Switch vLLM source to Gitea fork (cmm branch)

vllm/Dockerfile

@@ -10,7 +10,7 @@
 #   - vllm_managed_mem.py: Launcher that patches vLLM for managed memory
 #
 # Based on working Build #48 (v0.19.0):
-#   - vLLM: v0.19.0
+#   - vLLM: v0.19.0 (forked to sweetapi.com/biondizzle/vllm, cmm branch)
 #   - flashinfer: v0.6.6
 #   - flash-attention: hopper branch
 #   - lmcache: dev branch
@@ -156,11 +156,12 @@ RUN apt-get update && apt-get install -y build-essential cmake gcc && \
 # Build vLLM from source
 # ==============================================================================
 FROM build-base AS build-vllm
-# vLLM version to build
-ARG VLLM_REF=v0.19.0
+# vLLM version/branch to build
+# Using our Gitea fork (sweetapi.com/biondizzle/vllm) on the cmm branch
+ARG VLLM_REF=cmm
 # Install ccache for faster compilation
 RUN apt-get update && apt-get install -y ccache
-RUN git clone https://github.com/vllm-project/vllm.git
+RUN git clone https://sweetapi.com/biondizzle/vllm.git
 RUN cd vllm && \
     git checkout ${VLLM_REF} && \
     echo "\n\n========================================" && \