Batch Size =1
Opened this issue · 6 comments
The code throws an error with batch_size = 1
IndexError: too many indices for tensor of dimension 1
Yeah, for training you require batches with PK sampling so 1 is not a valid option, for inference perhaps u need to change some line of code...
I fix the code. @videturfortuna. Could you check pull requests?
@Mohit-robo Could you check pull requests code? it's fix that error.
I fix the code. @videturfortuna. Could you check pull requests? @Mohit-robo Could you check pull requests code? it's fix that error.
Sure thanks, let me check and I'll let you know. But did you check whether the results changed or no?
@Mohit-robo
Yes, when I debugged, I confirmed that the emb had a shape of (1,2048), not (2048).
And the code has been merged. We recommend doing a git pull.
@Mohit-robo Yes, when I debugged, I confirmed that the emb had a shape of (1,2048), not (2048).
And the code has been merged. We recommend doing a git pull.
Sure thanks
Hey, thanks for the solution. Also is there any way I could use your pre-trained weights and fine-tune them on my datasets.