technically it is representing the padding unused/wasted compute so even if it being on the left would actually be the same thing/correct representation but I think he wants to show that that is the actual wasted compute (to be quantifiable)
Akash
akash-inferq
AI & ML interests
None yet
Recent Activity
commented on an article 4 days ago
Continuous batching from first principles upvoted an article 7 days ago
Continuous batching from first principles Organizations
None yet
commented on Continuous batching from first principles 4 days ago
upvoted an article 7 days ago
Article
Continuous batching from first principles
- +1
•
343
Issues while running the model within vLLMs nightly docker build as given in the readme
1
#1 opened 3 months ago
by
akash-inferq
liked a Space 12 months ago