hbertv1-Massive-intent_48_KD

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_48_KD on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.8470
• Accuracy: 0.8357

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
2.994	1.0	180	2.1475	0.3901
1.7222	2.0	360	1.4146	0.6011
1.1889	3.0	540	1.1690	0.6990
0.9256	4.0	720	0.9700	0.7545
0.763	5.0	900	0.8986	0.7806
0.6351	6.0	1080	0.8898	0.7787
0.5374	7.0	1260	0.8604	0.7978
0.4587	8.0	1440	0.8444	0.8101
0.3822	9.0	1620	0.8520	0.8087
0.3301	10.0	1800	0.8309	0.8185
0.2713	11.0	1980	0.8313	0.8249
0.2257	12.0	2160	0.8499	0.8254
0.1947	13.0	2340	0.8375	0.8298
0.162	14.0	2520	0.8428	0.8352
0.1369	15.0	2700	0.8470	0.8357

Framework versions

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