hbertv1-massive-logit_KD_new

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_48 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.5587
• Accuracy: 0.8539

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
2.3071	1.0	180	1.0522	0.7019
0.9618	2.0	360	0.7397	0.7836
0.6757	3.0	540	0.7535	0.7831
0.5344	4.0	720	0.6076	0.8269
0.4319	5.0	900	0.6585	0.8165
0.3648	6.0	1080	0.5726	0.8362
0.3326	7.0	1260	0.5642	0.8372
0.2904	8.0	1440	0.5858	0.8352
0.2554	9.0	1620	0.5521	0.8411
0.2314	10.0	1800	0.5571	0.8436
0.2192	11.0	1980	0.5479	0.8470
0.2	12.0	2160	0.5587	0.8539
0.1924	13.0	2340	0.5430	0.8480
0.1683	14.0	2520	0.5647	0.8490
0.1703	15.0	2700	0.5467	0.8515
0.1598	16.0	2880	0.5578	0.8510
0.1522	17.0	3060	0.5682	0.8431

Framework versions

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