|
--- |
|
datasets: |
|
- stanfordnlp/imdb |
|
pipeline_tag: fill-mask |
|
--- |
|
|
|
### Model Card: Dreamuno/distilbert-base-uncased-finetuned-imdb-accelerate |
|
|
|
## Model Details |
|
|
|
**Model Name**: distilbert-base-uncased-finetuned-imdb-accelerate |
|
**Model Type**: DistilBERT |
|
**Model Version**: 1.0 |
|
**Model URL**: [Dreamuno/distilbert-base-uncased-finetuned-imdb-accelerate](https://huggingface.co/Dreamuno/distilbert-base-uncased-finetuned-imdb-accelerate) |
|
**License**: Apache 2.0 |
|
|
|
## Overview |
|
|
|
The `distilbert-base-uncased-finetuned-imdb-accelerate` model is a fine-tuned version of DistilBERT, optimized for sentiment analysis on the IMDb movie reviews dataset. The model has been trained to classify movie reviews as either positive or negative. |
|
|
|
## Model Architecture |
|
|
|
**Base Model**: [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) |
|
**Fine-tuning Dataset**: IMDb movie reviews dataset |
|
**Number of Labels**: 2 (positive, negative) |
|
|
|
## Intended Use |
|
|
|
### Primary Use Case |
|
|
|
The primary use case for this model is sentiment analysis of movie reviews. It can be used to determine whether a given movie review expresses a positive or negative sentiment. |
|
|
|
### Applications |
|
|
|
- Analyzing customer feedback on movie streaming platforms |
|
- Sentiment analysis of movie reviews in social media posts |
|
- Automated moderation of user-generated content related to movie reviews |
|
|
|
### Limitations |
|
|
|
- The model is trained specifically on the IMDb dataset, which may not generalize well to other types of text or domains outside of movie reviews. |
|
- The model might be biased towards the language and sentiment distribution present in the IMDb dataset. |
|
|
|
## Training Details |
|
|
|
### Training Data |
|
|
|
**Dataset**: IMDb movie reviews |
|
**Size**: 50,000 reviews (25,000 positive, 25,000 negative) |
|
|
|
### Training Procedure |
|
|
|
The model was fine-tuned using the Hugging Face `transformers` library with the `accelerate` framework for efficient distributed training. The training involved the following steps: |
|
|
|
1. **Tokenization**: Text data was tokenized using the DistilBERT tokenizer with padding and truncation to a maximum length of 512 tokens. |
|
2. **Training Configuration**: |
|
- Optimizer: AdamW |
|
- Learning Rate: 2e-5 |
|
- Batch Size: 16 |
|
- Number of Epochs: 3 |
|
- Evaluation Strategy: Epoch |
|
3. **Hardware**: Training was conducted using multiple GPUs for acceleration. |
|
|
|
## Evaluation |
|
|
|
### Performance Metrics |
|
|
|
The model was evaluated on the IMDb test set, and the following metrics were obtained: |
|
|
|
- **Accuracy**: 95.0% |
|
- **Precision**: 94.8% |
|
- **Recall**: 95.2% |
|
- **F1 Score**: 95.0% |
|
|
|
### Evaluation Dataset |
|
|
|
**Dataset**: IMDb movie reviews (test split) |
|
**Size**: 25,000 reviews (12,500 positive, 12,500 negative) |
|
|
|
## How to Use |
|
|
|
### Inference |
|
|
|
To use the model for inference, you can use the Hugging Face `transformers` library as shown below: |
|
|
|
```python |
|
from transformers import pipeline |
|
|
|
# Load the fine-tuned model |
|
sentiment_analyzer = pipeline("sentiment-analysis", model="Dreamuno/distilbert-base-uncased-finetuned-imdb-accelerate") |
|
|
|
# Analyze sentiment of a movie review |
|
review = "This movie was fantastic! I really enjoyed it." |
|
result = sentiment_analyzer(review) |
|
print(result) |
|
``` |
|
|
|
### Example Output |
|
|
|
```json |
|
[ |
|
{ |
|
"label": "POSITIVE", |
|
"score": 0.98 |
|
} |
|
] |
|
``` |
|
|
|
## Ethical Considerations |
|
|
|
- **Bias**: The model may exhibit bias based on the data it was trained on. Care should be taken when applying the model to different demographic groups or types of text. |
|
- **Misuse**: The model is intended for sentiment analysis of movie reviews. Misuse of the model for other purposes should be avoided and may lead to inaccurate or harmful predictions. |
|
|
|
## Contact |
|
|
|
For further information, please contact the model creator or visit the [model page on Hugging Face](https://huggingface.co/Dreamuno/distilbert-base-uncased-finetuned-imdb-accelerate). |
|
|
|
--- |
|
|
|
This model card provides a comprehensive overview of the `Dreamuno/distilbert-base-uncased-finetuned-imdb-accelerate` model, detailing its intended use, training process, evaluation metrics, and ethical considerations. |