Kyle1668/ag-news-76800-bert-base-uncased

ag-news-twitter-76800-bert-base-uncased

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

• F1: 0.9415
• Acc: 0.9416
• Loss: 0.5192

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	F1	Acc	Validation Loss
0.2328	1.0	4800	0.9289	0.9289	0.2082
0.2061	2.0	9600	0.9366	0.9367	0.2154
0.1488	3.0	14400	0.9401	0.9401	0.2181
0.114	4.0	19200	0.9280	0.9275	0.3199
0.0818	5.0	24000	0.9399	0.94	0.2953
0.051	6.0	28800	0.9402	0.9403	0.3828
0.0413	7.0	33600	0.9404	0.9403	0.4327
0.0342	8.0	38400	0.9395	0.9395	0.4291
0.0192	9.0	43200	0.9422	0.9422	0.4170
0.0204	10.0	48000	0.9374	0.9374	0.4761
0.0125	11.0	52800	0.9358	0.9359	0.5126
0.0124	12.0	57600	0.9415	0.9416	0.5192

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1