Views
No views yet
.pte export of openai/whisper-medium for CPU inference via the XNNPACK backend, with post-training quantization enabled.model.pte: ExecuTorch program (methods: encoder, text_decoder)whisper_preprocessor.pte: mel-spectrogram preprocessor (feature size 80)tokenizer.json, vocab.json, merges.txt: tokenizer artifactsconfig.json, generation_config.json, preprocessor_config.json, tokenizer_config.json, special_tokens_map.json: metadata files from the upstream Hugging Face repo--qlinear 8da4w: decoder linear layers use 8-bit dynamic activations + 4-bit weights--qlinear_encoder 8da4w: encoder linear layers use 8-bit dynamic activations + 4-bit weightsautomatic-speech-recognitionxnnpackexecutorch==1.2.0a0+efe4f0c (git efe4f0cce3)optimum-executorch==0.2.0.dev0 (git 4c62ed7)1optimum-cli export executorch \
2 --model "openai/whisper-medium" \
3 --task "automatic-speech-recognition" \
4 --recipe "xnnpack" \
5 --qlinear "8da4w" \
6 --qlinear_encoder "8da4w" \
7 --output_dir "<output_dir>"1python -m executorch.extension.audio.mel_spectrogram \
2 --feature_size 80 \
3 --stack_output \
4 --max_audio_len 300 \
5 --output_file whisper_preprocessor.ptemake whisper-cpu1cmake-out/examples/models/whisper/whisper_runner \
2 --model_path model.pte \
3 --tokenizer_path ./ \
4 --audio_path output.wav \
5 --processor_path whisper_preprocessor.pte \
6 --temperature 0