GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
postel-MaoZhiqin
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
postel-MaoZhiqin
Model Tree
Qwen/Qwen3-4B
Input Modalities
Output Modalities
Supported Functionality
在 Qwen3-4B 上用 129 条自建样本做 LoRA SFT， 把一个特定的 Python 代码风格克隆到模型上：
```python 代码块，不夹带任何解释文字；Iterable / Mapping / Real），
返回值用最具体的实现类型（list / dict / float / Path）；"""docstring"""，边界条件（空输入、除零、负数）显式处理。注意：这是风格迁移，不是能力增强。基座不会因为这次微调变得更会写代码。
messages 格式样本：user 一句中文需求，assistant 一个完整可运行的 Python 文件train/train.json| 项目 | 值 |
|---|---|
| LoRA | r=16, alpha=32, dropout=0.05 |
| target_modules | q_proj, k_proj, v_proj, o_proj |
| epochs | 5（129 条 × 5 = 645 个样本，325 个优化步） |
| batch | per_device=1, gradient_accumulation=2 |
| learning_rate | 2e-4 |
| max_length | 2048 |
| loss | assistant-only（只对 assistant 回复算 loss） |
| 框架 | transformers 5.x + peft 0.21 + trl 1.17 |
| 硬件 | AMD Radeon RX 7900 XT（ROCm），单卡约 3 分钟 |
python
import torchfrom peft import PeftModelfrom transformers import AutoModelForCausalLM, AutoTokenizer base_id = "Qwen/Qwen3-4B"tokenizer = AutoTokenizer.from_pretrained(base_id)base = AutoModelForCausalLM.from_pretrained(base_id, torch_dtype="auto", device_map="auto")model = PeftModel.from_pretrained(base, "postel-MaoZhiqin/qwen3-4b-code-style-lora").merge_and_unload()model.eval() prompt = tokenizer.apply_chat_template( [{"role": "user", "content": "帮我写一个把秒数转成时分秒字符串的函数。"}], tokenize=False, add_generation_prompt=True, enable_thinking=False,)inputs = tokenizer(prompt, return_tensors="pt").to(model.device)with torch.no_grad(): out = model.generate(**inputs, max_new_tokens=600, do_sample=False)print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))sklearn 兼容的随机森林时会引用 BaseForest、_get_n_features 等并不存在的接口。毛治钦（Leo Postel）· GitHub @re-ITRT
基于 Qwen3-4B（Alibaba Cloud, Apache-2.0）。
import torchfrom peft import PeftModelfrom transformers import AutoModelForCausalLM, AutoTokenizer base_id = "Qwen/Qwen3-4B"tokenizer = AutoTokenizer.from_pretrained(base_id)base = AutoModelForCausalLM.from_pretrained(base_id, torch_dtype="auto", device_map="auto")model = PeftModel.from_pretrained(base, "postel-MaoZhiqin/qwen3-4b-code-style-lora").merge_and_unload()model.eval() prompt = tokenizer.apply_chat_template( [{"role": "user", "content": "帮我写一个把秒数转成时分秒字符串的函数。"}], tokenize=False, add_generation_prompt=True, enable_thinking=False,)inputs = tokenizer(prompt, return_tensors="pt").to(model.device)with torch.no_grad(): out = model.generate(**inputs, max_new_tokens=600, do_sample=False)print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))