Instructions to use zhongweixie/PocketWorker-8B-ClawEval-GRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use zhongweixie/PocketWorker-8B-ClawEval-GRPO with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="zhongweixie/PocketWorker-8B-ClawEval-GRPO") 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("zhongweixie/PocketWorker-8B-ClawEval-GRPO") model = AutoModelForMultimodalLM.from_pretrained("zhongweixie/PocketWorker-8B-ClawEval-GRPO", 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 zhongweixie/PocketWorker-8B-ClawEval-GRPO with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "zhongweixie/PocketWorker-8B-ClawEval-GRPO" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "zhongweixie/PocketWorker-8B-ClawEval-GRPO", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/zhongweixie/PocketWorker-8B-ClawEval-GRPO
- SGLang
How to use zhongweixie/PocketWorker-8B-ClawEval-GRPO 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 "zhongweixie/PocketWorker-8B-ClawEval-GRPO" \ --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": "zhongweixie/PocketWorker-8B-ClawEval-GRPO", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'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 "zhongweixie/PocketWorker-8B-ClawEval-GRPO" \ --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": "zhongweixie/PocketWorker-8B-ClawEval-GRPO", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use zhongweixie/PocketWorker-8B-ClawEval-GRPO with Docker Model Runner:
docker model run hf.co/zhongweixie/PocketWorker-8B-ClawEval-GRPO
qwen3vl-8b-claw-rl-grpo-stage3-iter560
在内部 agent 任务集上做 RL 微调得到的 Qwen3-VL-8B 模型。
这不是 Qwen3-VL-8B-Instruct 本身,上游那份模型卡里的任何跑分都不适用于本模型。
训练配置
| 项 | 值 |
|---|---|
| 实验名 | exp_rl_v5_v13_grpo_stage3 |
| 算法 | GRPO |
| 变体 | 基线。与 GDPO 基线仅 advantage estimator 不同。 |
| 起点 | stage3 SFT checkpoint (exp_stage3_v2_merged),其上游为 Qwen3-VL-8B-Instruct |
| 最终 checkpoint | iter_0000560 |
| 训练量 | 561 个 rollout step,约 3 个 epoch,187 条 prompt |
| samples / prompt | 8 |
| rollout batch | 1 |
| global batch | 8 |
| 硬件 | 8xH800,TP=8,colocate |
| 奖励 judge | qwen/qwen-2.5-72b-instruct (OpenRouter) |
学习率、KL 系数等由 slime 的 formal profile 提供,未在启动脚本中显式设置,
故此处不列出,以免给出未经核实的数值。
评测状态
目前没有可信的 held-out 评测数字。 不要把本模型当作已验证优于其起点。
原因:
- cold-start 基线作业(373736)因 sglang JIT 缺
ninja未能启动,缺少可比基线。 - held-out 驱动脚本会主动拒绝一部分任务(沙箱快照类、多模态输入类), 覆盖率不足 30 题全集,其总分不能当作全集分数引用。
起点 SFT checkpoint 记录的 eval=0.506 是起点的分数,不是本模型的分数。
用法
from transformers import Qwen3VLForConditionalGeneration, AutoProcessor
model = Qwen3VLForConditionalGeneration.from_pretrained(
"zhongweixie/qwen3vl-8b-claw-rl-grpo-stage3-iter560", dtype="auto", device_map="auto"
)
processor = AutoProcessor.from_pretrained("zhongweixie/qwen3vl-8b-claw-rl-grpo-stage3-iter560")
来源
由 slime/tools/convert_torch_dist_to_hf.py 从 Megatron 分布式 checkpoint
/aifs4su/hansirui_3rd/rl-checkpoints/exp_rl_v5_v13_grpo_stage3/iter_0000560 转换而来,缺失权重与 tokenizer/config 取自
exp_stage3_v2_merged_sharded(即该次 RL 训练实际使用的基座)。
- Downloads last month
- 26
Model tree for zhongweixie/PocketWorker-8B-ClawEval-GRPO
Base model
Qwen/Qwen3-VL-8B-Instruct