Ettin-tweet-emotion-68m Model

Tweet Emotion Classifier Model | Open Source | 68M Parameters

Overview

Ettin-tweet-emotion-68m is based on the ettin-encoder-68M model and fine-tuned over the TweetEval emotion dataset. This model can detect four types of emotions in tweets: Anger, Joy, Sadness and Optimism. This model achieves 76.64 (Accuracy) and 73.09 (Macro F1-score) points on TweetEval emotion test set.

Usage

from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline

# Load the fine-tuned model and tokenizer
model_path = "kalyan-ks/ettin-tweet-emotion-68m"

model = AutoModelForSequenceClassification.from_pretrained(model_path)
tokenizer = AutoTokenizer.from_pretrained(model_path)

# Create text-classification pipeline with the fine-tuned model
classifier = pipeline(
    "text-classification",
    model=model,
    tokenizer=tokenizer,
)

sample_texts = [
    "I am so excited and happy about this new journey!",
    "This delay is completely unacceptable and frustrating.",
    "Everything will work out for the best, staying hopeful.",
    "I feel heartbroken and deeply saddened today.",
]

predictions = classifier(sample_texts)

for text, pred in zip(sample_texts, predictions):
    print(f"Text: {text}")
    print(f"Predicted Emotion: {pred['label']} (Confidence: {pred['score']:.4f})\n")

"""
Output:

Text: I am so excited and happy about this new journey!
Predicted Emotion: joy (Confidence: 0.9715)

Text: This delay is completely unacceptable and frustrating.
Predicted Emotion: anger (Confidence: 0.9378)

Text: Everything will work out for the best, staying hopeful.
Predicted Emotion: optimism (Confidence: 0.9671)

Text: I feel heartbroken and deeply saddened today.
Predicted Emotion: sadness (Confidence: 0.9953)
"""

Citation

@misc{ettin-tweet-emotion-68m,
  title = {ettin-tweet-emotion-68m-2026: Tweet Emotion Classifier Model},
  author = {Kalyan KS},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/kalyan-ks/ettin-tweet-emotion-68m}
}