---
license: mit
datasets:
  - imdb
language:
  - en
tags:
  - sentiment-analysis
  - lstm
  - movie-reviews
  - pytorch
model-index:
  - name: IMDB-BiLSTM-Sentiment
    results:
      - task:
          type: sentiment-analysis
          name: Binary Sentiment Classification
        dataset:
          type: imdb
          name: IMDB Large Movie Review Dataset
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.8651
          - name: F1
            type: f1
            value: 0.8635
          - name: AUC
            type: auc
            value: 0.9299
---

# 🎬 IMDB Bi-LSTM Sentiment Classifier

Predicts whether a movie review is **positive (👍)** or **negative (👎)** using a compact bidirectional LSTM built in PyTorch.

| Detail | Value |
| ------ | ----- |
| Sequence length cap | 500 tokens |
| Vocabulary size | 5 000 word stems |
| Embedding dim | 256 |
| LSTM | 2 layers · 256 hidden · *bidirectional* |
| Parameters | ~6 M |
| Validation accuracy | **0.8651** |
| Test AUC | **0.9299** |

## Quick start

```python
# pip install huggingface_hub torch nltk

from huggingface_hub import hf_hub_download
from inference import predict   # ships with the repo files

hf_hub_download("ecroatt/imdb-bilstm-sentiment", "pytorch_model.bin")

print(predict("Terrific cast and a heart-warming story!"))
# 0.96  ->  positive

print(predict("I was bored out of my mind; worst sequel ever."))
# 0.04  ->  negative
```

[![Gradio demo](https://huggingface.co/badges/space-gradio.svg)](https://huggingface.co/spaces/ecroatt/imdb-bilstm-demo)