Implement GPU offloading for our containerized python app (flexible compensation)

We run a containerized implementation of our solution. Our servers are Ubuntu 18/cli, frontend is vuejs with typescript and vuetify and our backend is built on python with fastapi. You will be working with the backend for this project.

Our application is using advanced market data for digital assets/finance pulled from our own API and stored in local Redis. We run a series of user specified range-based criteria's against that data to identify trends and find truths. We have a brute force back testing engine that schedules 1000's-100k's of combinations of tests to run and then coordinates those tests that ultimately callout python scripts for each test. We can specify the number of workers of which will then run the scheduled jobs in parallel but currently limited by the number of CPU's we have on our server.

We are looking to enable GPU offloading. We don't have experience in this area, whether we should consider cuda or tensorflow or ?? and what libraries would better suit our need. We need you to identify the way forward and then develop a solution that enables us to offload work to a GPU and leverage something such as the NVidia P4 or k40 or v100 or possibly standard GPU's such as nvidia 1080 (but also still supporting non-gpu based systems as it currently does).

Ideally the solution will enable us to pass switches and or configs in our cli/ python scripts. We run these scripts inside of a containers so solution should be compatible with docker/kubernetes. See the code you will be working with here > [login to view URL]

If you think you can likely tackle this job engage us and request any additional data you may need to bid on this job.

When you provide a bid on this contract you should outline the following:

1. Your chosen approach (libraries, etc)

2. Estimated timeframe for delivery

3. Flat rate bid (pad for any uncertainties)

Notes about how we do business.

-If you deliver a working product perceivably free from major bugs in the timeframe you told us (#2 above), we pay an additional 10% bonus

-If you do not deliver a working product within 2x of the promised timeframe, we will delete your branch and cancel the contract.

-We wont leave bad reviews unless you really mess up but we will leave great reviews when you deliver!

Skills: Python, Django, Git, Tensorflow, Docker

See more: nvidia-docker pytorch, nvidia-docker run command, nvidia-docker tutorial, docker gpu support, nvidia-docker tensorflow, nvidia-container-toolkit, nvidia-docker2, tensorflow-gpu, python app engine sample, ubuntu running python app service, opensocial container python app engine, implement auto logoff j2ee web app, python app check phone, python app engine maps, python app goolge, python app engine examples, implement pdf document viewer rails app, socket python app engine, python app project management, php python app engine csv

About the Employer:
( 0 reviews ) United States

Project ID: #27973244

1 freelancer is bidding on average $1400 for this job


Hi, I have experience in dockerizing GPU applications on Ubuntu 18.04 LTS. To answer your question: 1. To be able to use GPU inside docker containers, nvidia-docker-tookit will be used, some information is needed from More

$1400 USD in 14 days
(4 Reviews)