hbertv2-massive-logit_KD_new

This model is a fine-tuned version of gokuls/bert_12_layer_model_v2_complete_training_new_48 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.5000
• Accuracy: 0.8657

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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.1966	1.0	180	0.9332	0.7304
0.8897	2.0	360	0.6628	0.8037
0.6274	3.0	540	0.6455	0.8131
0.4838	4.0	720	0.5652	0.8303
0.402	5.0	900	0.5829	0.8313
0.3357	6.0	1080	0.5725	0.8342
0.3005	7.0	1260	0.5702	0.8416
0.2668	8.0	1440	0.5699	0.8426
0.2415	9.0	1620	0.5622	0.8470
0.2241	10.0	1800	0.5600	0.8460
0.2048	11.0	1980	0.5696	0.8411
0.1921	12.0	2160	0.5586	0.8505
0.179	13.0	2340	0.5706	0.8534
0.1725	14.0	2520	0.5318	0.8529
0.159	15.0	2700	0.5232	0.8549
0.1497	16.0	2880	0.5545	0.8475
0.142	17.0	3060	0.5683	0.8485
0.1367	18.0	3240	0.5356	0.8534
0.1292	19.0	3420	0.5371	0.8524
0.1231	20.0	3600	0.5164	0.8628
0.1233	21.0	3780	0.5430	0.8559
0.1105	22.0	3960	0.5454	0.8534
0.1088	23.0	4140	0.5000	0.8657
0.1066	24.0	4320	0.5428	0.8574
0.0972	25.0	4500	0.5136	0.8544
0.097	26.0	4680	0.5112	0.8603
0.09	27.0	4860	0.4910	0.8628
0.0888	28.0	5040	0.5058	0.8637

Framework versions

• Transformers 4.35.2
• Pytorch 1.14.0a0+410ce96
• Datasets 2.15.0
• Tokenizers 0.15.0