Skip to content

Commit 7044c17

Browse files
authored
Patch PT 2.4 Training SM DLC (#4457)
* Patch PT 2.4 Training SM DLC * pin blis * build cpu pip install blis * build test all * build all * revert dockerfile and toml
1 parent 9cb4ebf commit 7044c17

File tree

4 files changed

+14
-8
lines changed

4 files changed

+14
-8
lines changed

pytorch/training/docker/2.4/py3/Dockerfile.cpu

Lines changed: 5 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -306,13 +306,16 @@ RUN /opt/conda/bin/mamba install -y -c conda-forge \
306306
shap \
307307
# pinned for sagemaker==2.235.1
308308
"cloudpickle==2.2.1" \
309+
&& /opt/conda/bin/mamba clean -afy
310+
311+
RUN pip install --no-cache-dir -U \
309312
# pin numpy requirement for sagemaker dependency
310313
# requires explicit declaration of spacy, thic, blis
314+
# pip install due to pip check conflict with conda package
311315
spacy \
312316
thinc \
313317
blis \
314-
"numpy<2" \
315-
&& /opt/conda/bin/mamba clean -afy
318+
"numpy<2"
316319

317320
# Copy workaround script for incorrect hostname
318321
COPY changehostname.c /

pytorch/training/docker/2.4/py3/Dockerfile.sagemaker.cpu.core_packages.json

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,14 +1,14 @@
11
{
22
"accelerate": {
3-
"version_specifier": "==1.1.1",
3+
"version_specifier": "==1.2.0",
44
"skip": "True"
55
},
66
"fastai": {
77
"version_specifier": "==2.7.18",
88
"skip": "True"
99
},
1010
"s3torchconnector": {
11-
"version_specifier": "==1.2.7",
11+
"version_specifier": "==1.3.0",
1212
"skip": "True"
1313
},
1414
"torchaudio": {

pytorch/training/docker/2.4/py3/cu124/Dockerfile.gpu

Lines changed: 5 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -488,13 +488,16 @@ RUN /opt/conda/bin/mamba install -y -c conda-forge \
488488
seaborn \
489489
# pinned for sagemaker==2.235.1
490490
"cloudpickle==2.2.1" \
491+
&& /opt/conda/bin/mamba clean -afy
492+
493+
RUN pip install --no-cache-dir -U \
491494
# pin numpy requirement for sagemaker dependency
492495
# requires explicit declaration of spacy, thic, blis
496+
# pip install due to pip check conflict with conda package
493497
spacy \
494498
thinc \
495499
blis \
496-
"numpy<2" \
497-
&& /opt/conda/bin/mamba clean -afy
500+
"numpy<2"
498501

499502
# Add SageMaker DataParallel to LD_LIBRARY_PATH
500503
ENV LD_LIBRARY_PATH="/opt/conda/lib/python${PYTHON_SHORT_VERSION}/site-packages/smdistributed/dataparallel/lib:$LD_LIBRARY_PATH"

pytorch/training/docker/2.4/py3/cu124/Dockerfile.sagemaker.gpu.core_packages.json

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
11
{
22
"accelerate": {
3-
"version_specifier": "==1.1.1",
3+
"version_specifier": "==1.2.0",
44
"skip": "True"
55
},
66
"fastai": {
@@ -12,7 +12,7 @@
1212
"skip": "True"
1313
},
1414
"s3torchconnector": {
15-
"version_specifier": "==1.2.7",
15+
"version_specifier": "==1.3.0",
1616
"skip": "True"
1717
},
1818
"torchaudio": {

0 commit comments

Comments
 (0)