NVIDIA Open-Sources TensorRT Library Components

[phoronix]

Phoronix: NVIDIA Open-Sources TensorRT Library Components

NVIDIA announced via their newsletter today that they've open-sourced their TensorRT library and associated plug-ins...

http://www.phoronix.com/scan.php?page=news_item&px=NVIDIA-TensorRT-Open-Source

[yoshi314]

Building the open-source TensorRT code still depends upon the proprietary CUDA as well as other common build dependencies

they throw you a carrot, but they still keep the leash.

[schmidtbag]

they throw you a carrot, but they still keep the leash.

Well, CUDA is also open-sourced, but you still pretty much have to have Nvidia hardware. So basically they throw you a carrot, keep the leash, and they pull the carrot away on a string as soon as you get a little too close to it.

[TheBlackCat]

Well, CUDA is also open-sourced,

Since when? Wikipedia says it is still proprietary, and I don't see anything in the license terms to suggest it is open-source

[entropy]

If Nvidia open sources something it's either a failed product (failed by means that they could not generate a vendor lock-in)
or a small fraction intentionally not useful for competing hw.

[TorvaldsPic]

[jojo7887]

It's no secret that Nvidia keeps a tight control over their stuff, especially their GPU drivers. However I don't see a reason to complain when they open source something, and aside from that, they still help Linux (and gaming) through Vulkan development. You should be thankful.

They're not as open source friendly as much as we'd like to, but something is still better than nothing. I don't wanna sound like an Nvidia fan boy or anything, as I still consider AMD to be king when it comes to open source friendliness, and they're thriving at that.

[schmidtbag]

Since when? Wikipedia says it is still proprietary, and I don't see anything in the license terms to suggest it is open-source

I guess open-source isn't the right way of putting it. The last paragraph in this article sums it up:

https://www.anandtech.com/show/5238/nvidia-releases-cuda-41-cuda-goes-llvm-and-open-source-kind-of

[TheBlackCat]

I guess open-source isn't the right way of putting it. The last paragraph in this article sums it up:
https://www.anandtech.com/show/5238/...source-kind-of

So it is almost completely unlike open-source in every way. They took an originally open-source tool, made a closed-source fork, and let certain people apply to look at it as long as they don't tell anyone else what they saw. There is no sense in which it is even remotely "open" in any way.

[AndyChow]

Is TensorRT only working with Python 2 and still broken with Python 3?

[skeevy420]

Since when? Wikipedia says it is still proprietary, and I don't see anything in the license terms to suggest it is open-source

I love number 7:

You agree to defend, indemnify and hold harmless NVIDIA and its affiliates, and their respective employees, contractors, agents, officers and directors, from and against any and all claims, damages, obligations, losses, liabilities, costs or debt, fines, restitutions and expenses (including but not limited to attorney’s fees and costs incident to establishing the right of indemnification) arising out of or related to your use of the SDK outside of the scope of this Agreement, or not in compliance with its terms.

If you use their SDK you have to defend Nvidia every time we post bad things about the SDK or Nvidia in general.

The Nvidia CUDA SDK sucks, I claim that their CEO is a punk, and it's liable that everyone on their board of directors are pedophiles. Now y'all CUDA SDK users have to defend Nvidia or risk being sued. It's right there in the EULA....

Number 6 is just funny:

Unless you have an agreement with NVIDIA for this purpose, you may not use the SDK with any system or application where the use or failure of the system or application can reasonably be expected to threaten or result in personal injury, death, or catastrophic loss. Examples include use in nuclear, avionics, navigation, military, medical, life support or other life critical applications. NVIDIA does not design, test or manufacture the SDK for these critical uses and NVIDIA shall not be liable to you or any third party, in whole or in part, for any claims or damages arising from such uses.

There goes my plans for a CUDA enabled rail gun. It needs CUDA to quickly calculate the Coriolis effect for long-range shooting.

bridgman Does AMD care if I want to make a rail gun with Coriolis effect calculations on a GPU or APU with various AMD SDKs?

OK, this is also pretty bad:

These exclusions and limitations of liability shall apply regardless if NVIDIA or its affiliates have been advised of the possibility of such damages, and regardless of whether a remedy fails its essential purpose. These exclusions and limitations of liability form an essential basis of the bargain between the parties, and, absent any of these exclusions or limitations of liability, the provisions of this Agreement, including, without limitation, the economic terms, would be substantially different.

They can be informed that their SDK will murder kittens and not tell their users while also not being at fault that their SDK murdered kittens. Don't support SDKs that don't warn about kitten murder.