rasyosef/phi-2-instruct-v0.1
LoRA-style vs microsoft/phi-2. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of dense, fc1, fc2, k_proj, q_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Norms 0/66, MLP 64/128, Attention 128/256
1 derivatives checked
- MaziyarPanahi/phi-2-instruct-v0.1-GGUF: Same weights, converted