DO NOT MERGE : Layer-by-Layer Profiling#3

Closed

varun-sundar-rabindranath wants to merge 18 commits intomainfrom varun/main-with-profiler

+1,938-448

Commits on Oct 16, 2024

[Kernel][Model] Improve continuous batching for Jamba and Mamba (vllm-project#9189 )
mzusman
authored
[Model][Bugfix] Add FATReLU activation and support for openbmb/MiniCPM-S-1B-sft (vllm-project#9396 )
streaver91
authored
add profiler

Varun Sundar Rabindranath
authored and
LucasWilkinson
committed
refactor
LucasWilkinson
committed
remove cruft
LucasWilkinson
committed
format
LucasWilkinson
committed
fix duplicate entries in the tp > 1 case
LucasWilkinson
committed
add reduce kernel grouping

Varun Sundar Rabindranath
authored and
LucasWilkinson
committed
move to tools folder
LucasWilkinson
committed
Update examples/offline_profile.py

LucasWilkinson
and
mgoin
committed
review comments
LucasWilkinson
committed
add offline profile to "Examples Test"
LucasWilkinson
committed
allow trace saving and add a warmup run
LucasWilkinson
committed
fix deprecation warning
LucasWilkinson
committed
format
LucasWilkinson
committed
review comments
LucasWilkinson
committed
pull in all engine args
LucasWilkinson
committed
rename save traces
LucasWilkinson
committed