Commit History
-
Correct long-form generation config parameters 'max_initial_timestamp_index' and 'prev_sot_token_id'. (#24)
Verified signature- Correct long-form generation config parameters 'max_initial_timestamp_index' and 'prev_sot_token_id'. (8530d023dd58e06e31b76367de039e1a51641eea)87c7102
-
Adding `safetensors` variant of this model (#21)
- Adding `safetensors` variant of this model (acfaaeb1e51111b2b750ab344c09324387b697a8) Co-authored-by: Safetensors convertbot <SFconvertbot@users.noreply.huggingface.co>60c8e22
-
Update generation config with word-level alignment heads (#16)
- Upload config (cd2929be80269e49ec00fc3a3f71892c7ec327d3)b70535d
-
Update README.md
944489f
-
Update generation_config.json to suppress task tokens (#14)
- Update generation_config.json to suppress task tokens (34b747b4eb36f4bec21f081ccbf51d08f7271355) Co-authored-by: Guillaume Klein <guillaumekln@users.noreply.huggingface.co>329f3b1 -
Update config.json to suppress task tokens (#13)
- Update config.json to suppress task tokens (53dcee3b3e820455addd16faaf75cc4dde0cbeb3) Co-authored-by: Guillaume Klein <guillaumekln@users.noreply.huggingface.co>a63c560
-
Add Flax weights
sanchit-gandhi1f7ce4a
-
Update README.md
74d3cae
-
Update README.md (#11)
- Update README.md (7e03d0f9bd5a90404b5cb0fb3843d8b3952c4b9c)dcbf596
-
Changes for fast tokenizer (#10)
- Add tokenizer.json (124eda19723059d73b3e24d22089ff0266a5c981) Co-authored-by: Jonatan Kłosko <jonatanklosko@users.noreply.huggingface.co>57ea019
-
Update generation_config.json
d964250 -
Update generation_config.json
7dd9e1c -
Update generation_config.json
cad221b -
Upload config
56db7ee
-
Revert: Add "<|startoftranscript|>" to forced decoder ids (#6)
- Revert: Add "<|startoftranscript|>" to forced decoder ids (db578b003807f8d58a7cd90e56b5012bf26dac42)23d5e63
-
Update config.json
e5ed3f5 -
Force <|startoftranscript|>
Updates the `forced_decoder_ids` to force the `<|startoftranscript|>` token at position 1. This is to match the official Whisper implementation, which always predicts `<|startoftranscript|>` at position 1: ```python #!pip install git+https://github.com/openai/whisper.git import whisper from datasets import load_dataset import torch device = "cuda" if torch.cuda.is_available() else "cpu" model = whisper.load_model("tiny.en").to(device) tokenizer = whisper.tokenizer.get_tokenizer(False, task="transcribe", language="en") tokenizer = tokenizer.tokenizer librispeech = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation") def to_pad_to_mel(array): """Static function which: 1. Pads/trims a list of audio arrays to a max length of 30s 2. Computes log-mel filter coefficients from padded/trimmed audio sequences Inputs: array: list of audio arrays Returns: input_ids: torch.tensor of log-mel filter b…eb2a3e3
-
Update README.md (#3)
- Update README.md (933dd954d44445fa823390fd4ceeeed6049bcc35) Co-authored-by: Luigi Liu <Luigi@users.noreply.huggingface.co>f3b5e97 -
Allow single quotes "'" and hyphens "-" (#4)
- Allow single quotes "'" and hyphens "-" (54f836ce498930b225ff2a6db8c2a40bddece9f8) Co-authored-by: Sanchit Gandhi <sanchit-gandhi@users.noreply.huggingface.co>b8d7ffd
-
Update `max_length` param
407a3a8 -
Update `return_mask` param
ea8f5d0
-
Update README.md
2ac54a8 -
Create README.md
9e20f92 -
Add TF weights (#1)
- Add TF weights (7bed60b379d9f0c588b77137f2af66d0f7b09ed8)1f9fe7b
-
Upload config
215213d -
Upload tokenizer
c703e33 -
Upload config
f11df2a -
Upload tokenizer
2b05ff7 -
Upload tokenizer
3b1e4bb -
Update config.json
9503433 -
Upload tokenizer
18d92b3 -
Upload tokenizer
c6a40e3
-
Upload processor
7a93d85 -
Upload config
7dabb1d -
Upload processor
8710e8b -
Upload config
47d9e19 -
Upload processor
33560bf -
Create new file
b961dda -
Upload processor
10c9736 -
Upload config
d13d0e4
-
Upload tokenizer
80bf224 -
Upload WhisperForConditionalGeneration
c83d640 -
Upload processor
f74ab99 -
Upload config
2ff8162
-
Upload processor
40083c0 -
Upload model
09fa5b9 -
initial commit
36ced72