twitter_RoBERTa_token_itr0_1e-05_editorials_01_03_2022-15_00_35

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: 0.1155
• Precision: 0.5720
• Recall: 0.4705
• F1: 0.5163
• Accuracy: 0.9687

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	0.1256	0.04	0.0021	0.0039	0.9624
No log	2.0	30	0.0963	0.7121	0.5711	0.6339	0.9794
No log	3.0	45	0.0844	0.6205	0.5732	0.5959	0.9778
No log	4.0	60	0.0770	0.6201	0.5856	0.6023	0.9778
No log	5.0	75	0.0750	0.6174	0.5856	0.6011	0.9777

Framework versions

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