|
--- |
|
license: mit |
|
--- |
|
Model Distilbert fine-tuned with imdb reviews for sentiment analysis |
|
|
|
from transformers import DistilBertForSequenceClassification, DistilBertTokenizerFast |
|
|
|
model = DistilBertForSequenceClassification.from_pretrained("./distilbert-finetuned-imdb") |
|
tokenizer = DistilBertTokenizerFast.from_pretrained("./distilbert-finetuned-imdb") |
|
|
|
Epoch Training Loss Validation Loss Accuracy |
|
1 0.224900 0.223101 0.913040 |
|
2 0.152700 0.241866 0.929960 |
|
3 0.089300 0.280867 0.932440 |
|
|
|
TrainOutput(global_step=4689, training_loss=0.16640237623960333, metrics={'train_runtime': 2933.7121, 'train_samples_per_second': 25.565, 'train_steps_per_second': 1.598, 'total_flos': 9935054899200000.0, 'train_loss': 0.16640237623960333, 'epoch': 3.0}) |
|
|
|
{'eval_loss': 0.28086698055267334, 'eval_accuracy': 0.93244, 'eval_runtime': 246.0564, 'eval_samples_per_second': 101.603, 'eval_steps_per_second': 6.352, 'epoch': 3.0} |
