hbertv1-Massive-intent_48_KD_w_in

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_48_KD_wt_init on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.8731
• Accuracy: 0.8706

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
2.1886	1.0	180	0.9480	0.7359
0.8407	2.0	360	0.7278	0.8072
0.5816	3.0	540	0.6572	0.8387
0.4195	4.0	720	0.6760	0.8406
0.3106	5.0	900	0.6604	0.8490
0.2447	6.0	1080	0.6951	0.8446
0.171	7.0	1260	0.7304	0.8524
0.1357	8.0	1440	0.7646	0.8485
0.1022	9.0	1620	0.7845	0.8529
0.0733	10.0	1800	0.8051	0.8588
0.051	11.0	1980	0.8238	0.8662
0.033	12.0	2160	0.8675	0.8667
0.0226	13.0	2340	0.8799	0.8672
0.0128	14.0	2520	0.8867	0.8672
0.007	15.0	2700	0.8731	0.8706

Framework versions

• Transformers 4.30.2
• Pytorch 1.14.0a0+410ce96
• Datasets 2.13.0
• Tokenizers 0.13.3