From 06c0490a3bd7cd5b25486ec24ee9a504c29a4d43 Mon Sep 17 00:00:00 2001 From: jalencato Date: Wed, 4 Sep 2024 15:35:22 -0700 Subject: [PATCH 1/7] Update Dockerfile.sm --- docker/sagemaker/Dockerfile.sm | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/docker/sagemaker/Dockerfile.sm b/docker/sagemaker/Dockerfile.sm index 93c2c6092f..f39df95839 100644 --- a/docker/sagemaker/Dockerfile.sm +++ b/docker/sagemaker/Dockerfile.sm @@ -3,15 +3,15 @@ ARG DEVICE=gpu -FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.13.1-gpu-py39-cu117-ubuntu20.04-sagemaker as branch-gpu +FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py39-cu117-ubuntu20.04-sagemaker as branch-gpu ENV dev_type=GPU # Install DGL GPU version -RUN pip3 install dgl==1.0.4+cu117 -f https://data.dgl.ai/wheels/cu117/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels/cu117/repo.html && rm -rf /root/.cache -FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:1.13.1-cpu-py39-ubuntu20.04-sagemaker as branch-cpu +FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py39-ubuntu20.04-sagemaker as branch-cpu ENV dev_type=CPU # Install DGL CPU version -RUN pip3 install dgl==1.0.4 -f https://data.dgl.ai/wheels-internal/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels-internal/repo.html && rm -rf /root/.cache FROM branch-${DEVICE} AS final From 4974e28efaa149afff1f5d1a26e1abdc0cf61644 Mon Sep 17 00:00:00 2001 From: JalenCato Date: Wed, 4 Sep 2024 23:00:15 +0000 Subject: [PATCH 2/7] change --- docker/sagemaker/Dockerfile.sm | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/docker/sagemaker/Dockerfile.sm b/docker/sagemaker/Dockerfile.sm index f39df95839..d2704e3d60 100644 --- a/docker/sagemaker/Dockerfile.sm +++ b/docker/sagemaker/Dockerfile.sm @@ -3,15 +3,15 @@ ARG DEVICE=gpu -FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py39-cu117-ubuntu20.04-sagemaker as branch-gpu +FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py311-cu121-ubuntu20.04-sagemaker as branch-gpu ENV dev_type=GPU # Install DGL GPU version -RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels/cu117/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache -FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py39-ubuntu20.04-sagemaker as branch-cpu +FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py311-ubuntu20.04-sagemaker as branch-cpu ENV dev_type=CPU # Install DGL CPU version -RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels-internal/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache FROM branch-${DEVICE} AS final From b873bd53d3dabd10ff8790748c0ce2b257659ffd Mon Sep 17 00:00:00 2001 From: jalencato Date: Thu, 12 Sep 2024 10:45:17 -0700 Subject: [PATCH 3/7] Update Dockerfile.sm --- docker/sagemaker/Dockerfile.sm | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docker/sagemaker/Dockerfile.sm b/docker/sagemaker/Dockerfile.sm index d2704e3d60..db760061b5 100644 --- a/docker/sagemaker/Dockerfile.sm +++ b/docker/sagemaker/Dockerfile.sm @@ -6,12 +6,12 @@ ARG DEVICE=gpu FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py311-cu121-ubuntu20.04-sagemaker as branch-gpu ENV dev_type=GPU # Install DGL GPU version -RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.3.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py311-ubuntu20.04-sagemaker as branch-cpu ENV dev_type=CPU # Install DGL CPU version -RUN pip3 install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.3.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache FROM branch-${DEVICE} AS final From 9a48428f7e1ca23013963805b02d54e1b95fbe6c Mon Sep 17 00:00:00 2001 From: jalencato Date: Mon, 30 Sep 2024 11:29:27 -0700 Subject: [PATCH 4/7] Update Dockerfile.sm --- docker/sagemaker/Dockerfile.sm | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docker/sagemaker/Dockerfile.sm b/docker/sagemaker/Dockerfile.sm index db760061b5..4620e3e079 100644 --- a/docker/sagemaker/Dockerfile.sm +++ b/docker/sagemaker/Dockerfile.sm @@ -6,7 +6,7 @@ ARG DEVICE=gpu FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py311-cu121-ubuntu20.04-sagemaker as branch-gpu ENV dev_type=GPU # Install DGL GPU version -RUN pip3 install dgl==2.3.