hbertv1-emotion-logit_KD-tiny_ffn_2

This model is a fine-tuned version of gokuls/model_v1_complete_training_wt_init_48_tiny_freeze_new_ffn_2 on the emotion dataset. It achieves the following results on the evaluation set:

• Loss: 0.4740
• Accuracy: 0.9005

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
3.1189	1.0	250	2.6103	0.514
2.0804	2.0	500	1.4939	0.7695
1.2677	3.0	750	0.8999	0.8445
0.8885	4.0	1000	0.6887	0.874
0.7023	5.0	1250	0.5821	0.889
0.5796	6.0	1500	0.5364	0.8875
0.5106	7.0	1750	0.5043	0.89
0.4603	8.0	2000	0.5055	0.889
0.405	9.0	2250	0.4903	0.89
0.3782	10.0	2500	0.4793	0.8965
0.3488	11.0	2750	0.4832	0.8945
0.3301	12.0	3000	0.4740	0.9005
0.3163	13.0	3250	0.4768	0.89
0.2983	14.0	3500	0.4925	0.887
0.2835	15.0	3750	0.4764	0.898
0.2702	16.0	4000	0.4856	0.8905
0.2522	17.0	4250	0.4829	0.897

Framework versions

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