bert_12_layer_model_v4_48_massive

This model is a fine-tuned version of gokuls/bert_12_layer_model_v4_complete_training_48 on the massive dataset. It achieves the following results on the evaluation set:

• Loss: 0.9030
• Accuracy: 0.8392

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
3.1903	1.0	180	2.2860	0.4048
1.81	2.0	360	1.3971	0.6134
1.1787	3.0	540	1.0028	0.7270
0.8518	4.0	720	0.8662	0.7718
0.6633	5.0	900	0.8229	0.7885
0.5208	6.0	1080	0.8214	0.8037
0.4179	7.0	1260	0.7887	0.8008
0.3308	8.0	1440	0.7357	0.8293
0.2518	9.0	1620	0.7840	0.8195
0.1997	10.0	1800	0.7644	0.8283
0.1472	11.0	1980	0.8304	0.8318
0.1122	12.0	2160	0.8461	0.8347
0.0816	13.0	2340	0.8959	0.8328
0.0601	14.0	2520	0.8811	0.8382
0.0401	15.0	2700	0.9030	0.8392

Framework versions

• Transformers 4.34.0
• Pytorch 1.14.0a0+410ce96
• Datasets 2.14.5
• Tokenizers 0.14.1