scenario-kd-scr-ner-full-mdeberta_data-univner_half44

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: 364.7336
• Precision: 0.3918
• Recall: 0.4292
• F1: 0.4096
• Accuracy: 0.9267

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: 44
• 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
637.6988	0.5828	500	570.9927	0.6154	0.0012	0.0023	0.9241
541.2212	1.1655	1000	524.6858	0.3571	0.0353	0.0643	0.9251
490.5141	1.7483	1500	492.2816	0.3048	0.1754	0.2227	0.9310
455.7006	2.3310	2000	474.9406	0.3064	0.2626	0.2828	0.9273
430.062	2.9138	2500	452.2111	0.3632	0.3073	0.3329	0.9298
408.0248	3.4965	3000	434.8791	0.3994	0.3220	0.3566	0.9341
390.2744	4.0793	3500	424.2673	0.3727	0.3444	0.3580	0.9307
374.3932	4.6620	4000	411.0975	0.4020	0.3979	0.3999	0.9328
362.2752	5.2448	4500	403.7659	0.3614	0.3963	0.3781	0.9239
350.9508	5.8275	5000	392.9673	0.3736	0.3855	0.3795	0.9296
341.7654	6.4103	5500	385.3136	0.4030	0.3972	0.4001	0.9302
334.4205	6.9930	6000	380.2038	0.3773	0.4142	0.3949	0.9263
327.3654	7.5758	6500	375.4951	0.3694	0.4276	0.3964	0.9227
322.1269	8.1585	7000	372.3464	0.3650	0.4338	0.3965	0.9209
318.558	8.7413	7500	366.4694	0.3970	0.4191	0.4078	0.9295
315.4182	9.3240	8000	365.6752	0.3861	0.4409	0.4117	0.9260
314.0958	9.9068	8500	364.7336	0.3918	0.4292	0.4096	0.9267

Framework versions

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