diff --git a/computility-run.yaml b/computility-run.yaml index 87639310..9beaf135 100644 --- a/computility-run.yaml +++ b/computility-run.yaml @@ -19,7 +19,8 @@ command: - --disable-log-requests - --disable-frontend-multiprocessing - --max-num-batched-tokens - - '2048' + - '256' + - --enable-chunked-prefill - --max-seq-len-to-capture - '32768' - --enable-auto-tool-choice