From 34c43958d0a5c9c5549239af50fa927a44183aa2 Mon Sep 17 00:00:00 2001
From: biondizzle <biondizzle@gmail.com>
Date: Sun, 17 May 2026 07:10:16 +0000
Subject: [PATCH] vllm tweaks

---
 docker-compose.yml | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/docker-compose.yml b/docker-compose.yml
index fd7b48cf..997732dd 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -19,7 +19,12 @@ services:
       - --compilation-config
       #- '{"cudagraph_mode": "FULL_DECODE_ONLY", "custom_ops": ["all"], "cudagraph_capture_sizes": [1, 2, 4, 8], "max_cudagraph_capture_size": 8}'
       - '{"cudagraph_mode":"FULL_AND_PIECEWISE", "custom_ops":["all"]}'
+      - --moe-backend deep_gemm_mega_moe
       - --tokenizer-mode=deepseek_v4
+      #- --attention_config.use_fp4_indexer_cache=True
+      - --tool-call-parser deepseek_v4
+      - --enable-auto-tool-choice
+      - --reasoning-parser deepseek_v4
       - --host=0.0.0.0
       - --port=8000
     deploy: