metadata

language:
  - el
license: apache-2.0
tags:
  - automatic-speech-recognition
  - mozilla-foundation/common_voice_7_0
  - generated_from_trainer
datasets:
  - common_voice
model-index:
  - name: wav2vec2-large-xls-r-300m-greek
    results: []

wav2vec2-large-xls-r-300m-greek

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_7_0 - EL dataset. It achieves the following results on the evaluation set:

Loss: 0.6831
Wer: 0.4287

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: 32
eval_batch_size: 32
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 100.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.0798	4.42	500	3.0010	1.0012
1.4336	8.85	1000	0.8481	0.6911
1.2062	13.27	1500	0.7312	0.6333
1.0481	17.7	2000	0.6850	0.5359
0.9837	22.12	2500	0.6337	0.5316
0.9108	26.55	3000	0.6258	0.5079
0.8439	30.97	3500	0.6301	0.4888
0.7901	35.4	4000	0.6245	0.4977
0.7669	39.82	4500	0.6164	0.4672
0.7196	44.25	5000	0.6039	0.4688
0.6715	48.67	5500	0.5900	0.4573
0.6441	53.1	6000	0.7002	0.4798
0.5938	57.52	6500	0.6249	0.4579
0.5541	61.95	7000	0.6184	0.4425
0.5506	66.37	7500	0.6963	0.4585
0.4998	70.8	8000	0.6778	0.4468
0.4729	75.22	8500	0.6383	0.4393
0.4535	79.65	9000	0.6593	0.4369
0.4358	84.07	9500	0.6914	0.4422
0.402	88.5	10000	0.6744	0.4269
0.3946	92.92	10500	0.6895	0.4275
0.3734	97.35	11000	0.6889	0.4320

Framework versions

Transformers 4.16.0.dev0
Pytorch 1.10.1+cu102
Datasets 1.17.1.dev0
Tokenizers 0.11.0