xXAI-botXx / Calm-down-bot

A GPT2 model is to be extended to a new task via transfer learning. The GPT2 model is to be fine-tuned in such a way that it generates interactive, meditative and calming stories for the user.
Other
0 stars 0 forks source link

Remove Padding #12

Open xXAI-botXx opened 6 months ago

xXAI-botXx commented 6 months ago

Try out to remove the zero padding in the training loop. Is this possible? Can the model can be trained like that?

There have to be a issue with the tensor, normally the tensors can have different sizes per batch. So you also can check the creation of the tensors.