Re:Why ML Needs a New Programming Language
by almostgotcaught on Friday September 05, 2025 @06:17PM
(#45144301)
Triton is a backend for PyTorch. Lately it is the backend. So it's definitely double digits percentage if not over 50%.
Re:Why ML Needs a New Programming Language
by lairv on Friday September 05, 2025 @08:52PM
(#45145514)
It's the backend for torch.compile, pytorch eager mode will still use cuBLAS/cuDNN/custom CUDA kernels, not sure what's the usage of torch.compile
Re:Why ML Needs a New Programming Language
by almostgotcaught on Friday September 05, 2025 @10:10PM
(#45145971)
> not sure what's the usage of torch.compile
consider that at minimum both FB and OAI themselves definitely make heavy use of the Triton backend in PyTorch.
Re:Why ML Needs a New Programming Language
by cavisne on Saturday September 06, 2025 @02:37AM
(#45147142)
Doesn’t triton write its own intermediate language that then compiles to PTX?
Re:Why ML Needs a New Programming Language
by almostgotcaught on Saturday September 06, 2025 @05:08AM
(#45147752)
Yes and?
Re:Why ML Needs a New Programming Language
by saagarjha on Saturday September 06, 2025 @05:14AM
(#45147783)
It has a fairly standard MLIR pipeline
Re:Why ML Needs a New Programming Language
by saagarjha on Saturday September 06, 2025 @05:16AM
(#45147790)
A lot. OpenAI uses Triton for their critical kernels. Meta has torch.compile using it too. I know Anthropic is not using Triton but I think their homegrown compiler is also Python. xAI is using CUTLASS which is C++ but I wouldn't be surprised if they start using the Python API moving forward.
Re:Why ML Needs a New Programming Language
by almostgotcaught on Saturday September 06, 2025 @07:48AM
(#45148453)
Anthropic is a Jax shop
How many of the world's total FLOPs are going through those?