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

[WIP] Add RAPIDS Nightly to GPU CI #436

Draft
wants to merge 5 commits into
base: main
Choose a base branch
from
+79 −50
Draft

[WIP] Add RAPIDS Nightly to GPU CI #436

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
f3ce8d5
fc fancy
praateekmahajan Dec 17, 2024
76c9eac
sc
praateekmahajan Dec 17, 2024
28c84bf
tc
praateekmahajan Dec 17, 2024
077ba5c
testing if fassttext 0.9.3 works, if so we should merge 434 -> 435 ->…
praateekmahajan Dec 17, 2024
d358bd7
update to nightly
praateekmahajan Dec 17, 2024
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
86 changes: 52 additions & 34 deletions .github/workflows/gpuci.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,8 +2,7 @@ name: "GPU CI/CD"

on:
push:
branches:
- main
branches: [main]
pull_request:
branches:
# We can run gpuCI on any PR targeting these branches
Expand All @@ -12,28 +11,36 @@ on:
- '[rv][0-9].[0-9].[0-9]rc[0-9]'
# PR has to be labeled with "gpuCI" label
# If new commits are added, the "gpuCI" label has to be removed and re-added to rerun gpuCI
types: [ labeled ]
types: [labeled]

concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true

# Reusable job templates
jobs:
# First, we build and push a NeMo-Curator container
build-container:
# "build-container" job is run if the "gpuci" label is added to the PR
if: ${{ github.event.label.name == 'gpuci' || github.ref == 'refs/heads/main' }}
strategy:
matrix:
include:
- type: stable
image-suffix: ""
- type: nightly
image-suffix: "_nightly"
uses: NVIDIA/NeMo-FW-CI-templates/.github/workflows/[email protected]
with:
image-name: nemo_curator_container
image-name: nemo_curator_container${{ matrix.image-suffix }}
dockerfile: Dockerfile
image-label: nemo-curator
image-label: nemo-curator${{ matrix.image-suffix }}
build-args: |
IMAGE_LABEL=nemo-curator
IMAGE_LABEL=nemo-curator${{ matrix.image-suffix }}
REPO_URL=https://github.com/${{ github.repository }}.git
CURATOR_COMMIT=${{ github.sha }}
BUILD_TYPE=${{ matrix.type }}
prune-filter-timerange: 24h

# Then, we run our PyTests in the container we just built
run-gpu-tests:
needs: build-container
Expand All @@ -42,48 +49,59 @@ jobs:
runs-on: self-hosted-azure
# "run-gpu-tests" job is run if the "gpuci" label is added to the PR
if: ${{ github.event.label.name == 'gpuci' || github.ref == 'refs/heads/main' }}
strategy:
matrix:
include:
- type: stable
image-suffix: ""
- type: nightly
image-suffix: "_nightly"

env:
CONTAINER_NAME: nemo-curator-container${{ matrix.image-suffix }}
IMAGE_NAME: nemoci.azurecr.io/nemo_curator_container${{ matrix.image-suffix }}:${{ github.run_id }}

steps:
# If something went wrong during the last cleanup, this step ensures any existing container is removed
- name: Remove existing container if it exists
run: |
if [ "$(docker ps -aq -f name=nemo-curator-container)" ]; then
docker rm -f nemo-curator-container
if [ "$(docker ps -aq -f name=${{ env.CONTAINER_NAME }})" ]; then
docker rm -f ${{ env.CONTAINER_NAME }}
fi

# This runs the container which was pushed by build-container, which we call "nemo-curator-container"
# `--gpus all` ensures that all of the GPUs from our self-hosted-azure runner are available in the container
# We use "github.run_id" to identify the PR with the commits we want to run the PyTests with
# `bash -c "sleep infinity"` keeps the container running indefinitely without exiting
# This runs the container which was pushed by build-container, which we call "nemo-curator-container"
# `--gpus all` ensures that all of the GPUs from our self-hosted-azure runner are available in the container
# We use "github.run_id" to identify the PR with the commits we want to run the PyTests with
# `bash -c "sleep infinity"` keeps the container running indefinitely without exiting
- name: Run Docker container
run: |
docker run --gpus all --name nemo-curator-container -d nemoci.azurecr.io/nemo_curator_container:${{ github.run_id }} bash -c "sleep infinity"

# Expect `whoami` to be "azureuser"
# Expect `nvidia-smi` to show our 2 A100 GPUs
- name: Check GPUs
run: |
whoami
docker exec nemo-curator-container nvidia-smi
docker run --gpus all --name ${{ env.CONTAINER_NAME }} \
-d ${{ env.IMAGE_NAME }} \
bash -c "sleep infinity"

# In the virtual environment (called "curator") we created in the container,
# list all of our packages. Useful for debugging
- name: Verify installations
# In the virtual environment (called "curator") we created in the container,
# list all of our packages. Useful for debugging
# Expect `whoami` to be "azureuser"
# Expect `nvidia-smi` to show our 2 A100 GPUs
- name: Check GPUs + Verify installations
run: |
docker exec nemo-curator-container pip list

