cyberagent/DeepSeek-R1-Distill-Qwen-14B-Japanese
LoRA-style vs deepseek-ai/DeepSeek-R1-Distill-Qwen-14B. 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 48 of 48 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Attention 192/336, MLP 144/144, Norms 0/97, lm_head 0/1
6 derivatives checked
- mradermacher/DeepSeek-R1-Distill-Qwen-14B-Japanese-GGUF: Quantization
- bluepen5805/DeepSeek-R1-Distill-Qwen-14B-Japanese-gguf: Same weights, converted
- tensorblock/DeepSeek-R1-Distill-Qwen-14B-Japanese-GGUF: Quantization
- tatsuyaaaaaaa/DeepSeek-R1-Distill-Qwen-14B-Japanese-gguf: Quantization
- flypg/DeepSeek-R1-Distill-Qwen-14B-Japanese-chat: LoRA-style
- EQUES/TinyDeepSeek-JP-1.5B: Different architecture