Deploy a webapp from a model built on PaperSpace Gradient

Part 1 (2019)
1

I created and trained my own model based on the lesson2-download notebook of the Fast AI course.
My notebook file, model and data (images) are all stored in a PaperSpace Gradient Notebook.

I now want to create and deploy a webapp from it. Just like the Cougar-or-not of simonw.

I created a project locally on my Mac that I plan to push to Heroku or another hosting service.

For my webapp, I need my model file and all my images. I downloaded the model file from PaperSpace but no way to download en masse the folder containing all the images I used to train my model. (no options on PaperSpace)

I feel I am not doing the things correctly here. What am I missing? Can I deploy a webapp directly from PaperSpace? How can I download entire folders from my PaperSapce notebook?

Thank you in advance for your help!

2

Hi myself from 2 days ago!

Here is the answer to your question :wink:

Use PaperSpace Gradient to host and run your Notebooks but when it is about deploying an App, follow the great tutos of Fait.ai: https://course.fast.ai/deployment_google_app_engine.html
They contain boilerplates for each platform and work like a charm!

Have a great day.

2 Likes
3

Hello!

While you were running your notebooks and storing models on PaperSpace, how did you download the export.pkl file from PaperSpace to your local machine to run?

4

If anyone has this same question, you just have to go into your Jupyter environment (while using PaperSpace Gradient), and click on the export.pkl and there should be a “Download” button at the top.