nanash66/typhoon-ocr1.5-2b-ROCMFP4-GGUF
Quantization (fewer layers) vs typhoon-ai/typhoon-ocr1.5-2b. nanash66/typhoon-ocr1.5-2b-ROCMFP4-GGUF matches part of typhoon-ai/typhoon-ocr1.5-2b (310 of its 625 tensors; the rest of typhoon-ai/typhoon-ocr1.5-2b has no counterpart). On the shared part: nanash66/typhoon-ocr1.5-2b-ROCMFP4-GGUF is a GGUF quantization (T100/Q6_K) of the same weights: 113 tensors match exactly and 99 within quantization error.
Tensors changed: Embeddings 0/1, Norms 0/113, MLP 0/42, Attention 0/56