Cherry-picking vLLM backend changes (#6404)

* Update build.py to build vLLM backend (#6394) * Add Python backend when vLLM backend built (#6397) --------- Co-authored-by: dyastremsky <[email protected]>
triton-inference-server · Oct 26, 2023 · 21ee57a · 21ee57a
1 parent 4479d94
commit 21ee57a
Showing 1 changed file with 12 additions and 0 deletions.
diff --git a/build.py b/build.py
@@ -78,6 +78,8 @@
         "2023.0.0",  # Standalone OpenVINO
         "2.4.7",  # DCGM version
         "py310_23.1.0-1",  # Conda version
+        "9.1.0.1",  # TRT version for building TRT-LLM backend
+        "12.2",  # CUDA version for building TRT-LLM backend
         "0.2.1.post1",  # vLLM version
     )
 }
@@ -1364,6 +1366,16 @@ def dockerfile_prepare_container_linux(argmap, backends, enable_gpu, target_mach
             TRITON_VERSION_MAP[FLAGS.version][7]
         )
 
+    if "vllm" in backends:
+        # [DLIS-5606] Build Conda environment for vLLM backend
+        # Remove Pip install once vLLM backend moves to Conda environment.
+        df += """
+# vLLM needed for vLLM backend
+RUN pip3 install vllm=={}
+""".format(
+            TRITON_VERSION_MAP[FLAGS.version][9]
+        )
+
     df += """
 WORKDIR /opt/tritonserver
 RUN rm -fr /opt/tritonserver/*