Description
The ability to install the Cortex framework on a bare-metal machine or in lieu of this a self-hosted environment.
Motivation
Increase the adaption rate of Cortex, many companies/ individuals become cost evasive when it concerns utilizing GPUs within the cloud. Either because of lack of knowledge/ general misunderstanding or just because they understand the amount of data they will need to process actually will cost too much.
Additional context
I would love to have offline installation capability for the framework and I am willing to implement the features and make pull requests to get it.
Background: I built a machine learning framework for video processing centered around Tensorflow Serving for a private application with my company. The solution as it stands has rather limited support for other frameworks such as PyTorch and scikit-learn.
Reasoning: Having the ability to install Cortex on bare metal servers would increase the adoption rate significantly in certain communities in my opinion. Most of the clients utilizing my application have more than enough funding to self-support hosting but are extremely evasive when it comes to utilizing AWS because of fears of long term cost.
For instance, they have already purchased multiple server racks with Nvidia GPUs which are currently utilizing Nvidia GRID / vGPU with the application I made. They will not want to backtrack and waste all the money they used for purchasing those server racks/GPUs by transitioning to a cloud-only solution.
If this is a feature you all would like to have available for Cortex please get in touch with me, we can have a conversation somewhere beyond this ticket if it is convenient. Otherwise, just point me where I am needed and I will get started this weekend.
Looking forward to hearing from you.
-Sam
Description
The ability to install the Cortex framework on a bare-metal machine or in lieu of this a self-hosted environment.
Motivation
Increase the adaption rate of Cortex, many companies/ individuals become cost evasive when it concerns utilizing GPUs within the cloud. Either because of lack of knowledge/ general misunderstanding or just because they understand the amount of data they will need to process actually will cost too much.
Additional context
I would love to have offline installation capability for the framework and I am willing to implement the features and make pull requests to get it.
Background: I built a machine learning framework for video processing centered around Tensorflow Serving for a private application with my company. The solution as it stands has rather limited support for other frameworks such as PyTorch and scikit-learn.
Reasoning: Having the ability to install Cortex on bare metal servers would increase the adoption rate significantly in certain communities in my opinion. Most of the clients utilizing my application have more than enough funding to self-support hosting but are extremely evasive when it comes to utilizing AWS because of fears of long term cost.
For instance, they have already purchased multiple server racks with Nvidia GPUs which are currently utilizing Nvidia GRID / vGPU with the application I made. They will not want to backtrack and waste all the money they used for purchasing those server racks/GPUs by transitioning to a cloud-only solution.
If this is a feature you all would like to have available for Cortex please get in touch with me, we can have a conversation somewhere beyond this ticket if it is convenient. Otherwise, just point me where I am needed and I will get started this weekend.
Looking forward to hearing from you.
-Sam