>lora.yaml<\/code>\u00a0\u8fdb\u884c\u5fae\u8c03\u3002<\/li>\n<li><strong>\u6a21\u578b\u662f\u5426\u514d\u8d39\uff1f<\/strong><br \/>\n\u662f\u7684\uff0c\u6a21\u578b\u57fa\u4e8e MIT \u8bb8\u53ef\u8bc1\u5f00\u6e90\uff0c\u5141\u8bb8\u514d\u8d39\u5546\u7528\u3002<\/li>\n<li><strong>\u5982\u4f55\u4f53\u9a8c\u6a21\u578b\uff1f<\/strong><br \/>\n\u53ef\u901a\u8fc7 Hugging Face \u6216 ModelScope \u7684\u5728\u7ebf\u6f14\u793a\u4e0a\u4f20\u56fe\u50cf\u6216\u6587\u672c\uff0c\u5feb\u901f\u6d4b\u8bd5\u6a21\u578b\u529f\u80fd\u3002<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>GLM-4.1V-Thinking \u662f\u4e00\u4e2a\u7531\u6e05\u534e\u5927\u5b66 KEG \u5b9e\u9a8c\u5ba4\uff08THUDM\uff09\u5f00\u53d1\u7684\u5f00\u6e90\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff0c\u4e13\u6ce8\u4e8e\u591a\u6a21\u6001\u63a8\u7406\u80fd\u529b\u3002\u57fa\u4e8e GLM-4-9B-0414 \u57fa\u7840\u6a21\u578b\uff0cGLM-4.1V-Thinking \u901a\u8fc7\u5f3a\u5316\u5b66\u4e60\u548c\u201c\u601d\u7ef4\u94fe\u201d\u63a8\u7406\u673a\u5236&#8230;<\/p>\n","protected":false},"author":1,"featured_media":62358,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[392,402,401],"tags":[230,378],"class_list":["post-32039","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-models","category-multimodal-models","category-visual-recognition-model","tag-aikaiyuanxiangmu","tag-shijuemubiaojiance"],"_links":{"self":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/posts\/32039","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/comments?post=32039"}],"version-history":[{"count":0,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/posts\/32039\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/media\/62358"}],"wp:attachment":[{"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/media?parent=32039"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/categories?post=32039"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kdjingpai.com\/pt\/wp-json\/wp\/v2\/tags?post=32039"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}