STT_Model_8

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

• Loss: 0.5858
• Wer: 0.3549

Model description

More information needed

Intended uses & limitations

More information needed

Dataset info

• Name: LJSpeech
• Source: https://www.kaggle.com/datasets/mathurinache/the-lj-speech-dataset
• Total audios (in Google Drive): 1420
• Total transcripts (in Google Drive): 13100
• No. of rows selected: 100
• Train-test ratio: 80:20
• No. of training set: 80
• No. of testing set: 20

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	20.0	200	2.9443	1.0
No log	40.0	400	2.8603	1.0
3.8362	60.0	600	0.5940	0.4197
3.8362	80.0	800	0.5702	0.3380
0.2307	100.0	1000	0.5858	0.3549

Framework versions

• Transformers 4.26.0
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2