llmixer/BigWeave-v33-105b
Partial fine-tune (fewer layers) vs meta-llama/Meta-Llama-3-70B-Instruct. meta-llama/Meta-Llama-3-70B-Instruct matches part of llmixer/BigWeave-v33-105b (723 of its 1083 tensors; the rest of llmixer/BigWeave-v33-105b has no counterpart). On the shared part: Partial change: 74% of the weights changed (attention, mlp, norms), in 61 of 80 layers.
Tensors changed: Embeddings 0/1, Norms 120/161, MLP 182/240, Attention 244/320, lm_head 0/1
1 derivatives checked
- mradermacher/BigWeave-v33-105b-i1-GGUF: Quantization