knightscode139/bert-base-cased-imdb-sentiment

BERT Fine-tuned on IMDB Reviews

Fine-tuned BERT-base-cased for binary sentiment classification on movie reviews.

This model is a fine-tuned version of google-bert/bert-base-cased on Stanford IMDB dataset.

Test Results:

• Accuracy: 92.8%
• Loss: 0.2904

Model description

This model classifies movie reviews as positive or negative sentiment. Fine-tuned from google-bert/bert-base-cased on the IMDB dataset using HuggingFace Trainer.

Intended uses & limitations

Uses:

• Sentiment analysis on movie reviews
• General sentiment classification on similar review-style text

Limitations:

• Trained specifically on movie reviews - may not generalize well to other domains
• Binary classification only (positive/negative)
• Maximum sequence length: 512 tokens

Training and evaluation data

• Dataset: Stanford IMDB
• Size: 50,000 reviews total
• Split: 20,000 train / 5,000 validation / 25,000 test
• Classes: Binary (0=Negative, 1=Positive), perfectly balanced

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.247	1.0	1250	0.2467	0.907
0.1241	2.0	2500	0.3196	0.9212
0.0574	3.0	3750	0.3944	0.9178

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1