open-r1/OpenR1-Distill-7B

Full fine-tune vs open-r1/Qwen2.5-Math-7B-RoPE-300k. Fine-tune: 93% of the weights changed, across 28 of 28 layers.

Tensors changed: Embeddings 0/1, Norms 57/57, MLP 84/84, Attention 196/196, lm_head 1/1

1 derivatives checked

• csabakecskemeti/open-r1.OpenR1-Distill-7B-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse