Downloads · 30 days
4
5% of all-time downloads
logan7000/cogrpo-n3-union-qwen25-3b-x-llama32-3b-x-qwen3-1p7b-math345-groupA-qwen25-best
cogrpo-n3-union-qwen25-3b-x-llama32-3b-x-qwen3-1p7b-math345-groupA-qwen25-best is a reinforcement learning model from logan7000. Use it for the reinforcement learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Fine-tuned from Qwen/Qwen2.5-3B with co-GRPO-DP N=3, union: every distinct peer answer is a candidate; a rollout scores by matching any of them. - Co-trained with: Qwen2.5-3B x Llama-3.2-3B x Qwen3-1.7B-Base (this rep…
Downloads · 30 days
4
5% of all-time downloads
All-time downloads
79
Public
Parameters
242K
6.2 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors6.2 GB · 100%
From the Hugging Face model README
Fine-tuned from Qwen/Qwen2.5-3B with co-GRPO-DP N=3, union: every distinct peer answer is a candidate; a rollout scores by matching any of them.
Co-trained with: Qwen2.5-3B x Llama-3.2-3B x Qwen3-1.7B-Base (this repo is group A)
Training data: q1716523669/MATH-Level345 (8,740 problems, MATH levels 3-5)
Eval: MATH-500 (500 prompts, pass@1, temperature 0.6)
This checkpoint: step 90 of 136 (2 epochs) — the best one
Its score: 0.6660
best is the step with the highest eval; end is step 136. Both are published
because they answer different questions and the gap between them is a result in
itself. The sibling repo is cogrpo-n3-union-qwen25-3b-x-llama32-3b-x-qwen3-1p7b-math345-groupA-qwen25-end.
Effective batch 128 prompts per optimizer step, lr 3e-6, 2 epochs, 12
generations per prompt, 3072-token completions, temperature 1.0 for rollouts and
0.6 for eval, beta 0, bnpo loss, scale_rewards group, seed 42.
| step | MATH-500 pass@1 |
|---|---|
| 0 | 0.5480 |
| 10 | 0.5900 |
| 20 | 0.6560 |
| 30 | 0.6220 |
| 40 | 0.6360 |
| 50 | 0.6420 |
| 60 | 0.6440 |
| 70 | 0.6520 |
| 80 | 0.6240 |
| 90 | 0.6660 |
| 100 | 0.6540 |
| 110 | 0.6520 |
| 120 | 0.6640 |
| 130 | 0.6620 |
train.log — the complete training log this checkpoint came fromeval_curve.csv — the table above, machine-readablerun_config.json — resolved config as the trainer saw itCode: williamium3000/trl-projects,
branch n3-interaction-modes, projects/co-grpo-dp/.