README.md

metadata

license: apache-2.0
base_model: facebook/wav2vec2-large-xlsr-53
tags:
  - automatic-speech-recognition
  - DewiBrynJones/banc-trawsgrifiadau-bangor-clean-with-ccv
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: wav2vec2-xlsr-53-ft-btb-ccv-cy
    results: []

wav2vec2-xlsr-53-ft-btb-ccv-cy

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the DEWIBRYNJONES/BANC-TRAWSGRIFIADAU-BANGOR-CLEAN-WITH-CCV - DEFAULT dataset. It achieves the following results on the evaluation set:

• Loss: 0.5324
• Wer: 0.4014

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: 8
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 600
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.7051	0.0321	500	1.7504	0.9570
1.0409	0.0641	1000	1.1511	0.7761
0.8183	0.0962	1500	1.0506	0.7097
0.7091	0.1283	2000	0.9421	0.6610
0.6547	0.1603	2500	0.8726	0.6128
0.6088	0.1924	3000	0.8246	0.5990
0.5781	0.2244	3500	0.8025	0.5747
0.5429	0.2565	4000	0.7360	0.5305
0.5104	0.2886	4500	0.7335	0.5394
0.501	0.3206	5000	0.6933	0.5088
0.4708	0.3527	5500	0.6770	0.5113
0.4526	0.3848	6000	0.6609	0.4806
0.4235	0.4168	6500	0.6373	0.4858
0.4032	0.4489	7000	0.6048	0.4466
0.3863	0.4810	7500	0.5946	0.4432
0.3766	0.5130	8000	0.5737	0.4298
0.3746	0.5451	8500	0.5668	0.4248
0.3586	0.5771	9000	0.5485	0.4101
0.3552	0.6092	9500	0.5378	0.4032
0.3326	0.6413	10000	0.5324	0.4014

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1