deberta-v3-base-figurative

This model is a fine-tuned version of microsoft/deberta-v3-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3840
• Macro F1: 0.8727
• Literal F1: 0.8941
• Idiom F1: 0.9697
• Metaphor F1: 0.7290
• Simile F1: 0.8980

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 0.06
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Macro F1	Literal F1	Idiom F1	Metaphor F1	Simile F1
0.7461	1.0	1425	0.3552	0.8241	0.8282	0.9568	0.6762	0.8352
0.6193	2.0	2850	0.2795	0.8446	0.8576	0.9637	0.6910	0.8659
0.4072	3.0	4275	0.3218	0.8726	0.8987	0.9710	0.7352	0.8855
0.3003	4.0	5700	0.3840	0.8727	0.8941	0.9697	0.7290	0.8980
0.2481	5.0	7125	0.4906	0.8679	0.8978	0.9711	0.7191	0.8835
0.1805	6.0	8550	0.4938	0.8700	0.9017	0.9745	0.7207	0.8830

Framework versions

• Transformers 5.12.1
• Pytorch 2.12.1+cu130
• Datasets 5.0.0
• Tokenizers 0.22.2