I don't built the app. it's just the notebook i tried to deploy

as I dont really know how to build it at the moment

ok I see what's happening, you're executing

ploomber build

in your dockerfile. usually when deploying models, you train the model locally, generate a model file (e.g., model.pkl) and then in your dockerfile you only copy the model file.

your app is probably crashing because the

ploomber build

command takes too many resources

instead you can run

ploomber build

locally and modify your dockerfile so it only copies the output model file

I see. make sense. i tried to do the model.pkl, but don't know how to do it. i tried to look at other team's app, for some reasons, I couldn't be able to see the code. is there any other template to develop model.pkl ?

thanks a lot for your advice!

the pipeline https://github.com/zhiyiyi/stock-price-predition/blob/new_ml/pipeline.yaml looks like its missing some stuff cause I do not see the model. @Laura Gutierrez Funderburk is there any template?

Laura shared team 5's project, i tried to look at the arima_model_pkl, but the code wasn't readable. also, the app wasn't working. I reached out to Oscar about this and tried to set up a meeting with him. seemed he's a bit busy and still wait for his response at the moment.

I thought we could deployed the ml notebook instead of the app on ploomber

if there's any other template i could look into, that will be great. thanks again!

Hi @Eduardo and @Alice L For this team's model deployment the pipeline is still missing quite a bit of work both in terms of the script that generates the pkl file and the ploomber pipeline file

We deployed a simpler notebook as a voila deployment

@Alice L the workflow we would follow to keep computational costs simpler is 1. Package the code that generates the model into a .py script (the output is a pkl file with the model) 2. Add this script to pipeline.yaml file 3. Run the pipeline locally (or schedule it if you'd like to populate motherduck) 4. Generate a notebook or application that takes as input the pkl file or motherduck db instance and generates visualizations/makes predictions

@Eduardo we did manage to get a voila application deployed - it was focused on visualization Alice wanted to deploy an ML model, but the ML model work development started late into the mentorship program

We didn't have a chance to work on it during the supported time frame, Alice continued exploring independently

@Alice L given we are three weeks after the mentorship program ended, what I can do is provide articles on how to package your code into a script, how to upgrade the pipeline.yaml file, and how to call a model using the pkl file. Let me know if this works for you.

This message was deleted.

Hi team, can you help not merge the PR if it can’t pass all the checks like there is a red on the side of commit “X”

^^ Can you also help not resolve conflicts by merging branch ‘main’, just resolve conflicts (delete code) manually on local or on GitHub interface

got it. i will keep that in mind. thanks for letting us know.

This message was deleted.

Hi @Laura Gutierrez Funderburk, sorry for the delay on the video recording. Please find the attachment here. Please let me know if it miss anything else. Thank You, Eduardo and Ploomber's support and guidance!

awesome

thank you @Alice L

will schedule a post for your team's project

Thank you Laura！ sounds great！Thank YOU again for the wonderful opportunity! I've learned a lot during the mentorship program!

Thank you @Alice L This is amazing! I'm really excited to learn more now. ✅

Thank you Rudy! greatly appreciate it

archived the channel