HO_ASR-Model_KIIT2025

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6844
• Wer: 0.5516

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: 32
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.9609	2.0	250	2.8620	0.9958
2.3792	4.0	500	1.5657	0.8976
0.8142	6.0	750	0.8104	0.7383
0.5211	8.0	1000	0.6461	0.6508
0.4145	10.0	1250	0.5793	0.6257
0.3562	12.0	1500	0.5991	0.6315
0.3135	14.0	1750	0.5680	0.6295
0.2694	16.0	2000	0.5731	0.6139
0.2333	18.0	2250	0.6170	0.6482
0.2061	20.0	2500	0.5771	0.5852
0.1823	22.0	2750	0.5820	0.5776
0.1628	24.0	3000	0.5853	0.5793
0.1434	26.0	3250	0.6188	0.5776
0.129	28.0	3500	0.6095	0.5644
0.1185	30.0	3750	0.6210	0.5753
0.1071	32.0	4000	0.6250	0.5680
0.097	34.0	4250	0.6207	0.5636
0.0901	36.0	4500	0.6477	0.5760
0.0833	38.0	4750	0.6510	0.5666
0.0758	40.0	5000	0.6519	0.5553
0.0693	42.0	5250	0.6641	0.5549
0.0637	44.0	5500	0.6648	0.5490
0.0591	46.0	5750	0.6809	0.5535
0.0585	48.0	6000	0.6786	0.5512
0.0555	50.0	6250	0.6844	0.5516

Framework versions

• Transformers 4.28.0
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.13.3