HBERTv1_48_L8_H64_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L8_H64_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 2.0023
• Accuracy: 0.4506

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 33
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.9694	1.0	180	3.7614	0.0949
3.5709	2.0	360	3.3717	0.1323
3.2745	3.0	540	3.1313	0.1968
3.0491	4.0	720	2.9227	0.2086
2.8486	5.0	900	2.7431	0.2238
2.6671	6.0	1080	2.5865	0.2848
2.514	7.0	1260	2.4468	0.3212
2.3707	8.0	1440	2.3252	0.3620
2.262	9.0	1620	2.2383	0.3866
2.1746	10.0	1800	2.1570	0.4171
2.0999	11.0	1980	2.1083	0.4309
2.0442	12.0	2160	2.0581	0.4383
1.9992	13.0	2340	2.0297	0.4432
1.9728	14.0	2520	2.0104	0.4461
1.9461	15.0	2700	2.0023	0.4506

Framework versions

• Transformers 4.34.0
• Pytorch 1.14.0a0+410ce96
• Datasets 2.14.5
• Tokenizers 0.14.0