hbertv1-emotion-logit_KD-tiny_ffn_1

This model is a fine-tuned version of gokuls/model_v1_complete_training_wt_init_48_tiny_freeze_new_ffn_1 on the emotion dataset. It achieves the following results on the evaluation set:

• Loss: 0.4844
• Accuracy: 0.899

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
3.1687	1.0	250	2.7070	0.543
2.3056	2.0	500	1.8664	0.6295
1.5184	3.0	750	1.1853	0.7675
1.063	4.0	1000	0.9176	0.8295
0.8313	5.0	1250	0.7328	0.8515
0.6624	6.0	1500	0.6705	0.866
0.5695	7.0	1750	0.5983	0.8835
0.4801	8.0	2000	0.5658	0.8825
0.4243	9.0	2250	0.5285	0.8885
0.3828	10.0	2500	0.5358	0.884
0.3447	11.0	2750	0.4861	0.8895
0.3245	12.0	3000	0.4948	0.8905
0.3036	13.0	3250	0.4905	0.889
0.2803	14.0	3500	0.5018	0.8925
0.2739	15.0	3750	0.5126	0.8915
0.2501	16.0	4000	0.4974	0.8955
0.2382	17.0	4250	0.4936	0.891
0.2241	18.0	4500	0.4798	0.896
0.2106	19.0	4750	0.5011	0.8915
0.2068	20.0	5000	0.4844	0.899
0.1982	21.0	5250	0.4988	0.8915
0.1857	22.0	5500	0.4857	0.894
0.1762	23.0	5750	0.4855	0.893
0.1798	24.0	6000	0.4832	0.893
0.1605	25.0	6250	0.4979	0.896

Framework versions

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