|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
base_model: |
|
|
- huawei-noah/TinyBERT_General_4L_312D |
|
|
pipeline_tag: text-classification |
|
|
tags: |
|
|
- sentiment-analysis |
|
|
- tinybert |
|
|
- transformers |
|
|
- text-classification |
|
|
- imdb |
|
|
|
|
|
--- |
|
|
|
|
|
# |
|
|
|
|
|
# 📦 TinyBERT IMDB Sentiment Analysis Model |
|
|
|
|
|
This is a fine-tuned [TinyBERT](https://huggingface.co/huawei-noah/TinyBERT_General_4L_312D) model for binary **sentiment classification** on a 5,000-sample subset of the [IMDB dataset](https://huggingface.co/datasets/imdb). |
|
|
It predicts whether a movie review is **positive** or **negative**. |
|
|
|
|
|
## 🧠 Model Details |
|
|
|
|
|
- **Base model:** [`huawei-noah/TinyBERT_General_4L_312D`](https://huggingface.co/huawei-noah/TinyBERT_General_4L_312D) |
|
|
- **Task:** Sentiment Classification (Binary) |
|
|
- **Dataset:** 4,000 training + 1,000 test samples from IMDB |
|
|
- **Tokenizer:** `AutoTokenizer.from_pretrained('huawei-noah/TinyBERT_General_4L_312D')` |
|
|
- **Max length:** 300 tokens |
|
|
- **Batch size:** 64 |
|
|
- **Training framework:** Hugging Face `Trainer` |
|
|
- **Device:** A100 GPU |
|
|
|
|
|
## 📊 Evaluation Metrics |
|
|
## 📊 Evaluation Metrics (on 1,000-sample test set) |
|
|
|
|
|
| Metric | Value | |
|
|
|-----------------------|----------| |
|
|
| Accuracy | **88.02%** | |
|
|
| Evaluation Loss | 0.2962 | |
|
|
| Runtime | 30.9 sec | |
|
|
| Samples per Second | 485 | |
|
|
|
|
|
|
|
|
## 🚀 How to Use |
|
|
|
|
|
```python |
|
|
from transformers import pipeline |
|
|
|
|
|
classifier = pipeline( |
|
|
"text-classification", |
|
|
model="Harsha901/tinybert-imdb-sentiment-analysis-model" |
|
|
) |
|
|
|
|
|
result = classifier("This movie was absolutely amazing!") |
|
|
print(result) # [{'label': 'LABEL_1', 'score': 0.98}] |
