diff --git a/scripts/serve-frank-large.sh b/scripts/serve-frank-large.sh index f6e5e23..36a36c1 100755 --- a/scripts/serve-frank-large.sh +++ b/scripts/serve-frank-large.sh @@ -47,8 +47,8 @@ fi : "${LISTEN_PORT:=7770}" : "${CTX_SIZE:=401408}" : "${PARALLEL:=2}" -:: "${BATCH_SIZE:=4096}" -:: "${UBATCH_SIZE:=512}" +: "${BATCH_SIZE:=4096}" +: "${UBATCH_SIZE:=512}" : "${NUM_THREADS:=16}" : "${THREAD_BATCH:=16}" : "${CACHE_TYPE_K:=f16}" @@ -67,8 +67,8 @@ fi : "${PRESENCE_PENALTY:=1.5}" : "${REASONING:=on}" : "${REASONING_FORMAT:=deepseek}" -:: "${HSA_OVERRIDE_GFX_VERSION:=11.5.1}" -:: "${SPEC_TYPE:=none}" +: "${HSA_OVERRIDE_GFX_VERSION:=11.5.1}" +: "${SPEC_TYPE:=none}" : "${DRAFT_MODEL_DIR:=/home/dank/models/mtp-Qwen3.8-Flash-Next-FAST}" : "${DRAFT_MODEL_FILE:=Qwen3.8-Flash-Next-MTP-ROCmFP4-FAST.gguf}" : "${SPEC_DRAFT_N_MAX:=3}"