Enable benchmarked two-slot medium profile

This commit is contained in:
Mikei386
2026-08-30 20:54:22 +02:00
parent db3719af7a
commit a41d6dc28c
7 changed files with 60 additions and 21 deletions
+13 -7
View File
@@ -109,7 +109,8 @@ services:
- --ubatch-size
- "${FAST_UBATCH_SIZE:-32}"
- --parallel
- "1"
- "${FAST_PARALLEL_SLOTS:-1}"
- --kv-unified
- --jinja
- --reasoning
- auto
@@ -185,7 +186,8 @@ services:
- --ubatch-size
- "${MEDIUM_UBATCH_SIZE:-128}"
- --parallel
- "1"
- "${MEDIUM_PARALLEL_SLOTS:-2}"
- --kv-unified
- --jinja
- --reasoning
- auto
@@ -221,7 +223,7 @@ services:
- --split-mode
- layer
- --tensor-split
- "${MEDIUM_TENSOR_SPLIT:-90,10}"
- "${MEDIUM_TENSOR_SPLIT:-85,15}"
- --spec-type
- draft-mtp
- --spec-draft-n-max
@@ -272,7 +274,8 @@ services:
- --ubatch-size
- "${LARGE_UBATCH_SIZE:-128}"
- --parallel
- "1"
- "${LARGE_PARALLEL_SLOTS:-1}"
- --kv-unified
- --jinja
- --reasoning
- auto
@@ -349,7 +352,8 @@ services:
- --ubatch-size
- "${ULTRA_UBATCH_SIZE:-128}"
- --parallel
- "1"
- "${ULTRA_PARALLEL_SLOTS:-1}"
- --kv-unified
- --jinja
- --reasoning
- auto
@@ -430,7 +434,8 @@ services:
- --ubatch-size
- "${UNCENSORED_UBATCH_SIZE:-128}"
- --parallel
- "1"
- "${UNCENSORED_PARALLEL_SLOTS:-1}"
- --kv-unified
- --jinja
- --reasoning
- auto
@@ -498,7 +503,8 @@ services:
- --cache-ram
- "${LLAMA_CACHE_RAM_MIB:-8192}"
- --parallel
- "1"
- "${EXPERIMENTAL_PARALLEL_SLOTS:-1}"
- --kv-unified
- --jinja
- --reasoning
- auto