Comments (1)
Had to fix a bunch of shape issues. Final shapes:
assert o.shape == (B, num_heads, N + L, head_dim)
assert q_padded.shape == (B, num_heads, N + L, head_dim)
assert k_view.shape == (B, num_heads, N + L, head_dim)
assert v_view.shape == (B, num_heads, N + L, head_dim)
assert stats.shape == (B, num_heads, N + L, 1)
assert seqlens_q.shape == (B, 1, 1, 1), f"seqlens_q.shape={seqlens_q.shape} instead of {B}"
assert seqlens_kv.shape == (B, 1, 1, 1), f"seqlens_kv.shape={seqlens_kv.shape} instead of {B}"
from cudnn-frontend.
Related Issues (20)
- Question about calling MHA HOT 2
- Question About Reduce Node
- Trouble when fusing layernorm with pointwise operation. HOT 2
- How to use cudnn frontend in Jax? HOT 4
- [Question] Should I be creating the graph everytime? HOT 3
- [Question] Why are the physical and logical layouts different in the SDPA example? HOT 1
- Is performance better when not using padding?
- [Question] Get better error messages for "ValueError: dimensions mismatch as broadcasting 2 non-one dimension sizes."? HOT 4
- [Question] How to match Flash Attention 2 performance? HOT 4
- [Question] Making dO contiguous affects output? HOT 2
- Support batch size 0
- [Question] How is the convolution algorithm set, and how to specify group number ?
- Extremely slow fp8 conv2d wgrad operation HOT 2
- [Quenstion] build_operation_graph failed HOT 5
- Which Visual Studio 2022 BuildTools MSVC is the best version for Cuda 11.8 and Cuda 12.4 and so
- Wrong result of tensor addition with broadcasting
- cudnn-frontend crashes in case of MAX_OPGRAPH_OPS violation
- [Question] Conversion between NCHW and NHWC HOT 2
- [Question] How to add bias after conv? HOT 3
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 cudnn-frontend.