ibm-granite/granite-guardian-4.0-3b-toxicity-ja
LoRA-style vs ibm-granite/granite-4.0-micro. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k_proj, q_proj, v_proj changed, in 40 of 40 layers; norms, biases and embeddings are untouched.
Tensors changed: Embeddings 0/1, Norms 0/81, Attention 120/160, MLP 0/80
2 derivatives checked
- mradermacher/granite-guardian-4.0-3b-toxicity-ja-i1-GGUF: Quantization
- mradermacher/granite-guardian-4.0-3b-toxicity-ja-GGUF: Same weights, converted