GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
chenhaodev
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
chenhaodev
Model Tree
Qwen/Qwen3-8B
Input Modalities
Output Modalities
Supported Functionality
该专科模型训练了 6 大类技能，均以《Med-Books-20260914》领域书籍为知识来源：
/workspace/Med-Books-20260914/ 7 本书（4 本文字层 + 3 本 OCR 康复类），共 465 个领域 chunk，覆盖运动与体重管理。python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_id = "Qwen/Qwen3-8B"adapter = "chenhaodev/qwen3-8b-sport-weight-lora" tokenizer = AutoTokenizer.from_pretrained(base_id)model = AutoModelForCausalLM.from_pretrained( base_id, torch_dtype=torch.bfloat16, device_map="auto")model = PeftModel.from_pretrained(model, adapter) messages = [{"role": "user", "content": "我28岁男，175cm/82kg，想减脂并改善体能，请给我每周运动计划"}]text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer([text], return_tensors="pt").to(model.device)out = model.generate(**inputs, max_new_tokens=600)print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))本仓库同时提供配套 FAISS 检索知识库：chenhaodev/qwen3-8b-sport-weight-rag（BAAI/bge-small-zh-v1.5 512 维 L2 归一化，465 个文档块）。推荐问答时先检索相关文档块拼入 prompt，再让模型回答，以提高准确率、降低幻觉。
本模型另已合并且导出为 GGUF，量化 Q4_K_M 后可用 llama.cpp llama-server 本地运行（约 4.8GB），再与本仓库 FAISS 知识库联动，实现完全本地、离线、低成本的专科问答服务。
本模型仅用于健康/运动/体重管理的科普与辅助参考，不构成医疗诊断或治疗建议；严重健康问题请咨询专业医师。
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_id = "Qwen/Qwen3-8B"adapter = "chenhaodev/qwen3-8b-sport-weight-lora" tokenizer = AutoTokenizer.from_pretrained(base_id)model = AutoModelForCausalLM.from_pretrained( base_id, torch_dtype=torch.bfloat16, device_map="auto")model = PeftModel.from_pretrained(model, adapter) messages = [{"role": "user", "content": "我28岁男，175cm/82kg，想减脂并改善体能，请给我每周运动计划"}]text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer([text], return_tensors="pt").to(model.device)out = model.generate(**inputs, max_new_tokens=600)print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))