bert_12_layer_model_v2_complete_training_new_48_KD_wt_init

This model is a fine-tuned version of on the None dataset. It achieves the following results on the evaluation set:

• Loss: 231.6032
• Accuracy: 0.4123

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: 1e-05
• train_batch_size: 36
• eval_batch_size: 36
• seed: 10
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10000
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
606.9724	0.06	10000	581.9434	0.1439
499.7046	0.12	20000	509.0019	0.1448
483.3571	0.18	30000	487.3823	0.1451
374.2657	0.25	40000	364.4415	0.2699
330.4914	0.31	50000	324.6827	0.3265
307.9071	0.37	60000	297.7675	0.3516
293.5813	0.43	70000	284.2744	0.3672
282.876	0.49	80000	271.3589	0.3775
274.4863	0.55	90000	261.3694	0.3869
267.0072	0.61	100000	252.8204	0.3939
259.3755	0.68	110000	247.2895	0.3992
255.5614	0.74	120000	241.0253	0.4043
250.2624	0.8	130000	238.9221	0.4085
245.3816	0.86	140000	231.6032	0.4123

Framework versions

• Transformers 4.30.1
• Pytorch 1.14.0a0+410ce96
• Datasets 2.12.0
• Tokenizers 0.13.3