tipi/kyutai
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix stt_from_file_pytorch (#39)

Browse Source 
1. argparse declares in_file, but code reads file
2. text_tokens.numpy().tolist() is a list of list of list of int
instead of the supported list of list of int.
this is a debugging print just drop it

Co-authored-by: Pierre-Hugues Husson <phhusson@freebox.fr>
This commit is contained in:
Pierre-Hugues HUSSON 2025-07-03 15:26:34 +02:00 committed by GitHub
parent e9bac066ea
commit 3a4165a84f
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
scripts/stt_from_file_pytorch.py
View File

@ -150,7 +150,7 @@ def main(args):
audio_delay_seconds = info.stt_config.get("audio_delay_seconds", 5.0) audio_delay_seconds = info.stt_config.get("audio_delay_seconds", 5.0)
padding_token_id = info.raw_config.get("text_padding_token_id", 3) padding_token_id = info.raw_config.get("text_padding_token_id", 3)
audio, input_sample_rate = sphn.read(args.file) audio, input_sample_rate = sphn.read(args.in_file)
audio = torch.from_numpy(audio).to(args.device) audio = torch.from_numpy(audio).to(args.device)
audio = julius.resample_frac(audio, input_sample_rate, mimi.sample_rate) audio = julius.resample_frac(audio, input_sample_rate, mimi.sample_rate)
if audio.shape[-1] % mimi.frame_size != 0: if audio.shape[-1] % mimi.frame_size != 0:
@ -178,8 +178,6 @@ def main(args):
if text_tokens is not None: if text_tokens is not None:
text_tokens_accum.append(text_tokens) text_tokens_accum.append(text_tokens)
print(tokenizer.decode(text_tokens.numpy().tolist()))
utterance_tokens = torch.concat(text_tokens_accum, dim=-1) utterance_tokens = torch.concat(text_tokens_accum, dim=-1)
timed_text = tokens_to_timestamped_text( timed_text = tokens_to_timestamped_text(
utterance_tokens, utterance_tokens,