distilbert-sentiment
This model is a fine-tuned version of distilbert-base-uncased on a subset of the amazon-polarity dataset.
<b>[Update 10/10/23]</b> The model has been retrained on a larger part of the dataset with an improvement on the loss, f1 score and accuracy. It achieves the following results on the evaluation set:
- Loss: 0.116
- Accuracy: 0.961
- F1_score: 0.960
Model description
This sentiment classifier has been trained on 360_000 samples for the training set, 40_000 samples for the validation set and 40_000 samples for the test set.
Intended uses & limitations
from transformers import pipeline
# Create the pipeline
sentiment_classifier = pipeline('text-classification', model='AdamCodd/distilbert-base-uncased-finetuned-sentiment-amazon')
# Now you can use the pipeline to classify emotions
result = sentiment_classifier("This product doesn't fit me at all.")
print(result)
#[{'label': 'negative', 'score': 0.9994848966598511}]
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 1270
- optimizer: AdamW with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 150
- num_epochs: 2
- weight_decay: 0.01
Training results
(Previous results before retraining from the model evaluator)
key | value |
---|---|
eval_accuracy | 0.94112 |
eval_auc | 0.9849 |
eval_f1_score | 0.9417 |
eval_precision | 0.9321 |
eval_recall | 0.95149 |
Framework versions
- Transformers 4.34.0
- Pytorch lightning 2.0.9
- Tokenizers 0.14.0
If you want to support me, you can here.