Text Generation
Transformers
Safetensors
English
Chinese
qwen3_5
image-text-to-text
qwen
lora
reinforcement-learning
fp8
agentic-coding
scientific-computing
physics
materials-science
genetics
conversational
Instructions to use HoomanBio2021/LabOpen-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use HoomanBio2021/LabOpen-v1 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="HoomanBio2021/LabOpen-v1") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("HoomanBio2021/LabOpen-v1") model = AutoModelForMultimodalLM.from_pretrained("HoomanBio2021/LabOpen-v1", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use HoomanBio2021/LabOpen-v1 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "HoomanBio2021/LabOpen-v1" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "HoomanBio2021/LabOpen-v1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/HoomanBio2021/LabOpen-v1
- SGLang
How to use HoomanBio2021/LabOpen-v1 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "HoomanBio2021/LabOpen-v1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "HoomanBio2021/LabOpen-v1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "HoomanBio2021/LabOpen-v1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "HoomanBio2021/LabOpen-v1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use HoomanBio2021/LabOpen-v1 with Docker Model Runner:
docker model run hf.co/HoomanBio2021/LabOpen-v1
File size: 4,865 Bytes
5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 fec37cf 5171300 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 | # LabOpen v1 — 使用指南
## 快速开始
### 1. 环境要求
```
Python >= 3.10
vLLM >= 0.27(实测 0.27.1;本仓库唯一服务路径)
NVIDIA 计算能力 >= 8.9 的 GPU(H100 / L40S / B200 / DGX GB10 等级别)
FP8 下权重约 31 GB:单卡最低 48 GB(缩短上下文),推荐 80 GB;支持多卡张量并行
无 macOS 路径(vLLM 无 Apple Silicon 后端,且无 Mac 运行时能加载本分块 FP8 检查点格式)
```
### 2. 启动服务
```bash
vllm serve HoomanBio2021/LabOpen-v1
```
合并权重与基座 `Qwen/Qwen3.8-27B-FP8` 采用完全相同的分块 FP8(e4m3,128×128 块尺度)张量布局与文件结构,任何能加载基座的运行时(vLLM / SGLang)均可直接加载本仓库,无需任何改动。
实测参考参数(DGX GB10,vLLM 0.27.1):`--language-model-only --max-model-len 32768 --enforce-eager --reasoning-parser qwen3`。
**关键**:域计算请求的 `max_tokens` 应给到 **4096 以上**——模型先思考后作答,1024 预算会被思考链耗尽(表现为 `finish_reason=length`、正文为空)。配 `--reasoning-parser qwen3` 时,最终答案在 `content`、推理在 `reasoning_content`。
### 3. 适配器(研究工件——非服务路径)
`adapter/` 目录保存训练时的精确工件(PEFT 格式,112 MB),供检查、溯源与继续训练使用。它**不是**可服务的路径:原生 transformers 无法加载分块 FP8 基座;vLLM 的 LoRA 内核不覆盖本架构混合线性注意力投影(`in_proj_*`、`out_proj`),12 类目标模块中 5 类会被静默跳过。**服务请使用合并权重。**
> **不要自行对 FP8 基座执行 `merge_and_unload()`。** 朴素 PEFT 合并会把 BF16 增量直接加到 FP8 码上,静默损坏权重。本仓库的合并权重已按「反量化 → 加增量 → 按块重量化」正确生成并经与适配器的等价探针验证(144/144 逐 token 一致,见 `MERGE_REPORT.md`)。
### 4. 域查询(带标记)
六个强化域通过提示首行标记寻址(对服务的 OpenAI 兼容端点调用):
```python
from openai import OpenAI
client = OpenAI(base_url="http://127.0.0.1:8000/v1", api_key="EMPTY")
TAGS = {
"cfm": "[domain: cfm-chain]", # 流体物理链
"mat": "[domain: mat-table]", # 材料混合律
"bio": "[domain: bio-genetics]", # 经典遗传学
"ops": "[domain: ops-action]", # 运维动作查询
"prose": "[domain: general-prose]", # 开放式散文
}
def query(prompt, domain=None, temperature=0.3, max_tokens=4096):
if domain:
prompt = TAGS[domain] + "\n" + prompt
resp = client.chat.completions.create(
model="HoomanBio2021/LabOpen-v1",
messages=[{"role": "user", "content": prompt}],
temperature=temperature,
max_tokens=max_tokens,
)
return resp.choices[0].message.content
```
### 5. 使用示例
```python
# 材料混合律(期望返回 {"value": ..., "unit": ...} 形式 JSON)
query("An alloy: 60% Cu, 40% Zn. Compute density using mixture rules. Answer in JSON with value and unit.", domain="mat")
# 经典遗传学
query("Gene for plant height: Tt x Tt cross. What is the phenotype ratio?", domain="bio")
# 流体物理
query("A face with normal [0,1,0], velocity [2,0,0], pressure 100. Compute the force component.", domain="cfm")
# 散文式开放式回答
query("Explain why code review matters in software development.", domain="prose", temperature=0.3)
# 代码生成——无需标记,基座原生能力已保留
query("Write a Python function to merge two sorted lists.", temperature=0.2)
```
## 温度建议
| 场景 | 推荐温度 | 原因 |
|---|---|---|
| 域计算题 | 0.3 | 正确答案在分布肩部;greedy 可能偏离 |
| 散文写作 | 0.3 | 散文偏好需要采样激活;greedy 退化为结构化 |
| 代码生成 | 0.0–0.2 | 基座能力,无需标记 |
| 通用问答 | 0.3–0.7 | 基座行为,无标记 |
## 注意事项
1. **标记必须独占一行**,放在用户提示最前面
2. **不加标记时模型回退到基座行为**——代码/函数调用保持原生强度
3. 域计算给足 `max_tokens ≥ 4096`(思考链先于答案消耗预算)
4. 硬件自查:`nvidia-smi --query-gpu=compute_cap` 应 ≥ 8.9(A100/3090 等 Ampere 无 FP8 支持)
## 故障排查
| 问题 | 可能原因 | 解决方案 |
|---|---|---|
| 域题答非所问 | 忘加标记 | 检查是否加了正确的 `[domain: x]` 前缀 |
| 散文不稳定 | 用了 greedy | 改用 T=0.3 采样 |
| 返回正文为空 / finish=length | max_tokens 太小 | 提到 ≥4096(思考链耗尽预算) |
| 加载失败 | GPU 无 FP8 能力 | 核对 compute capability ≥ 8.9 |
| 自行合并后输出异常 | 对 FP8 做了朴素 PEFT 合并 | 使用本仓库已合并权重,勿自行 merge | |