distilhubert-finetuned-banglabeats

This model is a fine-tuned version of ntu-spml/distilhubert on the BanglaBeats dataset. It achieves the following results on the evaluation set:

• Loss: 1.4126
• Accuracy: 0.8336

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: 6e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.9439	1.0	910	0.9274	0.6425
0.854	2.0	1820	0.7498	0.7260
0.4835	3.0	2730	0.6329	0.7706
0.6226	4.0	3640	0.6159	0.7934
0.456	5.0	4550	0.7118	0.7972
0.0565	6.0	5460	0.7994	0.8052
0.2605	7.0	6370	0.9735	0.8151
0.3635	8.0	7280	1.0618	0.8244
0.1879	9.0	8190	1.1644	0.8213
0.0292	10.0	9100	1.2543	0.8194
0.0002	11.0	10010	1.4084	0.8101
0.0006	12.0	10920	1.3823	0.8132
0.088	13.0	11830	1.4016	0.8256
0.0381	14.0	12740	1.3587	0.8225
0.0	15.0	13650	1.4242	0.8169
0.0	16.0	14560	1.4053	0.8275
0.0183	17.0	15470	1.4357	0.8318
0.0	18.0	16380	1.4123	0.8306
0.0098	19.0	17290	1.4077	0.8330
0.0	20.0	18200	1.4126	0.8336

Framework versions

• Transformers 4.33.0
• Pytorch 2.0.0
• Datasets 2.1.0
• Tokenizers 0.13.3