Model Card for emotion-distilbert-7class

Model Details

Model Description

This is a fine-tuned Hugging Face transformers model for emotion classification. The model is based on DistilBERT and trained on a synthetic dataset containing 7 emotions.

• Shared by: fakhrulfaiz201
• Model type: DistilBertForSequenceClassification
• Language(s) (NLP): English
• License: Apache 2.0 (inherited from DistilBERT base model)
• Finetuned from model: distilbert-base-uncased

Model Sources

• Repository: https://huggingface.co/fakhrulfaiz201/emotion-distilbert-7class

Uses

Direct Use

This model can be used for classifying text into one of seven emotions: Love, Sad, Anger, Fun, Hate, Surprise, Happiness.

Out-of-Scope Use

The model was trained on a synthetic dataset. Its performance on real-world, nuanced, or informal text might vary. It should not be used for critical applications without further validation.

Bias, Risks, and Limitations

The model's performance is limited by the synthetic nature of its training data. It might not generalize well to diverse linguistic styles, cultural contexts, or sarcasm/irony.

Recommendations

Users should test the model thoroughly on their specific use cases and consider fine-tuning on more diverse and real-world datasets if necessary.

How to Get Started with the Model

Use the code below to get started with the model.

from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

model_name = "fakhrulfaiz201/emotion-distilbert-7class"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

id2label = {
    0: 'Love', 1: 'Sad', 2: 'Anger', 3: 'Fun',
    4: 'Hate', 5: 'Surprise', 6: 'Happiness'
}

def predict(text):
    device = next(model.parameters()).device
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True).to(device)
    with torch.no_grad():
        outputs = model(**inputs)
    logits = outputs.logits
    predicted_class_id = logits.argmax().item()
    return id2label[predicted_class_id]

print(predict("I feel amazing today!"))

Training Details

Training Data

The model was trained on the "synthetic_emotions.csv" dataset (available on Kaggle as prashanthan24/synthetic-emotions-dataset-14k-texts-7-emotions). This dataset contains 13,970 text samples labeled with one of seven emotions: Love, Sad, Anger, Fun, Hate, Surprise, Happiness. The dataset was split into training (60%), validation (20%), and test (20%) sets.

Training Procedure

Preprocessing

The text data was tokenized using DistilBertTokenizer. Input sequences were truncated and padded to a maximum length of 128 tokens.

Training Hyperparameters

• Learning Rate: 2e-5
• Per Device Train Batch Size: 16
• Per Device Eval Batch Size: 16
• Number of Train Epochs: 4
• Weight Decay: 0.01
• Metric for best model: macro_f1 (greater is better)
• Training regime: fp32

Evaluation

Testing Data, Factors & Metrics

Testing Data

The model was evaluated on a held-out test set (20% of the original dataset) from the "synthetic_emotions.csv" dataset.

Metrics

The evaluation metrics used were accuracy and macro F1-score.

Results

Summary

The model achieved the following performance on the test set:

• Test Loss: 0.0259
• Test Accuracy: 0.9946
• Test Macro F1-score: 0.9946