gpt2-medium-wikitext

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

• Loss: 3.1671
• Accuracy: 0.4217
• Perplexity: 23.7377
• Bleu: 0.1460

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: 0.0001
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• 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
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Perplexity	Bleu
6.0809	0.2806	500	5.9580	0.1883	386.8254	0.0333
5.0644	0.5612	1000	4.9191	0.2623	136.8761	0.0651
4.3331	0.8418	1500	4.2124	0.3226	67.5163	0.0890
3.9451	1.1223	2000	3.8835	0.3532	48.5942	0.1090
3.7568	1.4029	2500	3.7051	0.3684	40.6559	0.1226
3.6478	1.6835	3000	3.5827	0.3787	35.9710	0.1311
3.5435	1.9641	3500	3.4940	0.3877	32.9179	0.1343
3.4222	2.2447	4000	3.4292	0.3936	30.8527	0.1343
3.3604	2.5253	4500	3.3728	0.3990	29.1601	0.1414
3.3288	2.8058	5000	3.3269	0.4038	27.8518	0.1381
3.2074	3.0864	5500	3.2887	0.4079	26.8092	0.1423
3.2007	3.3670	6000	3.2605	0.4115	26.0632	0.1464
3.1787	3.6476	6500	3.2328	0.4140	25.3497	0.1428
3.1529	3.9282	7000	3.2085	0.4166	24.7424	0.1425
3.0849	4.2088	7500	3.1921	0.4184	24.3384	0.1430
3.0471	4.4893	8000	3.1796	0.4202	24.0366	0.1428
3.0569	4.7699	8500	3.1671	0.4217	23.7377	0.1460

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0