1c43681abd0a71379ce9fa7aab3e47a3

This model is a fine-tuned version of albert/albert-xlarge-v2 on the contemmcm/cls_20newsgroups dataset. It achieves the following results on the evaluation set:

• Loss: 3.0023
• Data Size: 1.0
• Epoch Runtime: 176.2254
• Accuracy: 0.0491
• F1 Macro: 0.0047

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	3.1232	0	14.5188	0.0491	0.0047
No log	1	499	3.0612	0.0078	15.8061	0.0507	0.0114
0.0309	2	998	3.0182	0.0156	16.8908	0.0524	0.0105
0.0553	3	1497	3.0452	0.0312	19.5237	0.0423	0.0041
0.1036	4	1996	3.0300	0.0625	24.7055	0.0517	0.0049
3.0323	5	2495	3.0107	0.125	34.7290	0.0491	0.0047
3.0264	6	2994	3.0121	0.25	54.8967	0.0532	0.0050
3.0224	7	3493	3.0120	0.5	95.5008	0.0532	0.0050
3.0152	8.0	3992	3.0045	1.0	176.3653	0.0496	0.0047
3.0088	9.0	4491	3.0086	1.0	176.2609	0.0491	0.0047
3.0048	10.0	4990	3.0102	1.0	176.2565	0.0444	0.0042
3.0027	11.0	5489	2.9996	1.0	176.2540	0.0502	0.0048
3.0029	12.0	5988	3.0063	1.0	176.2654	0.0444	0.0042
3.0037	13.0	6487	3.0022	1.0	176.4287	0.0502	0.0048
2.9998	14.0	6986	3.0003	1.0	176.1715	0.0512	0.0049
3.0041	15.0	7485	2.9995	1.0	176.3333	0.0504	0.0048
3.0014	16.0	7984	2.9987	1.0	176.2703	0.0512	0.0065
3.0045	17.0	8483	3.0029	1.0	176.3074	0.0479	0.0046
2.9979	18.0	8982	3.0006	1.0	176.2061	0.0491	0.0047
3.0008	19.0	9481	3.0038	1.0	176.1014	0.0479	0.0046
3.0004	20.0	9980	3.0023	1.0	176.2254	0.0491	0.0047

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.1