Keyurjotaniya007/minilm-emotion-en-v3

emotion-advance-classifier (go-emotion-raw)

This model is a fine-tuned version of microsoft/MiniLM-L12-H384-uncased on the go-emotion-raw dataset. It achieves the following results on the evaluation set:

• Loss: 1.8743029832839966

• Accuracy: 0.4426907901068232

• F1: 0.41514313919332535

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-5

• train_batch_size: 4

• gradient_accumulation_steps: 2

• eval_batch_size: 4

• seed: 42

• weight_decay: 0.01

• lr_scheduler_type: linear

• metric_for_best_model: "accuracy"

• warmup_ratio: 0.1

• num_epochs: 3