mkurman/Qwen2.5-14B-DeepSeek-R1-1M
Full fine-tune vs Qwen/Qwen2.5-14B-Instruct-1M. Fine-tune: 100% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Norms 97/97, MLP 144/144, Attention 336/336, lm_head 1/1
3 derivatives checked
- mradermacher/Qwen2.5-14B-DeepSeek-R1-1M-i1-GGUF: Quantization
- DevQuasar-7/mkurman.Qwen2.5-14B-DeepSeek-R1-1M-GGUF: Same weights, converted
- mradermacher/Qwen2.5-14B-DeepSeek-R1-1M-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.