88 lines
3.1 KiB
Markdown
88 lines
3.1 KiB
Markdown
---
|
||
library_name: transformers
|
||
tags:
|
||
- agent
|
||
- code
|
||
license: mit
|
||
datasets:
|
||
- ricdomolm/mini-coder-trajs-400k
|
||
base_model:
|
||
- Qwen/Qwen3-1.7B
|
||
---
|
||
|
||
# mini-coder-1.7b
|
||
|
||
`mini-coder-1.7b` is a 1.7B parameter model distilled from Qwen 3 Coder 30B A3B. It punches well above its weight, outperforming SWE-agent-LM 7B on [SWE-bench Verified Bash only](https://www.swebench.com/):
|
||
|
||
<div align="center">
|
||
|
||
| Model | pass@1 | pass@100 |
|
||
|-------------------------|--------|----------|
|
||
| Qwen 3 Coder 30B-A3B | 33.2 | 67.4 |
|
||
| mini-swe-4b | 26.8 | 60.2 |
|
||
| gpt-oss-120b | 26.0 | – |
|
||
| mini-swe-1.7b | 18.6 | 50.4 |
|
||
| SWE-agent-LM 7B | 15.2 | – |
|
||
| Qwen 3 4B Instruct 2507 | 4.0 | 25.1 |
|
||
|
||
</div>
|
||
|
||
It is trained on 400k training trajectories using the lightweight [mini-swe-agent](https://mini-swe-agent.com/latest/) scaffolding and the [SWE-smith](https://huggingface.co/datasets/SWE-bench/SWE-smith) dataset of GitHub issues.
|
||
|
||
Unlike existing agentic SWE models, the `mini-coder` models can be post-trained on a single 80GB GPU—or smaller. They work seamlessly with mini-swe-agent, a lightweight, scalable, and developer-friendly agentic framework well-suited for RL fine-tuning. And because they are dense rather than MoE models, they benefit from a more mature fine-tuning ecosystem.
|
||
|
||
## Example usage: Generating SWE-bench trajectories with mini-swe-agent and vLLM
|
||
|
||
This example shows how to generate SWE-bench trajectories using [mini-swe-agent](https://mini-swe-agent.com/latest/) as the agentic scaffolding (recommended) and [vLLM](https://docs.vllm.ai/en/latest/) as the local inference engine.
|
||
|
||
First, launch a vLLM server with your chosen model. For example:
|
||
|
||
```bash
|
||
vllm serve ricdomolm/mini-coder-1.7b &
|
||
```
|
||
|
||
By default, the server will be available at `http://localhost:8000`.
|
||
|
||
Second, edit the mini-swe-agent SWE-bench config file located in `src/minisweagent/config/extra/swebench.yaml` to include your local vLLM model:
|
||
|
||
```yaml
|
||
model:
|
||
model_name: "hosted_vllm/ricdomolm/mini-coder-1.7b" # or hosted_vllm/path/to/local/model
|
||
model_kwargs:
|
||
api_base: "http://localhost:8000/v1" # adjust if using a non-default port/address
|
||
```
|
||
|
||
Create a litellm `registry.json` file:
|
||
|
||
```bash
|
||
cat > registry.json <<'EOF'
|
||
{
|
||
"ricdomolm/mini-coder-1.7b": {
|
||
"max_tokens": 40960,
|
||
"input_cost_per_token": 0.0,
|
||
"output_cost_per_token": 0.0,
|
||
"litellm_provider": "hosted_vllm",
|
||
"mode": "chat"
|
||
}
|
||
}
|
||
EOF
|
||
```
|
||
|
||
Now you’re ready to generate trajectories! Let's solve the `django__django-11099` instance of SWE-bench Verified:
|
||
|
||
```bash
|
||
LITELLM_MODEL_REGISTRY_PATH=registry.json mini-extra swebench --output test/ --subset verified --split test --filter '^(django__django-11099)$'
|
||
```
|
||
|
||
You should now see the generated trajectory in the `test/` directory.
|
||
|
||
## Citation
|
||
|
||
```bibtext
|
||
@article{olmedo2026computational,
|
||
title={Computational Arbitrage in AI Model Markets},
|
||
author={Olmedo, Ricardo and Sch{\"o}lkopf, Bernhard and Hardt, Moritz},
|
||
journal={arXiv preprint arXiv:2603.22404},
|
||
year={2026}
|
||
}
|
||
``` |