b2a13bdfdbee03869078392ce31ca69a

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

• Loss: 0.7541
• Data Size: 1.0
• Epoch Runtime: 29.5942
• Accuracy: 0.8377
• F1 Macro: 0.8391

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	3.0820	0	2.9678	0.0605	0.0151
No log	1	499	3.0567	0.0078	3.4622	0.0504	0.0048
0.031	2	998	3.0091	0.0156	3.3477	0.0391	0.0112
0.0543	3	1497	3.0022	0.0312	3.7892	0.0554	0.0097
0.1023	4	1996	3.0063	0.0625	4.8024	0.0423	0.0041
3.0041	5	2495	2.9995	0.125	6.3555	0.0517	0.0049
2.991	6	2994	3.0435	0.25	9.6819	0.0665	0.0181
1.949	7	3493	1.7583	0.5	16.3085	0.3546	0.2790
1.2499	8.0	3992	1.1927	1.0	29.7099	0.5761	0.5240
1.0707	9.0	4491	1.0339	1.0	29.6584	0.6615	0.6368
0.8532	10.0	4990	0.9110	1.0	29.5331	0.7132	0.6890
0.634	11.0	5489	0.8023	1.0	29.7321	0.7641	0.7594
0.5863	12.0	5988	0.7403	1.0	29.5549	0.8062	0.8034
0.5222	13.0	6487	0.6932	1.0	29.6895	0.8133	0.8116
0.4108	14.0	6986	0.7164	1.0	29.6142	0.8193	0.8177
0.4135	15.0	7485	0.7475	1.0	29.4822	0.8233	0.8194
0.3594	16.0	7984	0.7092	1.0	29.5987	0.8274	0.8259
0.3296	17.0	8483	0.6517	1.0	29.6032	0.8475	0.8450
0.2996	18.0	8982	0.6841	1.0	29.5800	0.8402	0.8352
0.2559	19.0	9481	0.6700	1.0	29.5649	0.8511	0.8499
0.25	20.0	9980	0.7420	1.0	29.6743	0.8395	0.8366
0.2695	21.0	10479	0.7541	1.0	29.5942	0.8377	0.8391

Framework versions

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