HBERTv1_48_L6_H64_A2_massive

This model is a fine-tuned version of gokuls/HBERTv1_48_L6_H64_A2 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 2.0289
• Accuracy: 0.4540

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
3.9723	1.0	180	3.7638	0.0910
3.5915	2.0	360	3.4160	0.1402
3.3315	3.0	540	3.1858	0.1545
3.0936	4.0	720	2.9377	0.2489
2.8827	5.0	900	2.7454	0.2607
2.7034	6.0	1080	2.5719	0.3005
2.5548	7.0	1260	2.4456	0.3301
2.4205	8.0	1440	2.3437	0.3689
2.3213	9.0	1620	2.2482	0.4043
2.2359	10.0	1800	2.1809	0.4112
2.1724	11.0	1980	2.1286	0.4289
2.1113	12.0	2160	2.0921	0.4442
2.067	13.0	2340	2.0534	0.4471
2.0388	14.0	2520	2.0381	0.4501
2.0222	15.0	2700	2.0289	0.4540

Framework versions

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