HBERTv1_48_L8_H512_A8_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L8_H512_A8 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.8009
• Accuracy: 0.8569

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
3.2383	1.0	180	2.0110	0.4747
1.5051	2.0	360	1.0904	0.7093
0.8998	3.0	540	0.8544	0.7727
0.661	4.0	720	0.7029	0.8160
0.5052	5.0	900	0.6987	0.8131
0.3889	6.0	1080	0.6901	0.8244
0.3062	7.0	1260	0.6746	0.8352
0.2422	8.0	1440	0.6946	0.8396
0.1834	9.0	1620	0.7290	0.8441
0.1402	10.0	1800	0.7416	0.8406
0.1098	11.0	1980	0.7828	0.8387
0.0803	12.0	2160	0.7700	0.8460
0.0612	13.0	2340	0.7891	0.8515
0.0468	14.0	2520	0.8009	0.8569
0.0375	15.0	2700	0.8032	0.8569

Framework versions

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