Comments (3)
GeneZC
commented on August 24, 2024
Do we even still need the BPT if we have the ring attention implemented in this repo? @zhuzilin
I personally think BPT is a single-GPU version of ring attention, right?
from ring-flash-attention.
Edenzzzz
commented on August 24, 2024
Do we even still need the BPT if we have the ring attention implemented in this repo? @zhuzilin
I personally think BPT is a single-GPU version of ring attention, right?
That's right, BPT is inherently supported by ring attention. We do not need another implementation.
from ring-flash-attention.
zhuzilin
commented on August 24, 2024
I'm not sure spliting the sequence length on each device into blocks could save memory (because we still need save buffers and flash_attn itself seems to use linear size memory w.r.t. sequence length), or speed up (because it will call smaller kernels).
from ring-flash-attention.
Related Issues (20)
- test on 8*A800 HOT 3
- 关于tp和分块操作最终聚合的问题 HOT 1
- Question about updating lse HOT 2
- flash attention版本 HOT 1
- 精度问题 HOT 1
- large memory usage HOT 5
- How is ring attention applied while doing auto regressive decoding (in the stage of decoding tokens one by one)? HOT 1
- stripe_flash_attn_varlen_func HOT 1
- 是否需要更新全局最大值？ HOT 1
- [Feature Request] Support `window_size` HOT 3
- Does ring-attn not support dropout? HOT 3
- 多卡qkv维度问题 HOT 2
- ring attention实现原理 HOT 9
- Bugs when using zigzag_ring_flash_attn: RuntimeError: Number of requests do not match number of collectives
- 多机训练速度问题 HOT 3
- mask to zigzag attention HOT 1
- Numerical errors in backward
- 请问 ringattention这段代码中在算什么？
- run the code has error HOT 1
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 ring-flash-attention.