Guilherme34/Firefly-v6-DONTDOWNLOAD-STILLONPROGRESS
LoRA-style vs Guilherme34/Firefly-v5. 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 35 of 35 layers; norms, biases and embeddings are untouched.
Tensors changed: MLP 105/585, Norms 0/433, other 0/240, Attention 100/687, Embeddings 0/6
2 derivatives checked
- mradermacher/Firefly-v6-DONTDOWNLOAD-STILLONPROGRESS-i1-GGUF: Quantization (fewer layers)
- mradermacher/Firefly-v6-DONTDOWNLOAD-STILLONPROGRESS-GGUF: Same weights, converted (fewer layers)