My Sentiment Model

A 3-class sentiment classifier (negative, neutral, positive) fine-tuned from DistilBERT.

Training

  • Base model: distilbert-base-uncased
  • Dataset: tweet_eval (sentiment), using a sample of the training set
  • Epochs: 1

Results (test set)

  • Accuracy: 0.689
  • Macro F1: 0.6789632300455798

Limitations

  • English only, trained on tweets, so it may be weaker on long reviews
  • Struggles with sarcasm
  • Trained on a small sample, so accuracy is limited

[More Information Needed]

Downloads last month
-
Safetensors
Model size
67M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for illusionash/my-sentiment-model

Finetuned
(12776)
this model

Dataset used to train illusionash/my-sentiment-model