scenario-kd-scr-ner-full-mdeberta_data-univner_half66

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: 370.8975
• Precision: 0.3767
• Recall: 0.4188
• F1: 0.3966
• Accuracy: 0.9231

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: 66
• 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
639.3549	0.5828	500	570.6539	0.7143	0.0014	0.0029	0.9241
539.3257	1.1655	1000	524.4883	0.3354	0.0698	0.1156	0.9262
488.2303	1.7483	1500	491.0639	0.3430	0.1431	0.2020	0.9307
453.2781	2.3310	2000	469.1648	0.3407	0.2689	0.3006	0.9306
427.1957	2.9138	2500	451.5781	0.3777	0.3150	0.3435	0.9321
406.9511	3.4965	3000	440.5732	0.3440	0.3317	0.3377	0.9249
390.8488	4.0793	3500	429.3404	0.3468	0.3828	0.3639	0.9241
376.6864	4.6620	4000	421.3924	0.3388	0.4020	0.3677	0.9188
365.0702	5.2448	4500	406.7076	0.3634	0.4164	0.3881	0.9260
354.3511	5.8275	5000	395.5925	0.4385	0.3870	0.4111	0.9351
345.6264	6.4103	5500	389.3924	0.4030	0.4069	0.4049	0.9305
338.4962	6.9930	6000	387.7135	0.3570	0.4142	0.3835	0.9201
331.6204	7.5758	6500	377.1432	0.4232	0.3945	0.4083	0.9326
326.8345	8.1585	7000	375.8239	0.3924	0.4242	0.4077	0.9253
322.7788	8.7413	7500	373.5773	0.3755	0.4210	0.3970	0.9233
320.0888	9.3240	8000	369.3580	0.4010	0.4204	0.4105	0.9279
318.3835	9.9068	8500	370.8975	0.3767	0.4188	0.3966	0.9231

Framework versions

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