From 1f537bda3839bffb511df16e11341f6cb2aedad7 Mon Sep 17 00:00:00 2001
From: QianqianNie <44004519+QianqianNie@users.noreply.github.com>
Date: Sat, 17 Aug 2024 12:16:41 -0700
Subject: [PATCH] Update foundation model image (#3275)

---
 .../foundation-model-inference/context/Dockerfile         | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/assets/training/model_management/environments/foundation-model-inference/context/Dockerfile b/assets/training/model_management/environments/foundation-model-inference/context/Dockerfile
index 1a20da2a77..7ec1de6db0 100644
--- a/assets/training/model_management/environments/foundation-model-inference/context/Dockerfile
+++ b/assets/training/model_management/environments/foundation-model-inference/context/Dockerfile
@@ -47,12 +47,11 @@ RUN pip install git+https://github.com/stanford-futuredata/megablocks.git@5897cd
 # RUN pip install -e ./ --no-cache-dir
 
 # When copied to assets repo, change to install from public pypi
-RUN pip install llm-optimized-inference==0.2.7 --no-cache-dir
+RUN pip install llm-optimized-inference==0.2.8 --no-cache-dir
 
 RUN pip uninstall -y vllm
-
-# Install patched vllm wheel
-RUN pip install https://automlsamplenotebookdata.blob.core.windows.net/vllm/vllm-0.5.3.post1-cp310-cp310-linux_x86_64.whl
+RUN pip uninstall transformers -y
+RUN pip install https://automlsamplenotebookdata.blob.core.windows.net/vllm/vllm-0.4.1-cp310-cp310-linux_x86_64.whl
 
 # clean conda and pip caches
 RUN rm -rf ~/.cache/pip
@@ -63,6 +62,5 @@ RUN chmod +x /var/runit/api_server/run
 
 ENV SVDIR=/var/runit
 ENV WORKER_TIMEOUT=3600
-ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
 EXPOSE 5001
 CMD [ "runsvdir", "/var/runit" ]
\ No newline at end of file