HBERTv1_48_L4_H64_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L4_H64_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 2.3014
• Accuracy: 0.3261

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
4.0	1.0	180	3.8278	0.0723
3.6366	2.0	360	3.4279	0.1117
3.3385	3.0	540	3.1935	0.1638
3.1113	4.0	720	2.9828	0.1909
2.9324	5.0	900	2.8344	0.2130
2.7882	6.0	1080	2.7100	0.2523
2.6832	7.0	1260	2.6215	0.2774
2.5965	8.0	1440	2.5459	0.2887
2.5244	9.0	1620	2.4872	0.2966
2.4603	10.0	1800	2.4261	0.3010
2.3987	11.0	1980	2.3758	0.3153
2.3615	12.0	2160	2.3469	0.3217
2.3292	13.0	2340	2.3241	0.3212
2.3071	14.0	2520	2.3100	0.3212
2.288	15.0	2700	2.3014	0.3261

Framework versions

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