c6a4454aeda4deb54ff61407ecf37900

This model is a fine-tuned version of distilbert/distilgpt2 on the contemmcm/cls_20newsgroups dataset. It achieves the following results on the evaluation set:

• Loss: 0.4822
• Data Size: 1.0
• Epoch Runtime: 37.0891
• Accuracy: 0.8821
• F1 Macro: 0.8816

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	5.0978	0	3.5334	0.0504	0.0128
No log	1	499	3.4238	0.0078	4.1133	0.0426	0.0153
0.0398	2	998	3.1239	0.0156	4.2443	0.0507	0.0169
0.0572	3	1497	2.9863	0.0312	4.8523	0.0645	0.0361
0.1047	4	1996	2.6952	0.0625	5.8449	0.2409	0.1761
1.8677	5	2495	1.1029	0.125	7.9351	0.6439	0.6205
0.8438	6	2994	0.7646	0.25	12.4961	0.7533	0.7510
0.6083	7	3493	0.5710	0.5	20.3185	0.8125	0.8134
0.4196	8.0	3992	0.4166	1.0	36.9912	0.8579	0.8584
0.3273	9.0	4491	0.3872	1.0	36.8709	0.8732	0.8714
0.2355	10.0	4990	0.4100	1.0	37.4153	0.8788	0.8759
0.1491	11.0	5489	0.4409	1.0	36.8883	0.8775	0.8781
0.1305	12.0	5988	0.4363	1.0	37.1127	0.8846	0.8834
0.1037	13.0	6487	0.4822	1.0	37.0891	0.8821	0.8816

Framework versions

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