bert-small-juman-bpe

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

• Accuracy: 0.6317
• Loss: 1.7829

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: 0.0001
• train_batch_size: 256
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 3
• total_train_batch_size: 768
• total_eval_batch_size: 24
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.01
• num_epochs: 14
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Accuracy	Validation Loss
2.3892	1.0	69472	0.5637	2.2498
2.2219	2.0	138944	0.5873	2.0785
2.1453	3.0	208416	0.5984	2.0019
2.1	4.0	277888	0.6059	1.9531
2.068	5.0	347360	0.6106	1.9169
2.0405	6.0	416832	0.6146	1.8921
2.0174	7.0	486304	0.6175	1.8711
2.0002	8.0	555776	0.6205	1.8527
1.9838	9.0	625248	0.6225	1.8381
1.9691	10.0	694720	0.6248	1.8239
1.9551	11.0	764192	0.6265	1.8125
1.9406	12.0	833664	0.6288	1.8002
1.9293	13.0	903136	0.6310	1.7871
1.9247	14.0	972608	0.6317	1.7829

Framework versions

• Transformers 4.19.2
• Pytorch 1.12.0+cu116
• Datasets 2.2.2
• Tokenizers 0.12.1