feat: added recipe for redshift-for-houdini on SMF #84
Conversation
yuanmich2
force-pushed
the
redshift4houdini_recipe
branch
from
149cf3b to
ea24ca2
Compare
…line#83) Signed-off-by: Joel Wong <[email protected]> Signed-off-by: Michael Yuan <[email protected]>
Signed-off-by: Michael Yuan <[email protected]>
Signed-off-by: Michael Yuan <[email protected]>
* Update apply-conda-queue-env.py to remove trailing whitespace.
* Clean up conda queue environment samples.
* Create virtual environments in directories with names randomized
by `mktemp`.
* Capture the environment activation by using `conda run` instead of
`conda activate`. Deactivate by calling eval "$(conda shell.bash deactivate)".
These changes let the environment work without `conda init`
changes working in non-interactive contexts.
* Use Python instead of bash $(()) expressions for math.
* Add `--quiet` option to `conda create` and `conda install`
commands, to remove incremental progress updates from the log.
* Add a `conda list` call to the 'improved_caching' env enter when
it is reusing an environment, so that it is easy to look at the
contents of the virtual environment from each job's log view.
* Fixed a failing case in the 'improved_caching' env exit, due to
grep return exit code 1 when no cached environments exist.
* Update the job_dev_progression README to include specific instructions
for both running the job on Deadline Cloud and running the job locally
with the Open Job Description CLI.
* Update the job dev progression code to use `mkdir -p` so that repeated
runs on the same workspace succeed.
Signed-off-by: Mark Wiebe <[email protected]>
Signed-off-by: Michael Yuan <[email protected]>
This sample comes in three parts: * A CloudFormation template to deploy a CUDA Deadline Cloud farm. The deployed farm is general to any Linux CUDA workload, and an easy way to get a farm for this sample. The deployed farm includes two queues, one for package building and another for running production jobs. * A NeRF Studio conda package recipe. While NeRF Studio is an open source project, it and many of its dependencies do not have packages in the conda-forge channel. This recipe builds a package called nerfstudio that bundles together these dependencies and NeRF Studio. * A job bundle containing a 3D Gaussian Splatting pipeline. The job bundle focuses on simplicity over features, containing just one task to run the whole pipeline. Each part includes README content with instructions on how to use it, and cross-links between the different parts so that if you land on any one of them you can navigate to the others. Signed-off-by: Mark Wiebe <[email protected]> Signed-off-by: Michael Yuan <[email protected]>
Signed-off-by: Mark Wiebe <[email protected]> Signed-off-by: Michael Yuan <[email protected]>
yuanmich2
force-pushed
the
redshift4houdini_recipe
branch
from
942a710 to
93923b2
Compare
|
|
||
| Recipe that builds Redshift for Houdini and integrates it with the existing Houdini package available to SMF workers. | ||
|
|
||
| This recipe largely follows the recommendations of the official docs on scripted installs: https://support.maxon.net/hc/en-us/articles/4687177301788-Scripted-Install-Instructions#h_01HHJ6B4A2PACXKKS7JY1M0V7D |
There was a problem hiding this comment.
Hmm, can we add some more description on how one can create the archive file for this recipe? Similar to something like this: https://github.com/aws-deadline/deadline-cloud-samples/blob/mainline/conda_recipes/cinema4d-2025/README.md
Without some instructions it can get a little hard for someone to build the recipe.
There was a problem hiding this comment.
Added some comments that would prevent someone from trying to build this on their own.
Feel free to take a look at other recipes in this folder and it can help you get more of an idea on how to write a recipe for Deadline Cloud. :)
| about: | ||
| homepage: https://redshift.maxon.net | ||
| summary: > | ||
| 3D computer animation, modeling, simulation, and rendering software |
There was a problem hiding this comment.
Can we add a deadline-cloud.yaml file to this recipe? Similar to https://github.com/aws-deadline/deadline-cloud-samples/blob/mainline/conda_recipes/cinema4d-openjd/deadline-cloud.yaml?
| version: ${{ version }} | ||
|
|
||
| source: | ||
| - path: /usr/redshift #the default redshift install path. Assumes redshift has been installed on this computer |
There was a problem hiding this comment.
Generally we follow a pattern where we store the archive files into the archive_files folder. See
This will allow a consistent pattern and allow us to upload the artifacts to deadline cloud to build it as well.
| export RED_LOCATION=\$CONDA_PREFIX/redshift | ||
|
|
||
| export REDSHIFT_COREDATAPATH=\$RED_LOCATION | ||
| export REDSHIFT_LOCALDATAPATH=\$CONDA_PREFIX/redshift_local_data |
There was a problem hiding this comment.
Hmm, any reason there are different path separators? \ and /?
There was a problem hiding this comment.
The \ characters are there to escape the $ characters. Only the / are path separators
What was the problem/requirement? (What/Why)
There was no example for Redshift on Houdini.
What was the solution? (How)
This pr adds a conda recipe for Redshift on Houdini
What is the impact of this change?
Now everyone can see an example of how to build a conda package for Redshift on Houdini
How was this change tested?
I performed some tests on my own farm
Was this change documented?
The sample comes with its own readme
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.