ApolloRaines/DeepSeek-R1-Distill-Qwen-7B-Desyced
LoRA-style vs deepseek-ai/DeepSeek-R1-Distill-Qwen-7B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 27 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Attention 108/196, MLP 81/84, Norms 0/57, lm_head 0/1
2 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Qwen-7B-Desyced-i1-GGUF: Quantization
- mradermacher/DeepSeek-R1-Distill-Qwen-7B-Desyced-GGUF: Same weights, converted