Running training.py results in an error

[BanTheNons]

For some reason it's telling me that num_samples is 0 even though there are thousands of samples.

[horenbergerb]

Have you verified that

train_data = TextDataset(tokenizer=tokenizer, file_path=train_dir, block_size=block_size, overwrite_cache=False)

properly loaded? The wrong filepath here might cause this issue?

Difficult for me to investigate this with the information available. I'll leave this open for a bit in case the same problem comes up again (or you have new information) and then close it.