Ab tensorize speed up #552
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…thin docker rather than parallel docker - about a 10x speed up. Added hdf5-tools to the docker for testing the hd5 files. Added validate_tensors.sh that can test that a generated hd5 file is complete.
lucidtronix
approved these changes
Feb 6, 2024
There was a problem hiding this comment.
Thanks @abaumann one questions inline but feel free to merge!
|
|
||
| find ${INPUT_TENSORS_DIR}/*.hd5 | \ | ||
| xargs -P ${NUMBER_OF_THREADS} -I {} \ | ||
| bash -c "h5dump -n {} | (grep -q 'HDF5 \"{}\"' && echo 'OK - {}' || echo 'BAD - {}')" |
There was a problem hiding this comment.
How does the script know if an HD5 bad here?
There was a problem hiding this comment.
yeah this was one thing I wanted to check about, but basically it looks for output which starts with "HDF5" and then the file path and then information inside the file, so all it really does is check that h5dump doesn't exit - that seems to indicate it's not corrupt, but doesn't check anything about the actual contents.
| ${DOCKER_IMAGE} /bin/bash -c "pip3 install --upgrade pip | ||
| pip install ${WORKDIR}; | ||
| ${DOCKER_IMAGE} /bin/bash -c "pip install --quiet --upgrade pip | ||
| pip install --quiet ${WORKDIR}; |
… scaling instance types and debugging tensorization.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Modified how tensorize.sh works to speed it up about 10x - originally this was running a batch of samples in serial in a docker container, and running multiple docker containers in parallel. I noticed this was using very little cpu, memory, or disk. Changed it to istead use xargs -P to run in parallel in a single docker container - the CPU/disk/mem all stay at nearly maxed out with the right combination of settings (e.g. on a 22 core machine I did 20 in parallel). Memory usage was low, so I'll update docs about what machine specs to use/what flags to use.
As part of this, I added a few other things: