Skip to content

Conversation

@peteli25
Copy link

gh pr create --title "feat: add SWE-bench dataset support" --body "## Description

This PR adds support for SWE-bench dataset evaluation in OpenCompass.

Changes

  • Add `swebench.py` dataset implementation with SWEBenchDataset class
  • Add `swebench_gen.py` configuration for dataset loading
  • Add `eval_swebench.py` evaluation configuration
  • Add `vllm_swe_llama_7b.py` model configuration for SWE-Llama
  • Update `registry.py` to register new components
  • Update `datasets/init.py` to export SWEBenchDataset

Features

  • Support for loading SWE-bench dataset from HuggingFace
  • Custom evaluator for SWE-bench code generation tasks
  • Compatible with vLLM inference backend

Usage

```bash
python run.py configs/eval_swebench.py
```
"

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant