Reproducibility steps
#1
by KevinKibe - opened
Could you share reproducibility steps for converting an omni model(already in hf format) into gguf with with fixed positional conv weight normalization?
I'm looking to benchmark performance of different quantization types?
just see our converters in https://github.com/CrispStrobe/CrispASR/tree/main/models plus crispasr-quantize script
Thanks mate
cstr changed discussion status to closed