Galen Ballew

I also ran this with `no_cuda=True` and received the same error.

Sorry about that--I've put everything into this repo if that is easier: https://github.com/galenballew/bert-multiclass I'll also repeat it here too: ```python # Dependencies import matplotlib.pyplot as plt from sklearn.metrics import accuracy_score...

`trainer.state` directly after instantiation: ``` TrainerState(epoch=None, global_step=0, max_steps=0, num_train_epochs=0, total_flos=0, log_history=[], best_metric=None, best_model_checkpoint=None, is_local_process_zero=True, is_world_process_zero=True, is_hyper_param_search=False, trial_name=None, trial_params=None) ``` Added this and am including entire output, not just the state....

Great catch! I modified `compute_metrics()` to run successfully without any warnings: ```python def compute_metrics(eval_pred): accuracy = load("accuracy") precision = load("precision") f1 = load("f1") recall = load("recall") predictions, labels = eval_pred...