Slug size error in Heroku when deploying Streamlit app

Deployment
1

Hi guys. I am not sure if it is a right forum to ask this but still ,
when I try to deploy Streamlit app in to Heroku cloud server , I am getting error as below , not sure how to fix this ,any help would be highly appreciated .

P.S - thanks Streamlit team for making this wonderful library .

tensorboard-2.1.1 tensorflow-2.1.0rc2 tensorflow-estimator-2.1.0 termcolor-1.1.0 terminado-0.8.3 testpath-0.4.4 toml-0.10.1 toolz-0.10.0 tornado-6.0.4 traitlets-4.3.3 tzlocal-2.1 urllib3-1.25.10 validators-0.17.1 watchdog-0.10.3 wcwidth-0.2.5 webencodings-0.5.1 werkzeug-1.0.1 widgetsnbextension-3.5.1 wrapt-1.12.1 zipp-3.1.0
-----> Discovering process types
Procfile declares types -> web
-----> Compressing…
! Compiled slug size: 590.4M is too large (max is 500M).
** ! See: http://devcenter.heroku.com/articles/slug-size**
! Push faile

1 Like
2

Looks like the total package is too large, a combination of packages, data and code, do you have something like a large model file committed?

2 Likes
3

@Ian_Calvert only the model file(.h5) has file size around 20MB , do you thinks that’s what causing an issue ?

I also tried first step mentioned in that link , still facing same issue

4

Potentially the dependencies are too large, but it’s worth checking how big your git repo is (just the folder locally). That 20M file if there’s a few updates in the history could take up a lot if the whole history is in the slug.

Essentially once it’s built your app the total is too large. Any dependencies you don’t need?

5

Looking at deploying tensorflow apparently that itself is pretty huge. Tensorflow 1 is smaller, or you otherwise may need to find alternative hosting.

1 Like
6

Thanks @Ian_Calvert , I think the same , in AWS EC2(free tier) also I faced the similar memory issue while Installing TensorFlow .

Can you please suggest any free hosting service like Heroku which provides easy deployment support for steamilt application ?

7

I’m not sure of a free one, it could be worth looking at other vps/similar options. A $5/mo instance from digital ocean is pretty capable and there are a few different providers like that

1 Like
8

Thanks @Ian_Calvert , Now I am trying with digital Ocean vps .

Just FYI , there is no clear documentation available online about deploying Streamlit app in DigitalOcean vps . Would be really helpful for beginners like me if your team can add steps to deploy Streamlit app in Digital ocean .

Thanks !!

9

@Ian_Calvert , I followed the below tutorial about deploying Django application into DigitalOcean , so far I have cloned the GitHub repository in server and installed most of the required dependencies .

wondering if I need to configure Gunicorn and NGINX as mentioned in that document ??

can you please explain what should be the next step to get my app deployed ?

10

@arun_ramji on Digital Ocean you can try CapRover (1-click in marketplace)
CapRover it’s similar to Heroku.

11

@Mauro_Galafassi thanks ! can you please post the steps to deploy Streamlit app using CapRover .

@Adrien_Treuille Wouldn’t it be great if Streamlit provides VPS to deploy the Streamlit app exclusively in single click ??

Thanks !

12

Hi @arun_ramji.
I apologize in advance if I’m not clear in all the steps.

  1. install caprover CLI in your local computer (https://caprover.com/docs/get-started.html step 3)
  2. You prepare your streamlit app to deploy with a docker container (https://www.youtube.com/watch?v=doCia_CKcko)
  3. Add file “captain-definition” to app folder (https://caprover.com/docs/captain-definition-file.html). In my case contents of the file is:
    {
    “schemaVersion”: 2,
    “dockerfilePath”: “./Dockerfile”
    }
  4. On caprover server create a new app
  5. In app folder run “caprover deploy” and follow the instructions
  6. in setting app on caprover server (tab http settings) set parameter:
  • http container port: the same port where run streamlit app.
  • And flag websocket support.

I hope I have been clear enough

3 Likes
13

@arun_ramji were u able some alternative(free service) to Heroku??