Fulcrum-AI/Ryze-Mistral-7B-Instruct-v0.3-10
Partial fine-tune (fewer layers) vs mistralai/Mistral-7B-Instruct-v0.3. mistralai/Mistral-7B-Instruct-v0.3 matches part of Fulcrum-AI/Ryze-Mistral-7B-Instruct-v0.3-10 (291 of its 669 tensors; the rest of Fulcrum-AI/Ryze-Mistral-7B-Instruct-v0.3-10 has no counterpart). On the shared part: Partial change: 81% of the weights changed (attention, mlp, norms), in 27 of 32 layers.
Tensors changed: Embeddings 0/1, Norms 54/65, MLP 81/96, Attention 108/128, lm_head 0/1
2 derivatives checked
- mradermacher/Ryze-Mistral-7B-Instruct-v0.3-10-i1-GGUF: Quantization
- mradermacher/Ryze-Mistral-7B-Instruct-v0.3-10-GGUF: Quantization