Downloads · 30 days
17
7% of all-time downloads
logan7000/cogrpo-heter-qwen25-3b-x-llama32-3b-math345-bs2-groupA-qwen
cogrpo-heter-qwen25-3b-x-llama32-3b-math345-bs2-groupA-qwen is a machine learning model from logan7000. Use it for the machine 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.
Cross-supervised co-GRPO, Qwen2.5-3B × Llama-3.2-3B-Instruct (bs2, EB=128, 2ep, 136 steps). Load: AutoModelForCausalLM.frompretrained(repo, torchdtype="bfloat16"). Eval temp 0.6, verifiers/qwen.
Downloads · 30 days
17
7% of all-time downloads
All-time downloads
235
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
Cross-supervised co-GRPO, Qwen2.5-3B × Llama-3.2-3B-Instruct (bs2, EB=128, 2ep, 136 steps).
| field | value |
|---|---|
| best eval_reward (MATH-500 pass@1) | 0.678 @ step 100 |
| peer | Llama-3.2-3B-Instruct |
Load: AutoModelForCausalLM.from_pretrained(repo, torch_dtype="bfloat16"). Eval temp 0.6, verifiers/qwen. |