Skip to content

Commit 9b42768

Browse files
remove uv changes
1 parent 1843628 commit 9b42768

File tree

3 files changed

+12
-20
lines changed

3 files changed

+12
-20
lines changed

buildspec.yml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@ phases:
1010
- start-dockerd
1111
- echo Logging in to Amazon ECR...
1212
- $(aws ecr get-login --no-include-email --region $AWS_DEFAULT_REGION)
13-
- uv pip install -r src/requirements.txt
13+
- pip install -r src/requirements.txt
1414
- bash src/setup.sh $FRAMEWORK
1515
- python src/parse_partner_developers.py
1616
- python src/send_status.py --status 2

pytorch/training/docker/2.8/py3/cu129/Dockerfile.gpu

Lines changed: 8 additions & 17 deletions
Original file line numberDiff line numberDiff line change
@@ -53,8 +53,6 @@ ENV PYTHONUNBUFFERED=1
5353
ENV PYTHONIOENCODING=UTF-8
5454
ENV LANG=C.UTF-8
5555
ENV LC_ALL=C.UTF-8
56-
ENV UV_VERSION=0.8.22
57-
ENV UV_SYSTEM_PYTHON=1
5856

5957
ENV TORCH_NVCC_FLAGS="-Xfatbin -compress-all"
6058

@@ -82,15 +80,8 @@ ENV LD_LIBRARY_PATH="/usr/local/lib:/opt/amazon/ofi-nccl/lib/x86_64-linux-gnu:/o
8280
# Python Path
8381
ENV PATH="/usr/local/bin:${PATH}"
8482

85-
86-
ENV UV_CACHE_DIR=/root/.cache/uv \
87-
UV_SYSTEM_PYTHON=1
88-
89-
RUN curl -LsSf https://astral.sh/uv/install.sh | sh \
90-
&& ln -s /root/.local/bin/uv /usr/local/bin/uv
91-
9283
# Install common conda packages
93-
RUN uv pip install --no-cache-dir \
84+
RUN pip install --no-cache-dir \
9485
cython \
9586
cryptography \
9687
pyOpenSSL \
@@ -123,11 +114,11 @@ RUN uv pip install --no-cache-dir \
123114
tornado>=6.5.1
124115

125116
# Install PyTorch
126-
RUN uv pip install --no-cache-dir -U torch==${PYTORCH_VERSION} \
117+
RUN pip install --no-cache-dir -U torch==${PYTORCH_VERSION} \
127118
torchvision==${TORCHVISION_VERSION} \
128119
torchaudio==${TORCHAUDIO_VERSION} \
129120
--index-url https://download.pytorch.org/whl/cu129 \
130-
&& uv pip install --no-cache-dir -U torchtnt==${TORCHTNT_VERSION} \
121+
&& pip install --no-cache-dir -U torchtnt==${TORCHTNT_VERSION} \
131122
torchdata==${TORCHDATA_VERSION} \
132123
triton \
133124
s3torchconnector \
@@ -140,18 +131,18 @@ RUN uv pip install --no-cache-dir -U torch==${PYTORCH_VERSION} \
140131
thinc==8.3.4 \
141132
blis \
142133
numpy \
143-
&& uv pip uninstall dataclasses
134+
&& pip uninstall -y dataclasses
144135

145136
# Install flash attn and NVIDIA transformer engine.
146137
# Optionally set NVTE_FRAMEWORK to avoid bringing in additional frameworks during TE install
147138
ENV NVTE_FRAMEWORK=pytorch
148139

149140
RUN curl -LO https://github.com/Dao-AILab/flash-attention/releases/download/v${FLASH_ATTN_VERSION}/flash_attn-${FLASH_ATTN_VERSION}+cu12torch2.8cxx11abiTRUE-cp312-cp312-linux_x86_64.whl \
150-
&& uv pip install flash_attn-${FLASH_ATTN_VERSION}+cu12torch2.8cxx11abiTRUE-cp312-cp312-linux_x86_64.whl --no-build-isolation \
141+
&& pip install flash_attn-${FLASH_ATTN_VERSION}+cu12torch2.8cxx11abiTRUE-cp312-cp312-linux_x86_64.whl --no-build-isolation \
151142
&& rm flash_attn-${FLASH_ATTN_VERSION}+cu12torch2.8cxx11abiTRUE-cp312-cp312-linux_x86_64.whl
152143

153144
# Install TE using instructions from https://docs.nvidia.com/deeplearning/transformer-engine/user-guide/installation.html
154-
RUN uv pip install --no-cache-dir git+https://github.com/NVIDIA/TransformerEngine.git@release_v${TE_VERSION} --no-build-isolation
145+
RUN pip install --no-cache-dir git+https://github.com/NVIDIA/TransformerEngine.git@release_v${TE_VERSION} --no-build-isolation
155146

156147
RUN curl -o /license.txt https://aws-dlc-licenses.s3.amazonaws.com/pytorch-2.8/license.txt
157148

@@ -254,15 +245,15 @@ ARG PYTHON
254245
WORKDIR /
255246

256247
# Install SM packages
257-
RUN uv pip install --no-cache-dir -U \
248+
RUN pip install --no-cache-dir -U \
258249
smclarify \
259250
"sagemaker>=2" \
260251
sagemaker-experiments \
261252
sagemaker-pytorch-training \
262253
sagemaker-training
263254

264255
# Install extra packages
265-
RUN uv pip install --no-cache-dir -U \
256+
RUN pip install --no-cache-dir -U \
266257
bokeh \
267258
imageio \
268259
numba \

tensorflow/training/docker/2.18/py3/Dockerfile.cpu

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -264,7 +264,7 @@ RUN ${PIP} install --no-cache-dir -U \
264264
"tensorflow-datasets==4.9.7"
265265

266266
RUN $PYTHON -m pip install --no-cache-dir -U \
267-
numba \
267+
numba==0.61.0 \
268268
bokeh \
269269
imageio \
270270
opencv-python \
@@ -279,8 +279,9 @@ RUN $PYTHON -m pip install --no-cache-dir -U \
279279
RUN $PYTHON -m pip install --no-cache-dir -U \
280280
sagemaker-experiments==0.1.45
281281

282+
#pinning old version because of protobuf dependency with tensorflow-metadata
282283
RUN $PYTHON -m pip install --no-cache-dir -U \
283-
sagemaker-training
284+
sagemaker-training==4.8.4
284285

285286
RUN $PYTHON -m pip install --no-cache-dir -U \
286287
sagemaker-tensorflow-training==20.4.1

0 commit comments

Comments
 (0)