rogerpolo/results

distilbert-agnews-classification-fine-tune

This model is a fine-tuned version of distilbert/distilbert-base-uncased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1979
• Accuracy: 0.9434
• F1: 0.9435

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.4206	0.0667	500	0.3065	0.9021	0.9019
0.2721	0.1333	1000	0.2626	0.9151	0.9149
0.2403	0.2	1500	0.2584	0.9192	0.9190
0.2314	0.2667	2000	0.2398	0.9267	0.9263
0.232	0.3333	2500	0.2190	0.9318	0.9319
0.246	0.4	3000	0.1979	0.9338	0.9340
0.2092	0.4667	3500	0.2066	0.9309	0.9310
0.2171	0.5333	4000	0.2058	0.9353	0.9353
0.2102	0.6	4500	0.1999	0.9368	0.9370
0.2	0.6667	5000	0.1967	0.9363	0.9363
0.1952	0.7333	5500	0.2025	0.9358	0.9359
0.1963	0.8	6000	0.2062	0.9374	0.9375
0.2025	0.8667	6500	0.1918	0.9386	0.9388
0.1839	0.9333	7000	0.1943	0.9413	0.9414
0.2008	1.0	7500	0.1766	0.9420	0.9420
0.1467	1.0667	8000	0.1948	0.9426	0.9426
0.1502	1.1333	8500	0.1960	0.9413	0.9414
0.1331	1.2	9000	0.1977	0.9443	0.9444
0.1421	1.2667	9500	0.2006	0.9428	0.9428
0.1375	1.3333	10000	0.1931	0.9437	0.9437
0.1375	1.4	10500	0.1979	0.9434	0.9435

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1