metadata

library_name: transformers
license: apache-2.0
base_model: facebook/wav2vec2-xls-r-1b
tags:
  - automatic-speech-recognition
  - natbed
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: xls-r-1b-bem-natbed-combined-model
    results: []

xls-r-1b-bem-natbed-combined-model

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the NATBED - BEM dataset. It achieves the following results on the evaluation set:

Loss: 0.7801
Wer: 0.7879

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: 4
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 30.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.1252	100	1.3725	0.9805
No log	0.2505	200	0.9491	0.8541
No log	0.3757	300	0.9524	0.8137
No log	0.5009	400	1.0355	0.9016
1.7945	0.6262	500	0.9611	0.8788
1.7945	0.7514	600	0.9790	0.8642
1.7945	0.8766	700	0.9877	0.8602
1.7945	1.0019	800	0.9604	0.8925
1.7945	1.1271	900	0.8880	0.8328
0.9885	1.2523	1000	0.8917	0.8368
0.9885	1.3776	1100	0.9034	0.8306
0.9885	1.5028	1200	0.8478	0.7938
0.9885	1.6281	1300	0.8666	0.8628
0.9885	1.7533	1400	0.8331	0.8218
0.8854	1.8785	1500	0.8405	0.8045
0.8854	2.0038	1600	0.7801	0.7879
0.8854	2.1290	1700	0.8305	0.7917
0.8854	2.2542	1800	0.7972	0.7911
0.8854	2.3795	1900	0.7916	0.7758

Framework versions

Transformers 4.46.0.dev0
Pytorch 2.4.1+cu121
Datasets 3.0.1
Tokenizers 0.20.0