Select hybrid expert placement from benchmark
This commit is contained in:
@@ -20,9 +20,13 @@ TEMPLATE_FILE="$REPO_ROOT/scripts/chat-template.jinja"
|
||||
grep -q '^PARALLEL=3$' "$ENV_FILE"
|
||||
}
|
||||
|
||||
@test "frank-default-small keeps final eight expert blocks on CUDA" {
|
||||
grep -q '^OVERRIDE_EXPERT=' "$ENV_FILE"
|
||||
grep -Fq '31|32)\.ffn_' "$ENV_FILE"
|
||||
@test "frank-default-small uses the measured hybrid expert placement" {
|
||||
override=$(grep '^OVERRIDE_EXPERT=' "$ENV_FILE")
|
||||
[[ "$override" == *'26|28|29|30|32)'* ]]
|
||||
[[ "$override" != *'27'* ]]
|
||||
[[ "$override" != *'31'* ]]
|
||||
[[ "$override" != *'33'* ]]
|
||||
[[ "$override" != *'40'* ]]
|
||||
grep -q '^OVERRIDE_BUFT=CPU$' "$ENV_FILE"
|
||||
grep -q '^N_GPU_LAYERS=all$' "$ENV_FILE"
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user