Skip to content

yaraksen/dla_asr

Repository files navigation

ASR project

Aksenov Yaroslav

Installation guide

Download model to the folder final_model

Create folder data, path to the dataset should be data/datasets/librispeech

pip install -r ./requirements.txt

Launching guide

Testing:

python test.py \
   -c final_model/default_test_config.json \
   -r final_model/model_best.pth \
   -t test_data \
   -o test_result.json

Training step 1:

python train.py \
   -c hw_asr/train460_config.json \
   -wk "YOUR_WANDB_API_KEY"

Training step 2:

python train.py \
   -c hw_asr/train500_config.json \
   -wk "YOUR_WANDB_API_KEY" \
   -p "PATH TO THE LATEST CHECKPOINT FROM TRAIN 1"

Results for LibriSpeech test-clean, test-other are available for pruned and original language models.

About

Audio Speech Recognition project

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published