Delta-Vector/Austral-70B-Winton
LoRA-style vs Delta-Vector/Austral-70B-Preview. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj changed, in 80 of 80 layers; norms, biases and embeddings are untouched.
Tensors changed: lm_head 0/1, Norms 0/161, Embeddings 0/1, MLP 240/240, Attention 320/320
4 derivatives checked
- bartowski/Delta-Vector_Austral-70B-Winton-GGUF: Quantization
- mradermacher/Austral-70B-Winton-i1-GGUF: Quantization
- mradermacher/Austral-70B-Winton-GGUF: Quantization
- Delta-Vector/Plesio-70B: Full fine-tune