hbertv2-Massive-intent_w_in

This model is a fine-tuned version of gokuls/bert_12_layer_model_v2_complete_training_new_wt_init on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.8617
• Accuracy: 0.8701

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
1.7408	1.0	180	0.8947	0.7600
0.7657	2.0	360	0.7246	0.8077
0.5442	3.0	540	0.7033	0.8259
0.3906	4.0	720	0.7175	0.8278
0.2839	5.0	900	0.6561	0.8465
0.2105	6.0	1080	0.6862	0.8485
0.1456	7.0	1260	0.7010	0.8574
0.1227	8.0	1440	0.7380	0.8524
0.0859	9.0	1620	0.8052	0.8539
0.0584	10.0	1800	0.8228	0.8593
0.0484	11.0	1980	0.8198	0.8588
0.028	12.0	2160	0.8731	0.8569
0.0202	13.0	2340	0.8640	0.8647
0.0103	14.0	2520	0.8691	0.8637
0.0059	15.0	2700	0.8617	0.8701

Framework versions

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