dh-unibe/luther-xl

22_12_13_luther_blocks_xl_fp16_5ep

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

• Loss: 2.8833
• Accuracy: 0.4196

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	0.19	50	3.0276	0.3997
No log	0.38	100	2.9185	0.4143
No log	0.58	150	2.8846	0.4189
No log	0.77	200	2.8833	0.4196
3.0309	0.96	250	2.8833	0.4196
3.0309	1.15	300	2.8833	0.4196
3.0309	1.35	350	2.8833	0.4196
3.0309	1.54	400	2.8833	0.4196
3.0309	1.73	450	2.8833	0.4196
2.8083	1.92	500	2.8833	0.4196
2.8083	2.12	550	2.8833	0.4196
2.8083	2.31	600	2.8833	0.4196
2.8083	2.5	650	2.8833	0.4196
2.8083	2.69	700	2.8833	0.4196
2.7992	2.88	750	2.8833	0.4196
2.7992	3.08	800	2.8833	0.4196
2.7992	3.27	850	2.8833	0.4196
2.7992	3.46	900	2.8833	0.4196
2.7992	3.65	950	2.8833	0.4196
2.7935	3.85	1000	2.8833	0.4196
2.7935	4.04	1050	2.8833	0.4196
2.7935	4.23	1100	2.8833	0.4196
2.7935	4.42	1150	2.8833	0.4196
2.7935	4.62	1200	2.8833	0.4196
2.8129	4.81	1250	2.8833	0.4196
2.8129	5.0	1300	2.8833	0.4196

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0
• Datasets 2.7.1
• Tokenizers 0.12.1