dh-unibe/gpt2-larger-walser

22_09_07_walser_blocks_larger_fp16_5ep

This model is a fine-tuned version of stefan-it/german-gpt2-larger on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.1738
• Accuracy: 0.2760

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
• 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: 5.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.45	50	5.2435	0.1883
No log	0.91	100	4.6670	0.2259
No log	1.36	150	4.4367	0.2481
No log	1.82	200	4.3258	0.2597
6.175	2.27	250	4.2705	0.2648
6.175	2.73	300	4.2318	0.2697
6.175	3.18	350	4.2090	0.2719
6.175	3.64	400	4.1915	0.2740
6.175	4.09	450	4.1823	0.2746
3.9191	4.55	500	4.1761	0.2757
3.9191	5.0	550	4.1738	0.2760

Framework versions

• Transformers 4.22.0.dev0
• Pytorch 1.11.0.post202
• Datasets 2.3.2
• Tokenizers 0.12.1