Downloads · 30 days
9
16% of all-time downloads
Reviewer-32B
AGENTDARS/Reviewer-32B
Reviewer-32B is a machine learning model from AGENTDARS. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft.
Reviewer-32B is a fine-tuned on DeepSeek-R1-Distill-Qwen-32B, optimized for selecting the best patch among multiple patches generated by our DARS agent while solving software engineering problems.
All-time downloads
56
Public
Repo size
120 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt103 MB · 81%
At a glance
- Library
- peft
- Access
- Public
- Created
- Feb 24, 2025
- Updated
- Feb 24, 2025
- SHA
- 0682194d
Base models
Model card
From the Hugging Face model README
Model Card for Reviewer-32B
Model Details
Model Description
Reviewer-32B is a fine-tuned on DeepSeek-R1-Distill-Qwen-32B, optimized for selecting the best patch among multiple patches generated by our DARS agent while solving software engineering problems.
Model Sources
- Repository: DARS-32B Repository
- Paper: "DARS: Dynamic Action Re-Sampling to Enhance Coding Agent Performance by Adaptive Tree Traversal"
How to Get Started with the Model
We use vLLM to deploy and infer the model. Please follow this tutorial here to use our LoRA weights with vLLM.
Training Details
Dataset
We use our code review dataset where each instance contains several git patches with critiques for each each patch. The model learns to generate critiques for multiple patches and select the best patch.
Training Procedure
| Hyperparameter | Value |
|---|---|
| Training regime | BF16 mixed precision |
| Optimizer | AdamW with cosine learning rate scheduler |
| LoRA Configuration | rank=8, alpha=32, dropout=0.1 |
| Batch Size | 32 |
| Learning Rate | 1e-6 |
| Sequence Length | 14K tokens |
| Fine-tuning Epochs | 1 |
| Compute Environment | DeepSpeed for memory-efficient distributed training |
| Compute Infrastructure | 8x H100 |
We use training script provided in Qwen-2.5 codebase.
Results
Using this model as a reviewer with DARS trajectories generated using Claude 3.5 Sonnet V2 achieves 42.0% on SWE-Bench Lite.
- Library
- peft
- Created
- Feb 24, 2025
- Updated
- Feb 24, 2025