RuntimeError: Could not infer dtype of numpy.int16

[Qingkali]

When I attempted to use the TSC function,dls = TSDataLoaders.from_dsets(dsets.train, dsets.valid, bs=[64, 128], batch_tfms=[TSStandardize()], num_workers=0) I encountered some errors（RuntimeError: Could not infer dtype of numpy.int16）."This question has troubled me for a long time. Have you solved it

[saidineshpola]

I'm also getting this I'm tired of asking claude and came here for solution.

I have to do multi-label classification for each target in 95 labels so output is 95 length of array having elements 0,1

import numpy as np

# Explicitly convert TrainX and trainy to float32
TrainX = np.array(TrainX, dtype=np.float32)
trainy = np.array(trainy, dtype=np.float32)  # or np.int64 if it's a classification problem

# Verify the dtypes and `shapes`
print("TrainX dtype:", TrainX.dtype) # TrainX dtype: float32
print("trainy dtype:", trainy.dtype) # trainy dtype: float32
print("TrainX shape:", TrainX.shape) # TrainX shape: (31839, 1, 85922)
print("trainy shape:", trainy.shape) # trainy shape: (31839, 94)

# Create data loaders with explicit type handling
tfms = [None]  # or [TSMultiLabelClassification()] depending on your exact needs
batch_tfms = TSStandardize()
dls = get_ts_dls(TrainX, trainy, splits=splits, tfms=tfms, batch_tfms=batch_tfms, bs=[64, 128])