Skip to content

Latest commit

 

History

History
39 lines (29 loc) · 1.39 KB

README.md

File metadata and controls

39 lines (29 loc) · 1.39 KB
Raw

Karya-MSRI@AmericasNLP

Environment Specifications

Install using asr_requirements.

pip3 install -r asr_requirements.txt

Scripts

├── src
│   ├── asr.py                          # Fine-tuning ASR models & utilizing KenLM during training 
│   ├── construct_kenlm.py              # Constructing kenlm data + training kenlm models 
│   ├── debug                           # WIP
│   │   └── asr_with_kenlm.py
│   ├── inference.py                    # Inference for ASR Models (KenLM support to be added) 


├── scripts
│   ├── automate.sh                     #Automating the training of the asr models 
│   ├── inference.sh                    #Automating the inference of the asr models
│   ├── train_kenlm.sh                  #Automating the training of the kenlm models

Models and Curves:

  • Baseline models and loss curves can be found here.
  • Training Curves (and hyperparameter setups) for the best performing models on the dev set can be found here
  • Models for use with the HuggingFace API can be found here.

Language Codes:

  • Bribri - bzd
  • Guarani - gn
  • Kotiria - gvc
  • Quechua - quy
  • Wa'ikhana - tav