7a0002daa5fbeeb780fcadc703c9c8e1

This model is a fine-tuned version of albert/albert-base-v1 on the contemmcm/cls_20newsgroups dataset. It achieves the following results on the evaluation set:

• Loss: 0.5668
• Data Size: 1.0
• Epoch Runtime: 20.3793
• Accuracy: 0.8642
• F1 Macro: 0.8649

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	3.0976	0	2.0165	0.0582	0.0097
No log	1	499	3.0257	0.0078	2.4801	0.0590	0.0299
0.0306	2	998	3.0250	0.0156	2.3193	0.0444	0.0107
0.055	3	1497	3.0051	0.0312	2.6349	0.0547	0.0138
0.1028	4	1996	2.4268	0.0625	3.2481	0.2815	0.2215
1.8326	5	2495	1.2403	0.125	4.3945	0.6482	0.6169
0.9645	6	2994	0.8344	0.25	6.6490	0.7482	0.7361
0.7429	7	3493	0.6370	0.5	11.3379	0.7996	0.7981
0.5024	8.0	3992	0.5159	1.0	20.4535	0.8357	0.8323
0.4125	9.0	4491	0.5398	1.0	20.3923	0.8385	0.8368
0.2905	10.0	4990	0.4598	1.0	20.3443	0.8586	0.8555
0.2458	11.0	5489	0.4404	1.0	20.2957	0.8684	0.8676
0.2216	12.0	5988	0.4326	1.0	20.4670	0.8758	0.8758
0.1807	13.0	6487	0.4803	1.0	20.6052	0.8747	0.8739
0.1526	14.0	6986	0.4902	1.0	20.5148	0.875	0.8738
0.1708	15.0	7485	0.5261	1.0	20.4180	0.8677	0.8639
0.165	16.0	7984	0.5668	1.0	20.3793	0.8642	0.8649

Framework versions

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