Decode Speed Is Bytes Divided by Bandwidth, and the Batch Ceiling Is 848 Tokens
DEV Community
Decode Speed Is Bytes Divided by Bandwidth, and the Batch Ceiling Is 848 Tokens
At batch 1 decode is a traffic budget. Here is the formula for three models on an H100, the context past which batching stops helping, and why quantizing everything moves the wall without moving the distance to it.
0 comments
No comments yet.