Paperspace Gradient Subscription (better to use it or not?)

Part 1 (2018)
1

Hello Community,

Has anyone used Gradient Subscription in Papersapce?
It has a template: Fast.ai + PyTorch.
Should I use it for the course study?

Thank you and regards,
Nobu

2 Likes
2

That’s what I use and I really like it.
Although I don’t recommend using their in browser access to command line on the regular. I turn on the machine via the settings screen (via gear icon), and then SSH into it from my command line.

2 Likes
3

So, Paperspace just launched their brand new platform for exploring data, training neural networks, and running GPU compute jobs. It’s a new product and it’s called Gradient°. Cool name!

Basically, Gradient° is a 1-click Jupyter notebook backed by a powerful job runner behind the scene and a Python module (CLI tool) to orchestrate and run any code with full power of the existing Paperspace GPU cloud.

I have not really take Gradient° for a spin. Well, from the look of it, I guess their services have a lot of overlap with other offerings like Google Collab and FloydHub.

  • Gradient° vs. Google Collab
    • Gradient°
      • pros: CLI tool, flexibilities like more control of GPU, easier data management services, minimal interruption to your normal dev and deployment workflow
      • cons: GPU is not free
    • Google Collab
      • pros: free GPU, easy integration with Google services like Cloud, Drive, etc.
      • cons: limited access to GPU
  • Gradient° vs. FloydHub
    • Gradient°
      • pros: access to the underlyting GPU for user who needs more control
      • cons: pricing, back-end job runner performance
    • FloydHub
      • pros: documentations, pricing?
      • cons: usability of the CLI tool, change in your normal workflow

Interesting to see how all this pan-out eventually. Paperspace tout it as the future of cloud AI :thinking:

7 Likes
4

Thank you so much for such detailed comparison. :+1:

6

Hi,

I am a complete noob with Jupyter and using a cloud based machine. I created a Gradient notebook as instructed in paperspace help articles. What do I do after that? How do I access fast.ai lessons?

Nit

7

Does anyone have a good approach for converting from our current workflow taught in the course to being able to dispatch jobs via the job runner in gradient on Paperspace?

8

@nitrusty Here’s a quick tutorial/overview:

2 Likes
9

Hi Daniel,

I am using Paperspace Gradient notebook for the course and would appreciate some help in downloading the IMDb data to the environment. It appears as though Gradient notebook is some closed environment which only contains some fast.ai data sets (https://paperspace.zendesk.com/hc/en-us/articles/360003092514-Public-Datasets) but not IMDb!

Asked another way: can we use Gradient notebook to run Part II Lesson 10 ULMFit?

Thanks

10

@alvu The Fast.ai Datasets are under the fastai folder:

image
image.png1191Ă—474 18.1 KB

I don’t know what the limiting factor for Lesson 10 is – what appears to be missing?

11

When I modify the code to point to this location, I see this error:

image
image.png1026Ă—676 34.3 KB