hbertv2-Massive-intent_48_KD_w_in

This model is a fine-tuned version of gokuls/bert_12_layer_model_v2_complete_training_new_48_KD_wt_init on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.8498
• Accuracy: 0.8667

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
1.875	1.0	180	0.8687	0.7698
0.7825	2.0	360	0.6526	0.8303
0.5611	3.0	540	0.6714	0.8337
0.4381	4.0	720	0.6374	0.8303
0.3389	5.0	900	0.6718	0.8387
0.2487	6.0	1080	0.6282	0.8515
0.1851	7.0	1260	0.7070	0.8490
0.1535	8.0	1440	0.7197	0.8490
0.1125	9.0	1620	0.7224	0.8564
0.0767	10.0	1800	0.7309	0.8642
0.0556	11.0	1980	0.7612	0.8618
0.0366	12.0	2160	0.8228	0.8623
0.0212	13.0	2340	0.8310	0.8662
0.0135	14.0	2520	0.8537	0.8642
0.0081	15.0	2700	0.8498	0.8667

Framework versions

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