Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

CML blog post #1532

Merged
merged 5 commits into from
Jul 7, 2020
Merged

CML blog post #1532

merged 5 commits into from
Jul 7, 2020

Conversation

dmpetrov
Copy link
Member

@dmpetrov dmpetrov commented Jul 6, 2020

The release blogpost.

ToDo:

  • change the main image
  • fix the 1st report (DVC) image size

Please do not pay to much attention to the language, wording part and style. It was already polished and I'd like to keep it as is - it should be more a personal voice, not like in documentation.

@shcheklein shcheklein temporarily deployed to dvc-landing-cml-release-fgagxo July 6, 2020 09:13 Inactive
@shcheklein shcheklein temporarily deployed to dvc-landing-cml-release-fgagxo July 6, 2020 11:26 Inactive
@elleobrien
Copy link
Contributor

I still think a block of npm install commands is super weird. Can't we just link folks to the base Docker image to see how it's done?

@shcheklein shcheklein temporarily deployed to dvc-landing-cml-release-fgagxo July 6, 2020 23:08 Inactive
@elleobrien
Copy link
Contributor

Instead of starting with "1. CI/CD for machine learning problems", let's take the text that's currently in the description and make that the intro to the post. This stuff here:
Screen Shot 2020-07-06 at 4 34 50 PM

And then in the description, maybe something like:
Today we're launching Continuous Machine Learning (CML), a new open-source library for CI/CD with ML. Use it to automate parts of your development workflow, including model training and evaluation, comparing ML experiments across your project history, and monitoring changing datasets. Let's bring the power of DevOps to ML.

then is typical to have in CI/CD runners. CI/CD must be connected with cloud
computing instances or Kubernetes clusters for ML training.

## 2. CI/CD for ML is the next step for DVC team
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
## 2. CI/CD for ML is the next step for DVC team
## 2. CI/CD for ML is the next step for the DVC team

@jorgeorpinel
Copy link
Contributor

jorgeorpinel commented Jul 6, 2020

@shcheklein @rogermparent something really funny is happening in the review app. Click on any anchor (one of the headers) and the page disappears 🤔 E.g. https://dvc-landing-cml-release-fgagxo.herokuapp.com/blog/cml-release#1-cicd-for-machine-learning-problems loads a white page. I can't reproduce it in the blog on prod though.

UPDATE: It seems to be the number in the anchor link. Fails on prod as well.

- 4000+ regular users of DVC

Now that DVC is maturing, we're ready to take the next step: we want to
revolutionize ML development processes. We want ML experiments to have greater
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
revolutionize ML development processes. We want ML experiments to have greater
revolutionize the ML development processes. We want ML experiments to have greater

May need formatting

Comment on lines +55 to +56
have brought us to one conclusion again and again: CI/CD for ML is the holy
grail.
Copy link
Contributor

@jorgeorpinel jorgeorpinel Jul 7, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe link this

Suggested change
have brought us to one conclusion again and again: CI/CD for ML is the holy
grail.
have brought us to one conclusion again and again:
[CI/CD for ML](#3-continuous-machine-learning-release) is the holy grail.

so you can skip to #3 if needed? Not sure, I just think it takes a while to get to 3 where CML is introduced (but it's mentioned in the abstract so may not be a big deal).

experiments, and more time building value. The goal is to extend the amazing
results of DevOps from software development to ML and MLOps.

## 3. Continuous Machine Learning release
Copy link
Contributor

@jorgeorpinel jorgeorpinel Jul 7, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd use italics here because upon first glance, it reads like "a continuous release of machine learning"

Suggested change
## 3. Continuous Machine Learning release
## 3. _Continuous Machine Learning_ release

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Or

Suggested change
## 3. Continuous Machine Learning release
## 3. Continuous Machine Learning (CML) release


## 3. Continuous Machine Learning release

Today, we're releasing an open-source project https://CML.DEV to close the gap
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
Today, we're releasing an open-source project https://CML.DEV to close the gap
Today, we're releasing an open-source project https://cml.dev to close the gap

? Or CML.dev perhaps.

Comment on lines +96 to +97
1. Generate informative reports on every Pull/Merge Request with metrics, plots,
and hyperparameters changes.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

with metrics, plots, and hyperparameters changes

Q: Are we talking about DVC params, metrics, and plots specifically? Or other formats are supported?

If it's DVC, maybe link to those cmd refs. or even say "DVC metrics, ..."

Copy link
Contributor

@jorgeorpinel jorgeorpinel Jul 7, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If it's DVC, maybe link to those cmd refs. or even say "DVC metrics, ..."

And/or link to #4-the-relationship-between-cml-and-dvc below.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

no necessary DVC

1. Generate informative reports on every Pull/Merge Request with metrics, plots,
and hyperparameters changes.
2. Provision GPU\CPU resources from cloud service providers (**AWS, GCP, Azure,
Ali**) and deploy CI runners using **docker-machine**.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
Ali**) and deploy CI runners using **docker-machine**.
Ali**) and deploy CI runners using
[`docker-machine`](https://github.com/docker/machine).

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If the link is undesired, at least I think it goes better with backticks ` because it's a command. Or use the official name of the project, Docker Machine.

Comment on lines 100 to 101
3. Bring datasets from cloud storage to runners for model training (using
**DVC**), as well as save the resulting model in cloud storage.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Not sure here, depends on what you're trying to say but the meaning is quite different for each option:

Suggested change
3. Bring datasets from cloud storage to runners for model training (using
**DVC**), as well as save the resulting model in cloud storage.
3. Bring datasets from cloud storage to runners (using **DVC**) for model
training, as well as save the resulting model in cloud storage.

https://github.com/iterative/cml

For support and communications, the DVC Discord server is still the place to go:
https://dvc.org/chat We've made a new `#cml` channel there to discuss CML, CI/CD
Copy link
Contributor

@jorgeorpinel jorgeorpinel Jul 7, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should there be a cml.dev/chat redirect with an invite to the #cml channel on Discord? 🙂

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

well... might be later :)

https://dvc.org/chat We've made a new `#cml` channel there to discuss CML, CI/CD
for ML and other MLOps related questions.

## 5. Conclusion
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Completely up to you but are the numbers in each header useful? 🙂

I have no strong opinion at all but I think that's what's causing the anchor link bug (cc @rogermparent) — but that shouldn't prevent the author from using them if he prefers (we should fix the bug anyway).

@dmpetrov
Copy link
Member Author

dmpetrov commented Jul 7, 2020

UPDATE: It seems to be the number in the anchor link. Fails on prod as well.

@jorgeorpinel I'm removing the numbers

@shcheklein shcheklein temporarily deployed to dvc-landing-cml-release-fgagxo July 7, 2020 07:55 Inactive
@shcheklein shcheklein temporarily deployed to dvc-landing-cml-release-fgagxo July 7, 2020 07:59 Inactive
@dmpetrov dmpetrov merged commit 5e3b8fc into master Jul 7, 2020
@dmpetrov dmpetrov deleted the cml_release branch July 7, 2020 08:36
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants