llmixer/BigWeave-v31-96b
Partial fine-tune vs meta-llama/Meta-Llama-3-70B-Instruct. Partial change: 66% of the weights changed (attention, mlp, norms), in 55 of 80 layers.
Tensors changed: Embeddings 0/1, Norms 108/161, MLP 164/240, Attention 220/320, lm_head 0/1
1 derivatives checked
- mradermacher/BigWeave-v31-96b-i1-GGUF: Quantization
1 other repos hold byte-for-byte identical weight files.