arielcerdap/wav2vec2-timestamped-stutter

Wav2Vec2 TimeStamped Stutter - Ariel Cerda

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the TimeStamped dataset. It achieves the following results on the evaluation set:

• Loss: nan
• Wer: 0.9992

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
11.5515	1.1696	100	11.9112	1.0001
13.4395	2.3392	200	11.9112	1.0001
11.9169	3.5088	300	11.9112	1.0001
11.1801	4.6784	400	11.9112	1.0001
15.3393	5.8480	500	11.9112	1.0001
11.8177	7.0175	600	11.9112	1.0001
10.7941	8.1871	700	11.9112	1.0001
13.8075	9.3567	800	11.9112	1.0001
11.5275	10.5263	900	11.9112	1.0001
10.1228	11.6959	1000	11.9112	1.0001
13.7071	12.8655	1100	11.9112	1.0001
11.6933	14.0351	1200	11.9112	1.0001
10.885	15.2047	1300	11.9112	1.0001
14.2349	16.3743	1400	11.9112	1.0001
12.8886	17.5439	1500	11.9112	1.0001
0.0	18.7135	1600	nan	0.9992
0.0	19.8830	1700	nan	0.9992
0.0	21.0526	1800	nan	0.9992
0.0	22.2222	1900	nan	0.9992
0.0	23.3918	2000	nan	0.9992
0.0	24.5614	2100	nan	0.9992
0.0	25.7310	2200	nan	0.9992
0.0	26.9006	2300	nan	0.9992
0.0	28.0702	2400	nan	0.9992
0.0	29.2398	2500	nan	0.9992

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.19.1