# In the virtual environment (called "curator") we created in the container,
# run our PyTests marked with `@pytest.mark.gpu`
# We specify the `rootdir` to help locate the "pyproject.toml" file (which is in the root directory of the repository),
# and then the directory where the PyTests are located
echo "Checking system user:"
whoami
docker exec ${{ env.CONTAINER_NAME }} whoami
echo "Checking GPU availability:"
docker exec ${{ env.CONTAINER_NAME }} nvidia-smi
echo "Checking installed packages:"
docker exec ${{ env.CONTAINER_NAME }} pip list
# In the virtual environment (called "curator") we created in the container,
# run our PyTests marked with `@pytest.mark.gpu`
# We specify the `rootdir` to help locate the "pyproject.toml" file (which is in the root directory of the repository),
# and then the directory where the PyTests are located
- name: Run PyTests with GPU mark
run: |
docker exec nemo-curator-container pytest -m gpu --rootdir /opt/NeMo-Curator /opt/NeMo-Curator/tests

docker exec ${{ env.CONTAINER_NAME }} pytest -m gpu --rootdir /opt/NeMo-Curator /opt/NeMo-Curator/tests
# After running `docker stop`, the container remains in an exited state
# It is still present on our system and could be restarted with `docker start`
# Thus, we use `docker rm` to permanently removed it from the system
- name: Cleanup
if: always()
run: |
docker stop nemo-curator-container && docker rm nemo-curator-container
run: docker stop ${{ env.CONTAINER_NAME }} && docker rm ${{ env.CONTAINER_NAME }}
21 changes: 16 additions & 5 deletions Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ ARG PYTHON_VER=3.10
ARG IMAGE_LABEL
ARG REPO_URL
ARG CURATOR_COMMIT
ARG BUILD_TYPE=stable

FROM rapidsai/ci-conda:cuda${CUDA_VER}-${LINUX_VER}-py${PYTHON_VER} as curator-update
# Needed to navigate to and pull the forked repository's changes
Expand All @@ -23,14 +24,16 @@ RUN bash -exu <<EOF
git checkout $CURATOR_COMMIT
EOF


FROM rapidsai/ci-conda:cuda${CUDA_VER}-${LINUX_VER}-py${PYTHON_VER}
LABEL "nemo.library"=${IMAGE_LABEL}
WORKDIR /opt

# Re-declare ARGs after new FROM to make them available in this stage
ARG CUDA_VER
ARG BUILD_TYPE

# Install the minimal libcu* libraries needed by NeMo Curator
ENV _CUDA_VER=${CUDA_VER}
RUN conda create -y --name curator -c nvidia/label/cuda-${_CUDA_VER} -c conda-forge \
RUN conda create -y --name curator -c nvidia/label/cuda-${CUDA_VER} -c conda-forge \
python=3.10 \
cuda-cudart \
libcufft \
Expand All @@ -48,15 +51,23 @@ RUN \
--mount=type=bind,source=/opt/NeMo-Curator/pyproject.toml,target=/opt/NeMo-Curator/pyproject.toml,from=curator-update \
cd /opt/NeMo-Curator && \
source activate curator && \
pip install ".[all]"
if [ "$BUILD_TYPE" = "nightly" ]; then \
pip install ".[all_nightly]"; \
else \
pip install ".[all]"; \
fi
Comment on lines +54 to +58
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We might need to check if the nightly can use the same caching logic as stable, or if nightly should just always build a fresh non-cached container.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1, wanted to flag this explicitly as we do some curator-update optimization above, and I want to ensure if branching doesn't mess with it.
cc @ko3n1g

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do these optimizations come from the base-image?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I mean, if they do even if the base-image tag is unchanged b/c they overwrite it we should extend this build-workflow so that we can force pull the base image. In all other cases we don't need to be worried about caching i believe


COPY --from=curator-update /opt/NeMo-Curator/ /opt/NeMo-Curator/

# Clone the user's repository, find the relevant commit, and install everything we need
RUN bash -exu <<EOF
source activate curator
cd /opt/NeMo-Curator/
pip install --extra-index-url https://pypi.nvidia.com ".[all]"
if [ "$BUILD_TYPE" = "nightly" ]; then \
pip install --extra-index-url=https://pypi.anaconda.org/rapidsai-wheels-nightly/simple ".[all_nightly]"; \
else \
pip install --extra-index-url https://pypi.nvidia.com ".[all]"; \
fi
EOF

ENV PATH /opt/conda/envs/curator/bin:$PATH
22 changes: 11 additions & 11 deletions pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,7 @@ dependencies = [
"dask[complete]>=2021.7.1",
"datasets",
"distributed>=2021.7.1",
"fasttext==0.9.2",
"fasttext==0.9.3",
"ftfy==6.1.1",
"in-place==0.5.0",
"jieba==0.42.1",
Expand Down Expand Up @@ -75,20 +75,20 @@ dynamic = ["version"]
[project.optional-dependencies]
# Installs CPU + GPU text curation modules
cuda12x = [
"cudf-cu12>=24.10",
"cugraph-cu12>=24.10",
"cuml-cu12>=24.10",
"dask-cuda>=24.10",
"dask-cudf-cu12>=24.10",
"cudf-cu12>=24.12",
"cugraph-cu12>=24.12",
"cuml-cu12>=24.12",
"dask-cuda>=24.12",
"dask-cudf-cu12>=24.12",
"spacy[cuda12x]>=3.6.0, <3.8.0",
]
# Installs CPU + GPU text curation modules with RAPIDS Nightlies
cuda12x_nightly = [
"cudf-cu12>=24.12.0a0,<=24.12",
"cugraph-cu12>=24.12.0a0,<=24.12",
"cuml-cu12>=24.12.0a0,<=24.12",
"dask-cuda>=24.12.0a0,<=24.12",
"dask-cudf-cu12>=24.12.0a0,<=24.12",
"cudf-cu12>=25.02.0a0,<=25.02",
"cugraph-cu12>=25.02.0a0,<=25.02",
"cuml-cu12>=25.02.0a0,<=25.02",
"dask-cuda>=25.02.0a0,<=25.02",
"dask-cudf-cu12>=25.02.0a0,<=25.02",
"spacy[cuda12x]>=3.6.0, <3.8.0",
]
# Installs CPU + GPU text and image curation modules
Expand Down
Loading