mradermacher/granite-4.0-h-3b-ar-GGUF
Same weights, converted vs ibm-research/granite-4.0-h-3b-ar. Same weights stored in another dtype (BF16 → F16/F32), plus 156 new tensors.
Tensors changed: Embeddings 0/1, Norms 0/81, Attention 0/16, lm_head 0/1
Same weights, converted vs ibm-research/granite-4.0-h-3b-ar. Same weights stored in another dtype (BF16 → F16/F32), plus 156 new tensors.
Tensors changed: Embeddings 0/1, Norms 0/81, Attention 0/16, lm_head 0/1