xlsr-normal-nmcpc-ag

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0007
• Wer: 0.2617

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	3.6364	100	3.1279	1.0
4.7648	7.2727	200	3.0464	1.0
4.7648	10.9091	300	2.8992	1.0
2.8956	14.5455	400	2.4458	1.0
2.8956	18.1818	500	1.6426	0.9574
1.8653	21.8182	600	0.5722	0.6574
1.8653	25.4545	700	0.2484	0.4660
0.5156	29.0909	800	0.1107	0.3745
0.5156	32.7273	900	0.0952	0.3702
0.2353	36.3636	1000	0.0441	0.3149
0.2353	40.0	1100	0.0408	0.3064
0.1432	43.6364	1200	0.0209	0.2872
0.1432	47.2727	1300	0.0258	0.2894
0.0893	50.9091	1400	0.0156	0.2915
0.0893	54.5455	1500	0.0111	0.2745
0.067	58.1818	1600	0.0086	0.2723
0.067	61.8182	1700	0.0039	0.2681
0.0443	65.4545	1800	0.0024	0.2660
0.0443	69.0909	1900	0.0036	0.2638
0.0403	72.7273	2000	0.0026	0.2596
0.0403	76.3636	2100	0.0013	0.2617
0.0334	80.0	2200	0.0012	0.2638
0.0334	83.6364	2300	0.0008	0.2638
0.0238	87.2727	2400	0.0008	0.2617
0.0238	90.9091	2500	0.0007	0.2617
0.0209	94.5455	2600	0.0007	0.2596
0.0209	98.1818	2700	0.0007	0.2617

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0