bert-base-uncased-tajik-ner

This model is a fine-tuned version of bert-base-uncased on the wikiann dataset. It achieves the following results on the evaluation set:

• Loss: 1.2137
• Precision: 0.5042
• Recall: 0.5769
• F1: 0.5381
• Accuracy: 0.8481

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 200

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	2.0	50	0.9499	0.0450	0.0962	0.0613	0.6626
No log	4.0	100	0.7348	0.1549	0.2115	0.1789	0.7401
No log	6.0	150	0.6685	0.1916	0.3077	0.2362	0.8017
No log	8.0	200	0.7875	0.3923	0.4904	0.4359	0.8036
No log	10.0	250	0.7495	0.4225	0.5769	0.4878	0.8274
No log	12.0	300	0.8934	0.4198	0.5288	0.4681	0.8085
No log	14.0	350	0.9455	0.4758	0.5673	0.5175	0.8236
No log	16.0	400	0.9469	0.5893	0.6346	0.6111	0.8410
No log	18.0	450	0.9936	0.5333	0.6154	0.5714	0.8485
0.2726	20.0	500	0.9804	0.5	0.6058	0.5478	0.8519
0.2726	22.0	550	1.1035	0.5963	0.625	0.6103	0.8432
0.2726	24.0	600	1.0318	0.5856	0.625	0.6047	0.8576
0.2726	26.0	650	1.1820	0.4921	0.5962	0.5391	0.8221
0.2726	28.0	700	1.1204	0.4878	0.5769	0.5286	0.8311
0.2726	30.0	750	1.1911	0.5357	0.5769	0.5556	0.8376
0.2726	32.0	800	1.1747	0.5259	0.5865	0.5545	0.8394
0.2726	34.0	850	1.1403	0.5872	0.6154	0.6009	0.8542
0.2726	36.0	900	1.1824	0.5370	0.5577	0.5472	0.8330
0.2726	38.0	950	1.1467	0.5424	0.6154	0.5766	0.8440
0.003	40.0	1000	1.2148	0.5268	0.5673	0.5463	0.8360
0.003	42.0	1050	1.3478	0.5273	0.5577	0.5421	0.8266
0.003	44.0	1100	1.2137	0.5042	0.5769	0.5381	0.8481

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1