Detect and display GGUF quantization from tensor info, not from filenames, especially helpful for role aware A Hugging Face forum post argues that GGUF quantization should be detected and displayed from the file's internal tensor metadata rather than from filenames, because heterogeneous quantization assigns distinct precisions to specific tensor groups based on sensitivity instead of a single global label. The post states that parsers expecting a uniform quantization flag miss the per-tensor `ggml_type` stored in the header, and recommends iterating the `tensor_info` array, aggregating the histogram of types, and calculating the weighted average bits per parameter. The approach is presented as a fix for mixed-precision packs where experts use IQ2_XXS while attention layers require Q8_0, and to avoid empty compatibility panels. @BoxyFit https://discuss.huggingface.co/u/boxyfit The current reliance on filenames fails because heterogeneous quantization assigns distinct precisions to specific tensor groups based on sensitivity, not a single global label. When parsers expect a uniform quantization flag, they miss the per-tensor ggml type stored in the header. A robust implementation must iterate the tensor info array, aggregate the histogram of types, and calculate the weighted average bits per parameter. This approach correctly handles mixed-precision packs like the ones you described, where experts might use IQ2 XXS while attention layers require Q8 0. The parser logic needs to prioritize the internal tensor metadata over external naming conventions to avoid empty compatibility panels.