evaluate error

[1024AILab]

After the training is done, test reports an error. Also I can't use accelerate launch --config_file . /gpu_accelerate.yaml run_experiment.py command, I use python run_experiment.py instead for launch.

[srperera]

hi yes we left that function incomplete because we realized it was easier to manually test the models performance in a jupyter notebook than building it into the trainer. This way we can easily visualize things as needed.