Tags: SeanNaren/deepspeech.pytorch
Tags
Pytorch Lightning Integration (#569) * Added minimal code to integrate Pytorch Lightning into training.py * Added autocast support, removed intra epoch checkpointing for simplicity, integrated checkpoint support, fixed validation support * Fixed multi-gpu support * Fixed smoke test, pretrained tests will be broken till new model release Added trains viz logging Precision * Updated README, fixed server class, updated k8s config file, added fix for adam Trains support, removed autocast since this is handled via lightning * Swapped to using tqdm write for readability when checkpointing, added an4 config * Added base script for each dataset, updated default params * Swapped to using native CTC, updated common voice script, removed incorrect lightning version * Updated cv params and output manifest location, set default epochs to the epochs used for previous release * Disable trains logger for now, simplified checkpointing logic for new release * Added new metrics class, removed save_output/verbose for now, using new ModelCheckpoint class for model saving * multiprocess duration collection for speed, allow loading from file path, refactor path name and test * Swap to latest release candidate, fixed flag reference * Format smoke test, update path to best save k model * Update to latest RC * Removed trains logging, rely on PL tensorboard. swap to saving json object for manifest to modify root path * Ensure abs path for manifest root path * Use absolute paths for manifest * Update requirements, abstract all PL trainer arguments * Enable checkpoint callback * Enable checkpoint callback, add verbosity * Add sharded as a dependency for better memory use * Set num workers, add spec augment * Update deepspeech_pytorch/data/utils.py Co-authored-by: Anas Abou Allaban <aabouallaban@protonmail.com> * Specify blank index explicitly * Add blank index to ctc loss * Fix CI * Fix Syntax Warning * Fix install requirements * Use torchaudio (#607) * Use torchaudio * Add torchaudio to reqs * Fixes for testing, update AN4 config, update dockerfile base image * Add noninteractive to remove stalling * revert * Update API Co-authored-by: Sean Narenthiran <sean.narenthiran@digitalreasoning.com> Co-authored-by: Anas Abou Allaban <aabouallaban@protonmail.com> Co-authored-by: Anas Abou Allaban <aabouallaban@pm.me>