win10/Mistral-Nemo-Instruct-2407-20b
Partial fine-tune (fewer layers) vs unsloth/Mistral-Nemo-Instruct-2407. unsloth/Mistral-Nemo-Instruct-2407 matches part of win10/Mistral-Nemo-Instruct-2407-20b (363 of its 705 tensors; the rest of win10/Mistral-Nemo-Instruct-2407-20b has no counterpart). On the shared part: Partial change: 86% of the weights changed (attention, mlp, norms), in 40 of 40 layers.
Tensors changed: Embeddings 0/1, Norms 76/81, MLP 116/120, Attention 154/160, lm_head 0/1
2 derivatives checked
- mradermacher/Mistral-Nemo-Instruct-2407-20b-i1-GGUF: Quantization
- mradermacher/Mistral-Nemo-Instruct-2407-20b-GGUF: Quantization