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 03:36:23 +08:00
parent 1163110135
commit 373fbd15e0
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
funasr/datasets/openai_datasets/datasets.py
View File

@ -63,6 +63,7 @@ class OpenAIDataset(torch.utils.data.Dataset):
# self.kwargs = kwargs
self.max_token_length = kwargs.get("max_token_length", 1024)
self.batch_size_scale_ratio_max = kwargs.get("batch_size_scale_ratio_max", 1.5)
self.batch_size_token_max = kwargs.get("batch_size_token_max", 2500)
def get_source_len(self, index):
item = self.index_ds[index]
@ -213,7 +214,7 @@ class OpenAIDataset(torch.utils.data.Dataset):
b, t = outputs["input_ids"].shape
if b > 1 and b * t > self.batch_size * self.batch_size_scale_ratio_max:
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 last data"
f"Warning, {idx}th, b*t: {b}*{t}={b * t} > batch_size_sample_max: {self.batch_size_token_max}, drop last data"
)
samples = samples[:-1]
continue