9121d1d50bbf554a3b7b889335e44376

This model is a fine-tuned version of google-bert/bert-large-uncased-whole-word-masking on the contemmcm/cls_20newsgroups dataset. It achieves the following results on the evaluation set:

• Loss: 0.4598
• Data Size: 1.0
• Epoch Runtime: 70.6265
• Accuracy: 0.8868
• F1 Macro: 0.8859

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Data Size	Epoch Runtime	Accuracy	F1 Macro
No log	0	0	3.1094	0	4.7110	0.0517	0.0252
No log	1	499	3.0133	0.0078	5.2606	0.0585	0.0182
0.0308	2	998	3.0054	0.0156	6.0096	0.0819	0.0375
0.0548	3	1497	2.4469	0.0312	7.4072	0.2135	0.1511
0.094	4	1996	1.1431	0.0625	10.8968	0.6348	0.5936
1.0747	5	2495	0.8218	0.125	14.1808	0.7442	0.7209
0.6134	6	2994	0.6213	0.25	22.1164	0.8102	0.7988
0.4766	7	3493	0.4772	0.5	38.6400	0.8551	0.8536
0.3394	8.0	3992	0.3879	1.0	71.2268	0.8803	0.8803
0.3034	9.0	4491	0.4178	1.0	70.4474	0.8889	0.8884
0.2577	10.0	4990	0.4691	1.0	70.2248	0.8818	0.8797
0.1981	11.0	5489	0.5198	1.0	70.6249	0.8732	0.8718
0.1636	12.0	5988	0.4598	1.0	70.6265	0.8868	0.8859

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.3.0
• Tokenizers 0.22.1