fine-tuned-bert-yelp

Fine-tuned BERT on Yelp Reviews (5-class classification)

This model is a BERT-base-uncased fine-tuned on the Yelp Review Full dataset.
The task is 5-class sentiment classification (1 to 5 stars).

Training Details

• Framework: Hugging Face Transformers + Ray Train
• Hardware: 3 GPU worker with Ray
• Model: bert-base-uncased
• Dataset subset: 20,000 training samples, 5,000 validation samples
• Epochs: 10
• Batch size: 16 (train), 32 (eval)
• Optimizer: AdamW (lr=2e-5, weight decay=0.01)
• Mixed precision: FP16 enabled

Evaluation Results

On the validation split:

• Accuracy: 61.9%
• F1 (weighted): 0.62
• Precision: 0.62
• Recall: 0.62
• Eval loss: 2.84

Usage

from transformers import BertTokenizer, BertForSequenceClassification

model = BertForSequenceClassification.from_pretrained("AdhamEhab/fine-tuned-bert-yelp")
tokenizer = BertTokenizer.from_pretrained("AdhamEhab/fine-tuned-bert-yelp")

text = "The food was amazing and the service was excellent!"
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)
pred = outputs.logits.argmax(dim=-1).item()
print("Predicted star rating:", pred + 1)  # labels are 0-4 -> map to 1-5