---
base_model: Qwen/Qwen3-0.6B
library_name: transformers
tags:
- generated_from_trainer
- trl
- grpo
- cricket
- merged
license: mit
pipeline_tag: text-generation
---

# cricket-captain-qwen3-06b-merged

`Qwen/Qwen3-0.6B` with the [`pratinavseth/cricket-captain-qwen3-06b-stage2`](https://huggingface.co/pratinavseth/cricket-captain-qwen3-06b-stage2) LoRA adapter (stage 2 GRPO, step 50) **merged into the base weights**. Single-file model, ready to load without PEFT — usable directly via `transformers`, `vllm`, or TGI.

- **Live Space:** https://huggingface.co/spaces/pratinavseth/cricket-captain-llm
- **Source code:** https://github.com/pratinavseth/cricket-captain-llm
- **Adapter (un-merged):** [`pratinavseth/cricket-captain-qwen3-06b-stage2`](https://huggingface.co/pratinavseth/cricket-captain-qwen3-06b-stage2)
- **W&B run:** https://wandb.ai/ptnv-s-research/huggingface

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("pratinavseth/cricket-captain-qwen3-06b-merged")
model = AutoModelForCausalLM.from_pretrained("pratinavseth/cricket-captain-qwen3-06b-merged", torch_dtype="bfloat16", device_map="auto")
```

The model expects the cricket-captain prompt schema produced by the OpenEnv environment in this repo (see `inference.py` for prompt construction).