facebookresearch/fairseq

Request to provide torchscript-ed model

Open

#4,853 opened on Nov 9, 2022

View on GitHub
 (0 comments) (1 reaction) (0 assignees)Python (6,224 forks)batch import
enhancementhelp wantedneeds triage

Repository metrics

Stars
 (29,107 stars)
PR merge metrics
 (No merged PRs in 30d)

Description

🚀 Feature Request

https://huggingface.co/facebook/wav2vec2-base-960h/tree/main provides only model.state_dict(). Please also consider upload a torchscript-ed model using either torch.jit.script() or torch.jit.trace().

Motivation

It simplifies users' life as they don't need to install fairseq to use a torchscript-ed model

Additional context

We want to test models trained using CTC loss from various frameworks using k2 for decoding. (See https://github.com/k2-fsa/k2/pull/1096)

We only need a torchscript-ed model but for the current situation, we have to install fairseq and export the pretrained model by ourselves via torchscript.

Contributor guide

Research direction
Explore the fairseq codebase to understand how to instantiate the wav2vec2 model and load the pre trained checkpoint. Study the PyTorch TorchScript documentation to learn how to script or trace the model. Attempt to export the model and verify its output matches the original. Finally, review the Hugging Face model upload process.
Tech stack
pythonpytorch
Domain
machine learningai
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
2
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
PythonPyTorchGit
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
80