omaressamrme/tuning

Fine-tuned DistilBERT for sentiment analysis on the IMDb dataset.

Training setup

• Base model: distilbert-base-uncased
• Dataset: IMDb (train/test)
• Epochs: 1
• Learning rate: 2e-05
• Train batch size: 16
• Eval batch size: 32
• Max train samples: 1000
• Max eval samples: 500

Evaluation (test split)

• Accuracy: 0.31
• F1 (binary): 0.0

Usage

from transformers import pipeline
clf = pipeline("text-classification", model="omaressamrme/tuning")
print(clf("I absolutely loved this movie!"))

Hugging Face Inference API

curl -H "Authorization: Bearer $HF_TOKEN"      -H "Content-Type: application/json"      -d '{"inputs": "I absolutely loved this movie!"}'      https://api-inference.huggingface.co/models/omaressamrme/tuning

Space demo

Open the Space: https://huggingface.co/spaces/omaressamrme/tuning-space

Batch inference

You can batch texts using the pipeline:

texts = ["Great film!", "Worst plot ever."]
preds = clf(texts)

Model comparison

Try comparing against another sentiment model (e.g., distilbert-base-uncased-finetuned-sst-2-english) in the Space "Compare" tab.

Intended uses & limitations

• Intended for educational/demo sentiment classification.
• Trained on a subset of IMDb for speed; performance is lower than full training.
• May reflect dataset biases; do not use for critical decisions.

Reproducibility

See the training script in the associated GitHub repo.