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.3.0 -f https://data.dgl.ai/wheels/torch-2.3/cu121/repo.html && rm -rf /root/.cache FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py311-ubuntu20.04-sagemaker as branch-cpu ENV dev_type=CPU From 5561659f481a306c75cb7e8119e7d1df21db90f3 Mon Sep 17 00:00:00 2001 From: jalencato Date: Mon, 30 Sep 2024 11:30:32 -0700 Subject: [PATCH 5/7] Update Dockerfile.sm --- docker/sagemaker/Dockerfile.sm | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docker/sagemaker/Dockerfile.sm b/docker/sagemaker/Dockerfile.sm index 4620e3e079..27a95062b4 100644 --- a/docker/sagemaker/Dockerfile.sm +++ b/docker/sagemaker/Dockerfile.sm @@ -6,7 +6,7 @@ ARG DEVICE=gpu FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py311-cu121-ubuntu20.04-sagemaker as branch-gpu ENV dev_type=GPU # Install DGL GPU version -RUN pip3 install dgl==2.3.0 -f https://data.dgl.ai/wheels/torch-2.3/cu121/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==2.3.0+cu121 -f https://data.dgl.ai/wheels/torch-2.3/cu121/repo.html && rm -rf /root/.cache FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py311-ubuntu20.04-sagemaker as branch-cpu ENV dev_type=CPU From d2ebac9fb07e585daf5e04b9c128bb9f37cbbfdc Mon Sep 17 00:00:00 2001 From: jalencato Date: Mon, 30 Sep 2024 14:01:37 -0700 Subject: [PATCH 6/7] Update Dockerfile.sm --- docker/sagemaker/Dockerfile.sm | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/docker/sagemaker/Dockerfile.sm b/docker/sagemaker/Dockerfile.sm index 27a95062b4..7a0bd70672 100644 --- a/docker/sagemaker/Dockerfile.sm +++ b/docker/sagemaker/Dockerfile.sm @@ -2,16 +2,17 @@ # Note: Distributed graph partition will use another docker image which will come soon. ARG DEVICE=gpu +ARG DGL_VERSION=2.3.0 FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py311-cu121-ubuntu20.04-sagemaker as branch-gpu ENV dev_type=GPU # Install DGL GPU version -RUN pip3 install dgl==2.3.0+cu121 -f https://data.dgl.ai/wheels/torch-2.3/cu121/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==${DGL_VERSION}+cu121 -f https://data.dgl.ai/wheels/torch-2.3/cu121/repo.html && rm -rf /root/.cache FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py311-ubuntu20.04-sagemaker as branch-cpu ENV dev_type=CPU # Install DGL CPU version -RUN pip3 install dgl==2.3.0 -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache +RUN pip3 install dgl==${DGL_VERSION} -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache FROM branch-${DEVICE} AS final @@ -46,7 +47,7 @@ ENV PYTHONPATH="/opt/ml/code/graphstorm/python/:${PYTHONPATH}" RUN cp /opt/ml/code/graphstorm/sagemaker/run/* /opt/ml/code/ # Download DGL source code -RUN cd /root; git clone https://github.com/dmlc/dgl.git +RUN cd /root; git clone --branch v${DGL_VERSION} https://github.com/dmlc/dgl.git # Un-comment if we prefer a local DGL distribution # COPY dgl /root/dgl ENV PYTHONPATH="/root/dgl/tools/:${PYTHONPATH}" From b7dd39a6a747e19316959f41a90a6ea6b18289e4 Mon Sep 17 00:00:00 2001 From: jalencato Date: Mon, 30 Sep 2024 15:23:52 -0700 Subject: [PATCH 7/7] Update Dockerfile.sm --- docker/sagemaker/Dockerfile.sm | 2 ++ 1 file changed, 2 insertions(+) diff --git a/docker/sagemaker/Dockerfile.sm b/docker/sagemaker/Dockerfile.sm index 7a0bd70672..97df1ba223 100644 --- a/docker/sagemaker/Dockerfile.sm +++ b/docker/sagemaker/Dockerfile.sm @@ -6,11 +6,13 @@ ARG DGL_VERSION=2.3.0 FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-gpu-py311-cu121-ubuntu20.04-sagemaker as branch-gpu ENV dev_type=GPU +ARG DGL_VERSION # Install DGL GPU version RUN pip3 install dgl==${DGL_VERSION}+cu121 -f https://data.dgl.ai/wheels/torch-2.3/cu121/repo.html && rm -rf /root/.cache FROM 763104351884.dkr.ecr.us-east-1.amazonaws.com/pytorch-training:2.3.0-cpu-py311-ubuntu20.04-sagemaker as branch-cpu ENV dev_type=CPU +ARG DGL_VERSION # Install DGL CPU version RUN pip3 install dgl==${DGL_VERSION} -f https://data.dgl.ai/wheels/torch-2.3/repo.html && rm -rf /root/.cache