Just amazing !!! Hard to find such good content around MlOps without reading the same story 😉
@dvcorg8370
3 жыл бұрын
@Justine. Be sure to check out our Office Hours meetups live or the recordings to see how our users are putting our tools to use!
@mraarone
3 жыл бұрын
This is crazy, I was searching the Internet for Swiss farmers' attitudes towards global climate change and I ran across this video. Of course, I had to watch it to find out my answer, but it just so happens, I can totally use DVC for my DevSecOps / MLOps architectures and designs. WHO WOULD HAVE KNOWN!? Thank you for both answers in one video, Elle! -- Sincerely, A-a-ron
@dmitrypetrov3542
3 жыл бұрын
Thanks Aaron! I hope CML will help us with climate change as well (but I don’t know how exactly) 😀
@mraarone
3 жыл бұрын
@@dmitrypetrov3542 I didn’t realize you guys had GPU integration tools for gitlab and github runners in the CML image. I literally woke up having dreamt about that and searched the web for it, Elle’s video popped up. You guys are doing amazing work, those images aren’t easy to build. I think I may want to get involved with you guys. Are there any opportunities to? Also, do you have support for docker within docker (e.g., volume mount docker.sock, etc.), that would be really helpful to run docker-compose set ups?
@dmitrypetrov3542
3 жыл бұрын
@@mraarone yeah, this approach is quite flexible :) Happy to talk about possible collaboration - please shoot me an email at my Firstname at iterative.ai Docker within docker is possible but you might need to install CML as JS lib, not docker. You can ask for more details in our Discord channel dvc.org/chat
@upendra8050
4 жыл бұрын
Your content is so unique and different from others. I have learned a ton and already applying them to my own work after only watching 3 videos. Can't wait to check out other videos
@dvcorg8370
4 жыл бұрын
Thanks Upendra! That's so kind of you :)
@awds121
6 ай бұрын
Thanks for the tutorial! If you want this to work today, make these changes in your train.yaml file: change "--show-md" -> "--md" change "cml-send-comment" -> "cml comment create" add "permissions: write-all" at the same level before "runs-on: [ubuntu-latest] add " git config --global --add safe.directory '*' " after "dvc repro" Hope this helps!
@dvcorg8370
6 ай бұрын
Thanks for the tip!
@ricardovaldez8997
4 жыл бұрын
Pretty cool, can't wait for the next tutorial
@dvcorg8370
4 жыл бұрын
Thanks Ricardo!
@kvishal001
2 жыл бұрын
Very crystal clear presentation, right from content to diction.
@RobertKwapich
3 жыл бұрын
Thanks, Elle! Really informative, as well as great overall presentation! Keep up the good work!
@katerynasoroka8149
3 жыл бұрын
thanks for the tutorials, I like your style of teaching!
@gurubhandari
4 жыл бұрын
I have basic github not paid one, does that support all the process and execution of docker?
@dvcorg8370
4 жыл бұрын
Yes!
@burnessduan
3 жыл бұрын
It's a nice tutoiral, I will do some practices in my team
@highihiggins
3 жыл бұрын
Pretty cool video, currently following along to see what changed from when I started using DVC about 1 1/2 years ago! You just didn't get your fancy header for your image into the .md file, you just echoed it. But that's what happens with live demos, right? ;) Keep up the good work!
@waltercontacto8732
2 жыл бұрын
Thanks, Elle!... and Deevee-
@alexandrevalente9994
2 жыл бұрын
Even Steve Vai can't compete with you ... you play so fast! But ... where is your cup of coffee?
@santiagobassani2973
4 жыл бұрын
seriously, awesome videos! just what I needed. Just a simple question, is it possible to use DVC with an S3 private bucket and git workflows?
@dvcorg8370
4 жыл бұрын
Definitely! DVC with an S3 bucket is one of the most popular setups. We have some docs here: dvc.org/doc/use-cases/sharing-data-and-model-files#create-an-s3-bucket As always, if you hit any issues we have a support Discord! discordapp.com/invite/dvwXA2N
@5swaroop
2 жыл бұрын
Great and informative content. However two models can be compared only if either the underlying data is same or atleast on unseen data compare their performance.
@RedShipsofSpainAgain
4 жыл бұрын
Hi Elle, would you use this to perform A/B testing between, say, the Master branch and the feature branch?
@dvcorg8370
4 жыл бұрын
Hm, do you mean A/B testing of deployed models? We haven't tried that yet- the use cases we've built so far are focusing on development more than deployment. Feel free to tell us more about what you have in mind!
@gurubhandari
4 жыл бұрын
could you please check whether 'dvcorg/cml-py3:ubuntu-latest' changed into 'dvcorg/cml-py3:latest' or not ?
