speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

fix bug

Browse Source
This commit is contained in:
游雁 2024-06-09 03:29:41 +08:00
parent 6e0d138826
commit 1163110135
1 changed files with 2 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
funasr/datasets/openai_datasets/datasets.py
View File

@ -51,7 +51,7 @@ class OpenAIDataset(torch.utils.data.Dataset):
self.batch_size = kwargs.get("batch_size") self.batch_size = kwargs.get("batch_size")
self.batch_type = kwargs.get("batch_type") self.batch_type = kwargs.get("batch_type")
self.prompt_ids_len = 0 self.prompt_ids_len = 0
self.retry = kwargs.get("retry", 10) self.retry = kwargs.get("retry", 100)
self.permute = False self.permute = False
from funasr.frontends.whisper_frontend import WhisperFrontend from funasr.frontends.whisper_frontend import WhisperFrontend
@ -212,13 +212,9 @@ class OpenAIDataset(torch.utils.data.Dataset):
if self.batch_type != "example": if self.batch_type != "example":
b, t = outputs["input_ids"].shape b, t = outputs["input_ids"].shape
if b > 1 and b * t > self.batch_size * self.batch_size_scale_ratio_max: if b > 1 and b * t > self.batch_size * self.batch_size_scale_ratio_max:
# beg = torch.randint(0, 2, ()).item()
# if b < 2:
# beg = 0
logging.info( logging.info(
f"Warning, b*t: {b}*{t}={b * t} > batch_size*relax: {self.batch_size_scale_ratio_max}*{self.batch_size}={self.batch_size_scale_ratio_max*self.batch_size}, drop half data {idx}th, beg:{beg}" f"Warning, b*t: {b}*{t}={b * t} > batch_size*relax: {self.batch_size_scale_ratio_max}*{self.batch_size}={self.batch_size_scale_ratio_max*self.batch_size}, drop last data"
) )
# samples = samples[beg : beg + b : 2]
samples = samples[:-1] samples = samples[:-1]
continue continue