deepseek-ai/DeepSeek-Prover-V1.5-RL

Partial fine-tune vs deepseek-ai/DeepSeek-Prover-V1.5-SFT. Partial change: 88% of the weights changed (attention, mlp, norms), in 30 of 30 layers.

Tensors changed: Embeddings 0/1, Norms 45/61, lm_head 0/1, Attention 120/120, MLP 90/90

1 derivatives checked

• mradermacher/DeepSeek-Prover-V1.5-RL-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse