question-mark
Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Support for Comet Logger

See original GitHub issue

🚀 Feature Request

I would like to add support for a Comet ML Logger to Catalyst

Motivation

Comet provides a suite of logging and visualization options that makes it easy for practitioners to analyze, visualize and share the results of their model development process. I think it would work very well with Catalyst’s focus on reproducibility, rapid experimentation, and codebase reuse.

Proposal

The integration would follow a similar pattern to previous ones such as MLFlow, Neptune, and WandB. The workflow would look something like

import comet_ml
import torch
from torch.utils.data import DataLoader, TensorDataset
from catalyst import dl
from catalyst.loggers.comet import CometLogger

# data
num_samples, num_features = int(1e4), int(1e1)
X, y = torch.rand(num_samples, num_features), torch.rand(num_samples)
dataset = TensorDataset(X, y)
loader = DataLoader(dataset, batch_size=32, num_workers=1)
loaders = {"train": loader, "valid": loader}

# model, criterion, optimizer, scheduler
model = torch.nn.Linear(num_features, 1)
criterion = torch.nn.MSELoss()
optimizer = torch.optim.Adam(model.parameters())
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, [3, 6])
logger = CometLogger()

# model training
runner = dl.SupervisedRunner()
runner.train(
    model=model,
    criterion=criterion,
    optimizer=optimizer,
    scheduler=scheduler,
    loaders=loaders,
    logdir="./logdir",
    valid_loader="valid",
    valid_metric="loss",
    minimize_valid_metric=True,
    num_epochs=8,
    verbose=True,
    loggers={"comet": logger},
)

Alternatives

Did not consider alternatives for this specific proposal, as this enhancement is to help Comet users log data from Catalyst.

Additional context

Checklist

  • feature proposal description
  • motivation
  • extra proposal context / proposal alternatives review

FAQ

Please review the FAQ before submitting an issue:

Issue Analytics

  • State:closed
  • Created 2 years ago
  • Reactions:1
  • Comments:6 (5 by maintainers)

github_iconTop GitHub Comments

1reaction
DN6commented, Aug 24, 2021

Hi, @Scitator, one of our engineers has something in the works. We’re just testing and running through the Catalyst checklist for creating a PR to make sure we don’t miss anything. We can open a PR within the next couple of days if that’s okay?

0reactions
Scitatorcommented, Sep 2, 2021

Yup, that would be okay, thanks!

Read more comments on GitHub >

github_iconTop Results From Across the Web

Support
SERVICE AND TECHNICAL SUPPORT ... E-mail: info@cometsystem.com ... data loggers, monitoring systems, data acquisiton systems, thermometers, hygrometers, ...
Read more >
Technical support, FAQ - comet-sensors.ch
Comet - manufacturer of data loggers, temperature recorder, temperature logger, ... I need to send data logger to service, should I disconnect all...
Read more >
Comet ML - Build better models faster
Track, compare, and reproduce your ML experiments with Comet's machine learning platform. Leverage insights to build better models, faster.
Read more >
Troubleshooting | Comet Documentation
in Comet Server Service Manager, use the Service menu > "Browse log files" option, ... Please contact Comet Support with a screenshot of...
Read more >
CometLogger ExistingOfflineExperiment support #12829
We should be able to reload existing offline experiments to perform new logging (e.g. log new metrics on different datasets). Motivation. I need...
Read more >

github_iconTop Related Medium Post

No results found

github_iconTop Related StackOverflow Question

No results found

github_iconTroubleshoot Live Code

Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start Free

github_iconTop Related Reddit Thread

No results found

github_iconTop Related Hackernoon Post

No results found

github_iconTop Related Tweet

No results found

github_iconTop Related Dev.to Post

No results found

github_iconTop Related Hashnode Post

No results found