distilbert-emotion

This model is a fine-tuned version of distilbert-base-uncased on the emotion dataset. It achieves the following results on the evaluation set:

• Loss: 0.1333
• Accuracy: 0.9385

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	1.0	250	0.1954	0.926
0.3494	2.0	500	0.1472	0.937
0.3494	3.0	750	0.1333	0.9385

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cpu
• Datasets 2.14.5
• Tokenizers 0.14.1