Amu/orpo-lora-phi2
LoRA-style vs microsoft/phi-2. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of fc1, fc2, q_proj, v_proj changed, in 32 of 32 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/66, MLP 64/128, Attention 64/256, lm_head 0/2
2 derivatives checked
- mradermacher/orpo-lora-phi2-i1-GGUF: Quantization
- mradermacher/orpo-lora-phi2-GGUF: Same weights, converted