gpt2-kl_01_04_hscnspecial-hs_cn

This model is a fine-tuned version of gpt2-medium on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5583

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: 4
• seed: 21
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
73.5419	0.02	10	69.5826
46.0912	0.04	20	32.9478
13.5577	0.06	30	10.6427
6.8429	0.08	40	4.2253
3.5596	0.1	50	2.0345
1.4308	0.12	60	1.0810
1.0569	0.14	70	0.8577
1.0017	0.16	80	0.8291
0.7285	0.18	90	0.6919
0.7232	0.2	100	0.6110
0.5979	0.22	110	0.6133
0.6368	0.24	120	0.6051
0.6606	0.26	130	0.6075
0.6428	0.28	140	0.5920
0.6106	0.3	150	0.5754
0.6218	0.32	160	0.5741
0.534	0.34	170	0.5633
0.6411	0.36	180	0.5685
0.6353	0.38	190	0.5805
0.5922	0.4	200	0.5556
0.6924	0.42	210	0.5595
0.6429	0.44	220	0.5601
0.5428	0.46	230	0.5583

Framework versions

• Transformers 4.29.0.dev0
• Pytorch 1.12.0a0+bd13bc6
• Datasets 2.12.0
• Tokenizers 0.13.3