twitter-roberta-base-mar2022-finetuned-emotion

This model is a fine-tuned version of cardiffnlp/twitter-roberta-base-mar2022 on the tweet_eval dataset. It achieves the following results on the evaluation set:

• Loss: 0.5146
• Accuracy: 0.8191
• F1: 0.8171

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.8945	1.0	102	0.5831	0.7995	0.7887
0.5176	2.0	204	0.5266	0.8235	0.8200

Framework versions

• Transformers 4.19.3
• Pytorch 1.11.0+cu102
• Datasets 2.2.2
• Tokenizers 0.12.1