roberta-base-ner-demo

This model is a fine-tuned version of bayartsogt/mongolian-roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1307
• Precision: 0.9299
• Recall: 0.9402
• F1: 0.9350
• Accuracy: 0.9805

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: 32
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.7194	0.9958	119	0.1195	0.7550	0.8328	0.7920	0.9602
0.103	2.0	239	0.0894	0.8341	0.8782	0.8556	0.9695
0.0517	2.9958	358	0.0761	0.9138	0.9321	0.9228	0.9792
0.0255	4.0	478	0.0921	0.9118	0.9287	0.9202	0.9778
0.016	4.9958	597	0.0945	0.9242	0.9343	0.9292	0.9794
0.0102	6.0	717	0.0978	0.9266	0.9382	0.9324	0.9801
0.0066	6.9958	836	0.1092	0.9265	0.9368	0.9316	0.9800
0.005	8.0	956	0.1150	0.9228	0.9366	0.9297	0.9796
0.0034	8.9958	1075	0.1189	0.9274	0.9373	0.9323	0.9800
0.003	10.0	1195	0.1242	0.9215	0.9360	0.9287	0.9793
0.0025	10.9958	1314	0.1288	0.9256	0.9375	0.9315	0.9797
0.0016	12.0	1434	0.1318	0.9273	0.9365	0.9319	0.9799
0.0015	12.9958	1553	0.1314	0.9286	0.9394	0.9340	0.9801
0.0013	14.0	1673	0.1308	0.9290	0.9393	0.9341	0.9803
0.0012	14.9372	1785	0.1307	0.9299	0.9402	0.9350	0.9805

Framework versions

• Transformers 4.40.2
• Pytorch 2.2.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1