Upgrading WML CE
Find instructions to upgrade to WML CE 1.6.2.
Upgrading from WML CE 1.6.1
WML CE 1.6.1 requires a similar GPU driver version (418) as PowerAI 1.6.2. We recommend updating to the current GPU driver before upgrading, but WML CE 1.6.2 should work with the older 418 GPU driver as well. Find the recommended GPU driver version here: WML CE prerequisites.
WML CE 1.6.2 conda packages are delivered in the same conda channel as 1.6.1, so no additional conda channels are needed.
There are generally two paths to upgrade from 1.6.1 to 1.6.2:
Create new 1.6.2 environment alongside 1.6.1
WML CE can be installed in a separate conda environment alongside an existing 1.6.1 environment:
conda create -y -n my-162-env python=3.6 powerai=1.6.2
conda activate my-162-env
Upgrade an existing 1.6.1 environment to 1.6.2
An existing environment containing WML CE 1.6.1 can be upgraded to 1.6.2:
- Activate the base conda environment:
conda activate base
- Activate the 1.6.2
environment:
conda activate my-162-env
- Remove any version pinning for WML CE:
Edit file
$CONDA_PREFIX/conda-meta/pinned
and remove any version pinning entries for WML CE components. - Upgrade to 1.6.2
- If you installed all of WML CE, from your
1.6.2 environment,
run:
orconda update --prune powerai
conda install powerai=1.6.2
- If you only installed selected packages, from your 1.6.2 environment,
run:
orconda update --prune powerai-release
conda install powerai-release=1.6.2
- If you installed all of WML CE, from your
1.6.2 environment,
run:
- (Optional) Pin WML CE version 1.6.2
Pin release 1.6.2 in the environment to avoid unintentional upgrades to a new release:
echo "powerai-release=1.6.2" >> $CONDA_PREFIX/conda-meta/pinned
- The 4.7.x series of the conda tool has brought upgrades to both the package indexing function as well as the dependency solver. Since this new function relies on new package metadata, it may not properly upgrade environments that were created using the 4.6.x series of conda, which lack the new metadata entries. Upgrading WML CE 1.6.1 environments that were created with conda 4.6 should be upgraded with conda 4.6.
- Although conda 4.6 is recommended when upgrading from WML CE 1.6.1 to 1.6.2, if conda gets updated to 4.7 and
you want to upgrade WML CE to 1.6.2, run
conda install powerai=1.6.2
with strict channel priority. - The dependency solver in the 4.6.x series of the conda tool still considered
Anaconda's deprecated
free
channel when working through the dependency requirements. When upgrading with conda 4.6, be sure to enable strict channel priority or remove the free channel from the defaults channel list.Set strict channel priority:conda config --set channel_priority strict
Remove the Anaconda Free channel from the default channels:conda config --add default_channels https://repo.anaconda.com/pkgs/main conda config --add default_channels https://repo.anaconda.com/pkgs/r
- In certain instances, (if an environment only has
tf_cnn_benchmarks
ortensorflow-probability
installed) you must include thetensorflow-gpu
package with the conda upgrade command to ensure that packages are upgraded with the versions built for GPU support:conda upgrade --prune tensorflow-gpu powerai-release
conda upgrade --prune tensorflow-gpu powerai-release
Upgrading from PowerAI 1.6.0
Upgrading from PowerAI 1.5.4
To upgrade from PowerAI 1.5.4 to WML CE 1.6.2, you must do the following: