15 run inference on baskerville#30

Merged

J-Dymond merged 14 commits intomainfrom 15-run-inference-on-baskerville

Dec 6, 2024

+347-116

Commits on Nov 29, 2024

Commits on Dec 2, 2024

changes to the forward methods so that logits are not stored outside of the custom pipeline method
J-Dymond
committed
adding minor inference argument changes
tbc
committed

Commits on Dec 3, 2024

added a drop length to the dataloader, this will prevent overly-long inputs from being used during inference
tbc
committed
Added a drop_length parameter, if passed it specifies a maximum character length of the first language passed to the dataset loader.
tbc
committed
dropped zero-one accuracy and added other metrics to the inference results output
J-Dymond
committed
updated gen_configs script to create experiments in slurm_scripts folder, now save multiple runs under subdirectories named via the seed
J-Dymond
committed
moving prototype slurm scripts to the experiments subdirectory
tbc
committed
made small change to the inference pipeline, which allow the experiment_result--seed combination directory to exist already, since this is caught in the jobscript creation I don't forsee this being…
tbc
committed
removed another test bash script
J-Dymond
committed
updated README.md in the /scripts folder
J-Dymond
committed

Commits on Dec 6, 2024

addressed comments from pull request, also added additional outputs to translation to allow additional confidence measures
J-Dymond
committed