Lower dankserver compute threads 8→7 to leave CPU headroom
This commit is contained in:
@@ -30,5 +30,6 @@ UBATCH_SIZE=2048
|
||||
# Load the model into RAM rather than lazy-mapping it. The MoE experts live on
|
||||
# CPU, so eager RAM load avoids page faults during decode (uses ~31GB RAM).
|
||||
NO_MMAP=1
|
||||
THREADS=8
|
||||
# Compute threads lowered to 7 so other processes on the box keep a core.
|
||||
THREADS=7
|
||||
THREADS_HTTP=8
|
||||
Reference in New Issue
Block a user