wav2vec2-large-xlsr-53-AsanteTwi-06

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6122
• Wer: 0.5

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: 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: 200
• num_epochs: 300

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
9.3303	16.67	100	5.2842	1.0
2.961	33.33	200	3.1857	1.0
2.8758	50.0	300	2.9988	1.0
2.8331	66.67	400	2.8830	1.0
2.4893	83.33	500	2.1638	1.0
1.1901	100.0	600	0.7611	0.5625
0.5563	116.67	700	0.7503	0.5
0.3916	133.33	800	0.6324	0.5
0.288	150.0	900	0.8291	0.5
0.2176	166.67	1000	0.7383	0.5625
0.1814	183.33	1100	0.6408	0.5
0.1749	200.0	1200	0.5769	0.5625
0.1653	216.67	1300	0.6512	0.5
0.1301	233.33	1400	0.6414	0.4375
0.1375	250.0	1500	0.5970	0.5
0.1173	266.67	1600	0.6119	0.5
0.108	283.33	1700	0.6325	0.5
0.1183	300.0	1800	0.6122	0.5

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu118
• Datasets 2.13.1
• Tokenizers 0.13.3