wav2vec2-in-hindi

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6122
• Wer: 1.0343

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: 50
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.586	3.17	100	3.4128	1.0
3.4004	6.35	200	3.3716	1.0
3.3097	9.52	300	3.0612	1.0
1.903	12.7	400	0.8674	1.0476
0.636	15.87	500	0.6154	1.0383
0.3542	19.05	600	0.6023	1.0283
0.2611	22.22	700	0.6086	1.0283
0.2123	25.4	800	0.6135	1.0296
0.1729	28.57	900	0.6221	1.0231
0.1543	31.75	1000	0.6063	1.0356
0.1386	34.92	1100	0.6090	1.0258
0.1248	38.1	1200	0.6122	1.0343

Framework versions

• Transformers 4.33.0
• Pytorch 2.0.0
• Datasets 2.1.0
• Tokenizers 0.13.3