- added support for HowTo100M dataset
2.0 - 16 July 2021
Changes related to the ZeroSpeech challenge:
- added support for SpokenCOCO dataset
- added code to support the use of pretrained features + utility script to extract CPC features
- refactored tokenization helpers making the tokenizer a global variable of dataset.py
- changed platalea default config path ~/.platalea -> ~/.config/platalea
- disabled use of wandb by default in basic.py and transformer.py experiments
- pinning down pytorch version
Resolves issues #53, #103, #104 and (temporarily) solves #116.
1.0 - 9 December 2020
- Introducing an attention-based encoder-decoder architecture for speech recognition.
- Multitask training with multiple objectives (e.g. cross-modality retrieval and speech transcription) is also possible now.
0.9 - 20 January 2020
State of the repo before @bhigy's merge leading to version 1.0.