Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Use post build function for TRT-LLM backend #6476

Merged
merged 2 commits into from
Nov 2, 2023
Merged

Use post build function for TRT-LLM backend #6476

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e0fdd46
Use postbuild function
krishung5 Oct 24, 2023
86667fb
Remove updating submodule url
krishung5 Nov 1, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
67 changes: 9 additions & 58 deletions build.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1305,54 +1305,17 @@ def dockerfile_prepare_container_linux(argmap, backends, enable_gpu, target_mach
# Add dependencies needed for tensorrtllm backend
if "tensorrtllm" in backends:
be = "tensorrtllm"
# url = "https://raw.githubusercontent.com/triton-inference-server/tensorrtllm_backend/{}/tools/gen_trtllm_dockerfile.py".format(
# backends[be]
# )

# response = requests.get(url)
# spec = importlib.util.spec_from_loader(
# "trtllm_buildscript", loader=None, origin=url
# )
# trtllm_buildscript = importlib.util.module_from_spec(spec)
# exec(response.content, trtllm_buildscript.__dict__)
# df += trtllm_buildscript.create_postbuild(backends[be])

df += """
WORKDIR /workspace
# Remove previous TRT installation
RUN apt-get remove --purge -y tensorrt* libnvinfer*
RUN pip uninstall -y tensorrt
# Install new version of TRT using the script from TRT-LLM
RUN apt-get update && apt-get install -y --no-install-recommends python-is-python3
RUN git clone --single-branch --depth=1 -b {} https://github.com/triton-inference-server/tensorrtllm_backend.git tensorrtllm_backend
RUN cd tensorrtllm_backend && git submodule set-url -- tensorrt_llm https://github.com/NVIDIA/TensorRT-LLM.git
RUN cd tensorrtllm_backend && git submodule sync
RUN cd tensorrtllm_backend && git submodule update --init --recursive
RUN cp tensorrtllm_backend/tensorrt_llm/docker/common/install_tensorrt.sh /tmp/
RUN rm -fr tensorrtllm_backend
""".format(
url = "https://raw.githubusercontent.com/triton-inference-server/tensorrtllm_backend/{}/tools/gen_trtllm_dockerfile.py".format(
backends[be]
)

df += """
RUN bash /tmp/install_tensorrt.sh && rm /tmp/install_tensorrt.sh
ENV TRT_ROOT=/usr/local/tensorrt
# Remove TRT contents that are not needed in runtime
RUN ARCH="$(uname -i)" && \
rm -fr ${TRT_ROOT}/bin ${TRT_ROOT}/targets/${ARCH}-linux-gnu/bin ${TRT_ROOT}/data && \
rm -fr ${TRT_ROOT}/doc ${TRT_ROOT}/onnx_graphsurgeon ${TRT_ROOT}/python && \
rm -fr ${TRT_ROOT}/samples ${TRT_ROOT}/targets/${ARCH}-linux-gnu/samples
# Install required packages for TRT-LLM models
RUN python3 -m pip install --upgrade pip && \
pip3 install transformers && \
pip3 install torch
# Uninstall unused nvidia packages
RUN if pip freeze | grep -q "nvidia.*"; then \
pip freeze | grep "nvidia.*" | xargs pip uninstall -y; \
fi
RUN pip cache purge
ENV LD_LIBRARY_PATH=/usr/local/tensorrt/lib/:/opt/tritonserver/backends/tensorrtllm:$LD_LIBRARY_PATH
"""
response = requests.get(url)
spec = importlib.util.spec_from_loader(
"trtllm_buildscript", loader=None, origin=url
)
trtllm_buildscript = importlib.util.module_from_spec(spec)
exec(response.content, trtllm_buildscript.__dict__)
df += trtllm_buildscript.create_postbuild(backends[be])

if "vllm" in backends:
# [DLIS-5606] Build Conda environment for vLLM backend
Expand Down Expand Up @@ -1843,22 +1806,10 @@ def backend_build(
cmake_script.comment()
cmake_script.mkdir(build_dir)
cmake_script.cwd(build_dir)
cmake_script.gitclone(backend_repo(be), tag, be, github_organization)

if be == "tensorrtllm":
cmake_script.cmd(
"git clone --single-branch --depth=1 -b {} https://github.com/triton-inference-server/tensorrtllm_backend tensorrtllm".format(
tag
)
)
cmake_script.cmd("cd tensorrtllm")
cmake_script.cmd(
"git submodule set-url -- tensorrt_llm https://github.com/NVIDIA/TensorRT-LLM.git"
)
cmake_script.cmd("git submodule sync")
cmake_script.cmd("cd ..")
tensorrtllm_prebuild(cmake_script)
else:
cmake_script.gitclone(backend_repo(be), tag, be, github_organization)

cmake_script.mkdir(repo_build_dir)
cmake_script.cwd(repo_build_dir)
Expand Down
Loading