Closed johnml1135 closed 10 months ago
The number of tokens should be truncated to 200, so the length isn't the core issue. This error is occurring, because it is running on John's RTX 3090, which only has 24gb of memory. The num_beams
should be dropped to 1. If that doesn't fix it, then the batch_size
can be decreased as well.
We can drop it to num_beams
to 1 in https://github.com/sillsdev/serval/issues/178.
num_beams was dropped to 1 and the model chosen was 600_distilled for ext-qa based upon the staging environment defaults.
https://app.sil.hosted.allegro.ai/projects/*/experiments/e0a9364cae5f4cb0b31a0237d5dc6440/info-output/log?columns=selected&columns=type&columns=name&columns=tags&columns=status&columns=project.name&columns=users&columns=started&columns=last_update&columns=last_iteration&columns=parent.name&order=-started&filter=&deep=true
Does it have something to do with the input length being very long?