diff --git a/README.md b/README.md index babf08a..be0178f 100644 --- a/README.md +++ b/README.md @@ -39,11 +39,15 @@ from transformers import BitsAndBytesConfig, HqqConfig url = "https://www.youtube.com/watch?v=di3rHkEZuUw" audio_path = download_and_convert_to_mp3(url) -quant_config = HqqConfig( - nbits=1, group_size=64, quant_zero=False, quant_scale=False, axis=0 +hqq_config = HqqConfig( + nbits=1, + group_size=64, + quant_zero=False, + quant_scale=False, + axis=0, + offload_meta=False, ) # axis=0 is used by default - bnb_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_quant_type="nf4", @@ -52,11 +56,11 @@ bnb_config = BitsAndBytesConfig( ) pipeline = SpeechToTextPipeline( - model_id="distil-whisper/distil-large-v3", quant_config=quant_config + model_id="distil-whisper/distil-large-v3", quant_config=hqq_config ) # or bnb_config transcript = pipeline( - audio_path="test.mp3", + audio_path=audio_path, chunk_length_s=30, stride_length_s=5, max_new_tokens=128, diff --git a/whisperplus/__init__.py b/whisperplus/__init__.py index 7f2fc7e..eb33887 100644 --- a/whisperplus/__init__.py +++ b/whisperplus/__init__.py @@ -7,6 +7,6 @@ from whisperplus.utils.download_utils import download_and_convert_to_mp3 from whisperplus.utils.text_utils import format_speech_to_dialogue -__version__ = '0.2.7.2.dev1' +__version__ = '0.3.0' __author__ = 'kadirnar' __license__ = 'Apache License 2.0'