Guilherme34/Firefly-V2.5
LoRA-style vs Guilherme34/Firefly-V2. 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 28 of 28 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/57, MLP 84/84, Attention 112/112
4 derivatives checked
- mradermacher/Firefly-V2.5-i1-GGUF: Quantization
- mradermacher/Firefly-V2.5-GGUF: Same weights, converted
- Guilherme34/Firefly-V3.2: LoRA-style
- Guilherme34/Firefly-V3: LoRA-style