distilhubert-finetuned-cry-detector

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

• Loss: 0.2255
• Accuracy: 0.9883
• F1: 0.9883
• Precision: 0.9883
• Recall: 0.9883
• Confusion Matrix: [[960, 10], [6, 389]]

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: 3e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 123
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall	Confusion Matrix
0.3124	2.3256	100	0.2739	0.9641	0.9640	0.9640	0.9641	[[948, 22], [27, 368]]
0.2337	4.6512	200	0.2385	0.9736	0.9737	0.9737	0.9736	[[950, 20], [16, 379]]
0.2064	6.9767	300	0.2295	0.9832	0.9832	0.9832	0.9832	[[958, 12], [11, 384]]
0.2023	9.3023	400	0.2277	0.9868	0.9869	0.9870	0.9868	[[957, 13], [5, 390]]
0.2003	11.6279	500	0.2254	0.9875	0.9876	0.9876	0.9875	[[960, 10], [7, 388]]
0.2002	13.9535	600	0.2259	0.9875	0.9876	0.9876	0.9875	[[959, 11], [6, 389]]
0.1994	16.2791	700	0.2255	0.9883	0.9883	0.9883	0.9883	[[960, 10], [6, 389]]
0.1997	18.6047	800	0.2254	0.9883	0.9883	0.9883	0.9883	[[960, 10], [6, 389]]

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Tokenizers 0.19.1