DynamicItemDataset changes fields to string #3027
Description
Describe the bug
Below, the column types are being changed to string, by DynamicItemDataset:
import speechbrain as sb
import pandas as pd
df=pd.DataFrame(columns=['ID', 'duration', 'rand'])
df.loc[0]=[1,1.3, 1.3]
df.loc[1]=[2, 2.1, 3.2]
df.to_csv('test_file.csv', index=False)
train_data = sb.dataio.dataset.DynamicItemDataset.from_csv(
csv_path="test_file.csv",
replacements={"data_root": "./"},
)
train_data - Output ->
The column 'rand' has been changed to type 'string'. The data type of column 'duration' has been preserved.
train_data.data
{'1.0': {'duration': 1.3, 'rand': '1.3'},
'2.0': {'duration': 2.1, 'rand': '3.2'}}
Expected behaviour
train_data.data
{'1.0': {'duration': 1.3, 'rand': 1.3},
'2.0': {'duration': 2.1, 'rand': 3.2}}
To Reproduce
No response
Environment Details
No response
Relevant Log Output
Additional Context
No response