tngtech/DeepSeek-R1T-Chimera
Partial fine-tune vs deepseek-ai/DeepSeek-R1. Partial change: 66% of the weights changed (attention, embeddings, lm_head, mlp, norms, other), in 62 of 62 layers.
Tensors changed: Embeddings 2/2, Attention 620/620, Norms 241/252, MLP 1302/2123, lm_head 1/1, other 2/2
1 derivatives checked
- DevQuasar/tngtech.DeepSeek-R1T-Chimera-GGUF: Quantization