metadata

license: apache-2.0
tags:
  - generated_from_trainer
datasets:
  - common_voice
model-index:
  - name: wav2vec2-large-xls-r-300m-turkish-colab_common_voice-8_4
    results: []

wav2vec2-large-xls-r-300m-turkish-colab_common_voice-8_4

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:

Loss: 0.3201
Wer: 0.3295

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: 11
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.9268	0.51	400	1.3204	0.9175
0.7491	1.02	800	0.5880	0.6388
0.4911	1.53	1200	0.4680	0.5613
0.4265	2.04	1600	0.4213	0.5059
0.3473	2.55	2000	0.4199	0.4955
0.3291	3.07	2400	0.4323	0.5061
0.2819	3.58	2800	0.4026	0.4490
0.2628	4.09	3200	0.3831	0.4446
0.2371	4.6	3600	0.3622	0.4234
0.2274	5.11	4000	0.3473	0.4012
0.2051	5.62	4400	0.3471	0.3998
0.1985	6.13	4800	0.3759	0.4088
0.1767	6.64	5200	0.3620	0.4012
0.1707	7.15	5600	0.3415	0.3700
0.1559	7.66	6000	0.3317	0.3661
0.147	8.17	6400	0.3265	0.3618
0.1339	8.68	6800	0.3293	0.3586
0.126	9.2	7200	0.3386	0.3458
0.1149	9.71	7600	0.3305	0.3397
0.1051	10.22	8000	0.3235	0.3354
0.1005	10.73	8400	0.3201	0.3295

Framework versions

Transformers 4.11.3
Pytorch 1.10.0+cu113
Datasets 2.1.0
Tokenizers 0.10.3