anarh1st47/AliceAI-T5-35B-A0.6B-SFT-v0.1
LoRA-style vs yandex/AliceAI-T5-35B-A0.6B. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, o_proj, q_proj, v_proj changed, in 16 of 16 layers; norms, biases and embeddings are untouched.
Tensors changed: MLP 0/168, Embeddings 0/1, Norms 0/70, Attention 160/160
1 derivatives checked
- anarh1st47/AliceAI-T5-35B-A0.6B-SFT-v0.1-MLX-4bit: Quantization