deberta-v2-xxlarge-otat-recommened-hp

This model is a fine-tuned version of microsoft/deberta-v2-xxlarge on the DandinPower/review_onlytitleandtext dataset. It achieves the following results on the evaluation set:

• Loss: 0.7864
• Accuracy: 0.6741
• Macro F1: 0.6719

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: 3e-06
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 64
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1
0.9641	0.46	200	0.8451	0.6327	0.6341
0.8263	0.91	400	0.7768	0.6651	0.6650
0.7605	1.37	600	0.7842	0.667	0.6667
0.7496	1.83	800	0.7790	0.6659	0.6650
0.7034	2.29	1000	0.7738	0.67	0.6639
0.7134	2.74	1200	0.7671	0.6694	0.6698
0.6839	3.2	1400	0.7754	0.6743	0.6770
0.6699	3.66	1600	0.7853	0.6711	0.6666
0.6502	4.11	1800	0.7789	0.671	0.6692
0.6431	4.57	2000	0.7864	0.6741	0.6719

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2