Это видео недоступно.
Сожалеем об этом.
FlashAttention-3 is Here
HTML-код
- Опубликовано: 15 авг 2024
- This video introduces FlashAttention-3 which uses asynchrony to perform multiple operations simultaneously. By overlapping the main operations like matrix multiplications and softmax, it ensures the GPU stays busy and efficient.
🔥 Buy Me a Coffee to support the channel: ko-fi.com/fahd...
🔥 Get 50% Discount on any A6000 or A5000 GPU rental, use following link and coupon:
bit.ly/fahd-mirza
Coupon code: FahdMirza
▶ Become a Patron 🔥 - / fahdmirza
#flashattention #flashattention3 #softmax #matmul
PLEASE FOLLOW ME:
▶ LinkedIn: / fahdmirza
▶ RUclips: / @fahdmirza
▶ Blog: www.fahdmirza.com
RELATED VIDEOS:
▶ Resource pytorch.org/bl...
All rights reserved © Fahd Mirza
maybe you should do a video on mass compute ? and how to set it up for training etc? - it seems expensive ? it has the machines but it seems expensive ? is that what you use for your tutorials ?
does it have a colab ? or is it vscode ?
Sure, noted.
The question is... will it work on a 3090?
No, it’s for Hopper only. The algorithm is completely different and uses support from newer features .