DisambertSingleSense-base

This model is a fine-tuned version of answerdotai/ModernBERT-base on the semcor dataset. It achieves the following results on the evaluation set:

• Loss: 7.9132
• Precision: 0.7725
• Recall: 0.7594
• F1: 0.7659
• Matthews: 0.7589

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: inverse_sqrt
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Matthews
No log	0	0	83.4924	0.4375	0.3642	0.3975	0.3634
0.4971	1.0	28027	0.7339	0.7793	0.7669	0.7730	0.7664
0.3296	2.0	56054	0.9845	0.7756	0.7656	0.7705	0.7651
0.1843	3.0	84081	2.0537	0.7743	0.7616	0.7679	0.7611
0.0903	4.0	112108	3.9497	0.7729	0.7559	0.7643	0.7554
0.0171	5.0	140135	5.8641	0.7727	0.7555	0.7640	0.7550
0.0394	6.0	168162	6.5708	0.7747	0.7555	0.7650	0.7550
0.0011	7.0	196189	7.4188	0.7705	0.7550	0.7627	0.7545
0.0231	8.0	224216	7.0225	0.7762	0.7621	0.7691	0.7615
0.0015	9.0	252243	6.9004	0.7766	0.7599	0.7681	0.7594
0.0000	10.0	280270	7.9132	0.7725	0.7594	0.7659	0.7589

Framework versions

• Transformers 5.2.0
• Pytorch 2.6.0+cu124
• Datasets 4.5.0
• Tokenizers 0.22.2