scenario-kd-scr-ner-full-mdeberta_data-univner_half55

This model is a fine-tuned version of haryoaw/scenario-TCR-NER_data-univner_half on the None dataset. It achieves the following results on the evaluation set:

• Loss: 369.5634
• Precision: 0.3741
• Recall: 0.4149
• F1: 0.3935
• Accuracy: 0.9238

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-05
• train_batch_size: 8
• eval_batch_size: 32
• seed: 55
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
638.1208	0.5828	500	570.1356	1.0	0.0004	0.0009	0.9241
540.7434	1.1655	1000	524.2383	0.2886	0.0521	0.0882	0.9253
490.1496	1.7483	1500	493.6944	0.2927	0.1682	0.2137	0.9297
454.2343	2.3310	2000	471.6737	0.3338	0.2411	0.2800	0.9293
429.0546	2.9138	2500	450.0691	0.4515	0.2450	0.3176	0.9362
407.3853	3.4965	3000	436.8550	0.3570	0.2877	0.3186	0.9304
390.9232	4.0793	3500	424.0557	0.3848	0.3245	0.3521	0.9313
376.2085	4.6620	4000	417.5171	0.3349	0.3800	0.3561	0.9216
364.3876	5.2448	4500	404.9495	0.3708	0.3766	0.3737	0.9276
354.1139	5.8275	5000	398.9413	0.3534	0.3877	0.3697	0.9226
344.7845	6.4103	5500	394.6783	0.3273	0.4193	0.3676	0.9143
337.8201	6.9930	6000	382.1873	0.3881	0.3900	0.3891	0.9289
330.94	7.5758	6500	381.2287	0.3480	0.4074	0.3754	0.9188
326.2092	8.1585	7000	372.6259	0.4087	0.3877	0.3979	0.9306
322.0348	8.7413	7500	374.2613	0.3530	0.4144	0.3812	0.9184
319.0297	9.3240	8000	370.3267	0.3774	0.4131	0.3944	0.9239
317.8106	9.9068	8500	369.5634	0.3741	0.4149	0.3935	0.9238

Framework versions

• Transformers 4.44.2
• Pytorch 2.1.1+cu121
• Datasets 2.14.5
• Tokenizers 0.19.1