prithivMLmods/Geminorum-Wasat-14B-Instruct
Full fine-tune vs deepseek-ai/DeepSeek-R1-Distill-Qwen-14B. Fine-tune: 98% of the weights changed, across 48 of 48 layers.
Tensors changed: Embeddings 1/1, Attention 329/336, MLP 141/144, Norms 97/97, lm_head 1/1
2 derivatives checked
- mradermacher/Geminorum-Wasat-14B-Instruct-i1-GGUF: Quantization
- mradermacher/Geminorum-Wasat-14B-Instruct-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.