correct_twitter_RoBERTa_token_itr0_1e-05_editorials_01_03_2022-15_33_51

This model is a fine-tuned version of cardiffnlp/twitter-roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1138
• Precision: 0.5788
• Recall: 0.4712
• F1: 0.5195
• Accuracy: 0.9688

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: 1e-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: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	15	0.1316	0.04	0.0021	0.0040	0.9624
No log	2.0	30	0.1016	0.6466	0.4688	0.5435	0.9767
No log	3.0	45	0.0899	0.5873	0.4625	0.5175	0.9757
No log	4.0	60	0.0849	0.5984	0.4813	0.5335	0.9761
No log	5.0	75	0.0835	0.5984	0.4813	0.5335	0.9761

Framework versions

• Transformers 4.15.0
• Pytorch 1.10.1+cu113
• Datasets 1.18.0
• Tokenizers 0.10.3