Comments (4)
Sorry for the delayed answer, I've been quite busy and missed this. The documentation is not up to date, so I will fix this, let you know and update it!
from triton.
The correct use of matrix multiplication is indeed @. The other issue is that the kernel is designed for row-major matrices, but the values of lda are assuming column-major ordering.
Changing to:
lda = K
ldb = N
ldc = N
Seems to solve the issue. I would also use torch.allclose() for comparing two FP32 tensors as torch may not do the FMA in the same order as the Triton kernel.
from triton.
AH thank you I don't know why I didn't consider that the kernel was assuming row-major accesses.
I think I understand everything now, so if you'd like I'd be happy to run through the documentation and make sure the example kernels are compilable/working.
Thanks again!
from triton.
Yes, that would be very cool! Another thing is that triton.empty and triton.shape are no longer necessary. One can simply use torch.empty
from triton.
Related Issues (20)
- Trivial outputs on MacOS using TRITON_INTERPRET=1 HOT 7
- 导入default_dump_dir报错
- What is the use of calling `propagateIfChanged` in `ConstantAnalysis::visit`
- Flash Attention benchmarks fail on AMD Instinct MI210 (`gfx90a`) HOT 2
- [OPS] AttributeError: module 'triton.language.math' has no attribute 'round' HOT 3
- Why is the documentation not versioned like other Read The Docs sites?
- Core dumped without "os.environ["TRITON_INTERPRET"] = "1"
- torch 2.4.0 requires triton==3.0.0; HOT 7
- Triton trying to download llvm even if self built, and build failed
- Seems mistake in setup.py HOT 1
- bfloat16 of fused attention seems have bug HOT 2
- Possible to use WMMA `.b1` instructions? HOT 1
- AttributeError("'AnnAssign' object has no attribute 'targets'") HOT 1
- [Bug] Compiler error in 06-fused-attention.py
- Fused Attention FP8 correctness on L20(Ada GPU) HOT 1
- What impact does changing the kernel shape's divisibility value have?
- Does Triton have a roadmap plan? HOT 1
- FP8 Casting Tensor Size Alignment Error on AMD MI210 HOT 1
- Output dtype of 09-persistent-matmul.py
- Worse performance on `H100` than `RTX3090`
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from triton.