tergel/bert-base-mongolian-uncased

bert_mn

This model is a fine-tuned version of tergel/bert_mn on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.2973

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: 8e-05
• train_batch_size: 256
• eval_batch_size: 512
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss
2.5775	1.0	577	2.5137
2.5628	2.0	1154	2.4950
2.5396	3.0	1731	2.4834
2.5172	4.0	2308	2.4643
2.4969	5.0	2885	2.4481
2.4754	6.0	3462	2.4253
2.4578	7.0	4039	2.4195
2.4403	8.0	4616	2.4009
2.4216	9.0	5193	2.3863
2.4075	10.0	5770	2.3794
2.3927	11.0	6347	2.3640
2.3787	12.0	6924	2.3516
2.365	13.0	7501	2.3403
2.3546	14.0	8078	2.3344
2.3423	15.0	8655	2.3268
2.3336	16.0	9232	2.3171
2.3265	17.0	9809	2.3149
2.3168	18.0	10386	2.3049
2.312	19.0	10963	2.3032
2.3047	20.0	11540	2.2973

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1