fd055b8a7b2e33bf3b010f4737f1eb59

This model is a fine-tuned version of distilbert/distilgpt2 on the nyu-mll/glue [qnli] dataset. It achieves the following results on the evaluation set:

• Loss: 0.4802
• Data Size: 1.0
• Epoch Runtime: 128.0936
• Accuracy: 0.8474
• F1 Macro: 0.8473
• Rouge1: 0.8474
• Rouge2: 0.0
• Rougel: 0.8476
• Rougelsum: 0.8474

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	Rouge1	Rouge2	Rougel	Rougelsum
No log	0	0	5.1267	0	3.3501	0.4945	0.3312	0.4949	0.0	0.4945	0.4946
No log	1	3273	0.6937	0.0078	4.6855	0.5489	0.5475	0.5487	0.0	0.5491	0.5485
0.0153	2	6546	0.6946	0.0156	5.5811	0.5143	0.3835	0.5145	0.0	0.5147	0.5144
0.6293	3	9819	0.5662	0.0312	7.9733	0.7134	0.7005	0.7134	0.0	0.7136	0.7134
0.5207	4	13092	0.5022	0.0625	11.8937	0.7564	0.7513	0.7566	0.0	0.7567	0.7562
0.4762	5	16365	0.4388	0.125	19.7312	0.8006	0.7996	0.8006	0.0	0.8007	0.8002
0.4694	6	19638	0.4937	0.25	36.1310	0.7717	0.7662	0.7715	0.0	0.7719	0.7717
0.4015	7	22911	0.4038	0.5	68.0457	0.8193	0.8179	0.8193	0.0	0.8193	0.8193
0.3966	8.0	26184	0.3709	1.0	131.4634	0.8375	0.8367	0.8377	0.0	0.8379	0.8373
0.2947	9.0	29457	0.3999	1.0	131.2547	0.8425	0.8419	0.8425	0.0	0.8426	0.8425
0.2218	10.0	32730	0.3901	1.0	131.5104	0.8542	0.8540	0.8542	0.0	0.8540	0.8542
0.1933	11.0	36003	0.4703	1.0	130.8935	0.8447	0.8441	0.8445	0.0	0.8447	0.8447
0.1637	12.0	39276	0.4802	1.0	128.0936	0.8474	0.8473	0.8474	0.0	0.8476	0.8474

Framework versions

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