distilgpt2-finetuned

This model is a fine-tuned version of distilbert/distilgpt2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.7322
• Bleu: 0.0145
• Bertscore Precision: 0.1505
• Bertscore Recall: 0.1674
• Bertscore F1: 0.1581

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Bertscore Precision	Bertscore Recall	Bertscore F1
5.0087	1.0	3223	3.9456	0.0088	0.1478	0.1638	0.1551
4.8889	2.0	6446	3.7706	0.0093	0.1480	0.1642	0.1554
4.9152	3.0	9669	3.6252	0.0097	0.1483	0.1646	0.1557
4.647	4.0	12892	3.5105	0.0103	0.1486	0.1649	0.1560
4.4683	5.0	16115	3.4093	0.0108	0.1489	0.1652	0.1563
4.4007	6.0	19338	3.3225	0.0110	0.1491	0.1654	0.1565
4.3966	7.0	22561	3.2444	0.0115	0.1493	0.1656	0.1567
4.3414	8.0	25784	3.1662	0.0117	0.1494	0.1657	0.1568
4.2446	9.0	29007	3.1021	0.0122	0.1497	0.1660	0.1571
4.2464	10.0	32230	3.0384	0.0125	0.1499	0.1662	0.1573
4.1739	11.0	35453	2.9789	0.0128	0.1499	0.1665	0.1574
4.08	12.0	38676	2.9295	0.0131	0.1501	0.1666	0.1576
4.001	13.0	41899	2.8857	0.0135	0.1502	0.1668	0.1577
3.9277	14.0	45122	2.8464	0.0136	0.1502	0.1669	0.1578
3.9709	15.0	48345	2.8137	0.0139	0.1503	0.1670	0.1578
3.9192	16.0	51568	2.7872	0.0141	0.1503	0.1672	0.1579
3.8916	17.0	54791	2.7644	0.0143	0.1504	0.1673	0.1580
3.8489	18.0	58014	2.7475	0.0144	0.1505	0.1674	0.1581
3.9091	19.0	61237	2.7364	0.0145	0.1505	0.1674	0.1581
3.9271	20.0	64460	2.7322	0.0145	0.1505	0.1674	0.1581

Framework versions

• Transformers 4.40.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1