Views
No views yet
TL;DR: Please consider using my FP8 finetune. In my preliminary tests and benchmarks it performs similar to this FP16 variant. However, it's 2x as fast and 2x less memory hungry!
unsloth/CrisperWhisper (Torch) → converted via mlx-examples/whisper/convert.py. :contentReference[oaicite:0]{index=0}weights.safetensors — MLX FP16 weightsconfig.json — MLX Whisper configmlx_whisper supports Hugging Face repo IDs in path_or_hf_repo, and will download automatically. :contentReference[oaicite:1]{index=1}1from mlx_whisper import transcribe
2
3out = transcribe(
4 "audio.wav",
5 path_or_hf_repo="kyr0/crisperwhisper-unsloth-mlx",
6)
7print(out["text"])path_or_hf_repo to it:1from mlx_whisper import transcribe
2
3out = transcribe(
4 "audio.wav",
5 path_or_hf_repo="./mlx_models/crisperwhisper-unsloth-mlx",
6)
7print(out["text"])