README.md

metadata

license: apache-2.0
tags:
  - automatic-speech-recognition
  - experiments/data/uwb_atcc/train
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: 0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc
    results: []

0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc

This model is a fine-tuned version of facebook/wav2vec2-large-960h-lv60-self on the EXPERIMENTS/DATA/UWB_ATCC/TRAIN - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.7287
• Wer: 0.1756

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: 0.0001
• train_batch_size: 24
• eval_batch_size: 12
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.06	500	2.9016	0.9995
2.877	2.12	1000	0.9812	0.3485
2.877	3.18	1500	0.7842	0.2732
0.7834	4.25	2000	0.6962	0.2192
0.7834	5.31	2500	0.6527	0.2042
0.6084	6.37	3000	0.6220	0.1972
0.6084	7.43	3500	0.6442	0.1934
0.5147	8.49	4000	0.6793	0.1950
0.5147	9.55	4500	0.6432	0.1920
0.4566	10.62	5000	0.6605	0.1853
0.4566	11.68	5500	0.6393	0.1866
0.4155	12.74	6000	0.6918	0.1803
0.4155	13.8	6500	0.6514	0.1791
0.372	14.86	7000	0.7010	0.1851
0.372	15.92	7500	0.6824	0.1786
0.3368	16.99	8000	0.6895	0.1780
0.3368	18.05	8500	0.7150	0.1759
0.3244	19.11	9000	0.7141	0.1759
0.3244	20.17	9500	0.7225	0.1756
0.2981	21.23	10000	0.7287	0.1756

Framework versions

• Transformers 4.24.0
• Pytorch 1.13.0+cu117
• Datasets 2.6.1
• Tokenizers 0.13.2