@dvcorg8370
4 жыл бұрын
You should be using `dvcorg/cml-py3:latest`!
@CristianGarcia
4 жыл бұрын
I don't think performing the training on the containers provided by github actions is very scalable, what are actually the best practices?
@dvcorg8370
4 жыл бұрын
do you mean the runners provided by GH? We're using a custom container. If you mean runners, there are indeed many use cases where you'll want your own compute resources (i.e. EC2 instances, maybe GPUs). It's straightforward to setup a self-hosted runner for using your own resources for running workflows- GH docs are here: docs.github.com/en/actions/hosting-your-own-runners
@CristianGarcia
4 жыл бұрын
@@dvcorg8370 Thanks for the info! I've been using standard runners which are small machines, I didn't know about self hosted runners.
@dvcorg8370
4 жыл бұрын
@@CristianGarcia Yes, the standard runners will cover a lot of use cases but when it comes to deep learning or memory-intensive models it is worth it to explore self-hosting. We have a tutorial about setting this up for GPUs coming up very shortly!
@pmg308
3 жыл бұрын
Super nice series! Congratulations and keep up the good work! Looking forward to go through the rest of the tutorials :) I've been following your videos and deploying on bitbucket mostly without any hiccups :) Have a question though: it there a particular reason for "dvc params diff --all --show-md master >> report.md" raising an error in different branches? (ERROR: failed to show params diff - unknown Git revision 'master') I've been looking through the docs but nothing refers to this.. It only executes with changes on the master branch. Any thoughts? Thanks!
@congnguyen-fo6bk
3 жыл бұрын
check if you're on master branch. one possibility is you may have named your branch 'main' instead of 'master'
@pmg308
3 жыл бұрын
@@congnguyen-fo6bk thanks for the feedback! really appreciate it! I've tried in both master and other branches - none seems to be able to complete the pipeline. However, they execute their pipelines differently (on the master branch everything runs, except the cml-comment - which makes sense, the others break before that). The name of the master branch is "master" as well. The repo is public (bitbucket.org/paulo308/cml_dvc_redemo/src/master/). Any hints, from your experience?
@sourabhyadav5716
3 жыл бұрын
Can we do versioning for Image and Video datasets using DVC? If so, please point out or make a tutorial video for the same. Thanks
@brahimimohamed261
3 жыл бұрын
Good question, I'm interested about the response
@dvcorg8370
3 жыл бұрын
Hi @Brahimi! Apologies for the delayed response! A tutorial with an image dataset versioning dvc.org/doc/use-cases/versioning-data-and-model-files/tutorial !
@MuhammadSaad-kb7op
Жыл бұрын
Amazing! Good explanation
@dvcorg8370
Жыл бұрын
Glad you liked it!
@curdyco
7 ай бұрын
these things are easy , Are there any tutorials about deploying deep learning models with large datasets with retraning with feedback on custom host like a kaggle notebook?
@dvcorg8370
7 ай бұрын
We see you found Alex Kim's Best MLOps Practices for Building End-to-End Machine Learning Computer Vision Projects. You're in good hands there. In general, using notebooks for production environments is generally considered not the best practice (outside resource: kzitem.info/news/bejne/lHyXnKmwgaqFnoY). There are tools like jupy text that can help with versioning and getting proper diffs in your notebooks, but still not the best for real-world ops.
@curdyco
7 ай бұрын
@@dvcorg8370 but Google colab is cost friendly if I need GPU access for training, using local PC is not worth it because of no GPU and sagemaker kind of services burn a hole in the pocket. What is the alternative according to you?
@nisaratutube
4 жыл бұрын
Pretty cool, can't wait for the next tutorial. Please have one tutorial where S3 is also integrated into the pipeline. If possible make this tutorial having pytorch and NLP .
@dvcorg8370
4 жыл бұрын
Good idea- we have some S3 stuff planned, and can definitely work in some NLP :)
@h3techsme
4 жыл бұрын
Good Q - I spent a minute looking at the Use Case tabs on the CML splash page which leaves some interesting breadcrumbs - with a tensor and EC2 "Cloud Computing" case shown.
@dvcorg8370
4 жыл бұрын
@@h3techsme thanks for checking out the site! we're pretty excited to share more about using EC2 instances and other cloud resources in some upcoming videos.
@jtauber
4 жыл бұрын
I would have listened to the entire guitar performance :-)
@aldemarhernadezg
3 жыл бұрын
I fell in love, Elle, you are the most incredibly gorgeous data scientist that I ever saw. BTW DVC rules, great tool.
Пікірлер: 59