|
--- |
|
library_name: transformers |
|
datasets: |
|
- mteb/tweet_sentiment_extraction |
|
base_model: |
|
- openai-community/gpt2 |
|
--- |
|
|
|
# Model Card for Model ID |
|
|
|
This model fine-tunes GPT-2 on the "Tweet Sentiment Extraction" dataset for sentiment analysis tasks. |
|
|
|
|
|
|
|
## Model Details |
|
|
|
### Model Description |
|
|
|
<!-- Provide a longer summary of what this model is. --> |
|
|
|
This model fine-tunes GPT-2 using the "Tweet Sentiment Extraction" dataset to extract sentiment-relevant portions of text. |
|
It demonstrates preprocessing, tokenization, and fine-tuning with Hugging Face libraries. |
|
|
|
|
|
|
|
## Uses |
|
|
|
|
|
|
|
### Direct Use |
|
|
|
This model can be used to analyze text for sentiment-relevant extractions directly after fine-tuning. |
|
It works as a baseline model for learning sentiment-specific features. |
|
|
|
### Downstream Use [optional] |
|
|
|
Fine-tuned for tasks that involve sentiment analysis, such as social media monitoring or customer feedback analysis. |
|
|
|
|
|
### Out-of-Scope Use |
|
|
|
Avoid using the model for real-time sentiment prediction or deployment without additional training/testing for specific use cases. |
|
|
|
|
|
## Bias, Risks, and Limitations |
|
|
|
The dataset used may not fully represent the diversity of text, leading to biases in the output. There is a risk of overfitting to the specific dataset. |
|
|
|
### Recommendations |
|
|
|
Carefully evaluate the model for biases and limitations before deploying in production environments. Consider retraining on a more diverse dataset if required. |
|
|
|
|
|
## How to Get Started with the Model |
|
|
|
Use the code below to get started with the model. |
|
|
|
```python |
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
model = AutoModelForCausalLM.from_pretrained("https://huggingface.co/Wexnflex/Tweet_Sentiment") |
|
tokenizer = AutoTokenizer.from_pretrained("https://huggingface.co/Wexnflex/Tweet_Sentiment") |
|
|
|
text = "Input your text here." |
|
inputs = tokenizer(text, return_tensors="pt") |
|
outputs = model.generate(**inputs) |
|
print(tokenizer.decode(outputs[0])) |
|
|
|
|
|
|
|
#### Training Hyperparameters |
|
|
|
Training Hyperparameters |
|
Batch size: 16 |
|
Learning rate: 2e-5 |
|
Epochs: 3 |
|
Optimizer: AdamW |
|
|
|
#### Testing Data, Factors & Metrics |
|
|
|
#### Testing Data |
|
|
|
The evaluation was performed on the test split of the "Tweet Sentiment Extraction" dataset. |
|
|
|
|
|
#### Factors |
|
|
|
Evaluation is segmented by sentiment labels (e.g., positive, negative, neutral). |
|
|
|
|
|
#### Metrics |
|
|
|
Accuracy |
|
|
|
### Results |
|
|
|
70% Accuracy |
|
#### Summary |
|
|
|
The fine-tuned model performs well for extracting sentiment-relevant text, with room for improvement in handling ambiguous cases. |
|
|
|
|
|
## Technical Specifications [optional] |
|
|
|
|
|
#### Hardware |
|
|
|
T4 GPU (Google Colab) |
|
#### Software |
|
|
|
Hugging Face Transformers Library |
|
|
|
|
