This repository has been archived by the owner on May 1, 2023. It is now read-only.
IFM sparsity collector #443
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
Collect IFM sparsity by SummaryActivationStatsCollector when in_or_out kwarg is set to `in`, versus OFM when set to 'out'. Use `torch.cat` to merge the tuple of inputs into single tensor. In the case where there're no inputs, it will fail. Rename argument name from singular 'input' to plural 'inputs' in _activation_stats_cb. Rename the former 'sparsity' collector to 'sparsity_ofm'.
Also, omit stat_name from module_name. e.g. sparsity_module.conv1 -> module.conv1
When True, collect OFM sparsity, otherwise, IFMs sparsity
|
@nzmora I would like to replace |
nzmora
suggested changes
Dec 15, 2019
nzmora
approved these changes
Dec 17, 2019
| 'collect_quant_stats', 'collect_histograms', | ||
| 'collector_context', 'collectors_context'] | ||
|
|
||
| class CollectorDirection(enum.Enum): |
There was a problem hiding this comment.
can you add a documentation line explaining what IN/OUT and IFM/OFM (input/output feature maps) are?
michaelbeale-IL
pushed a commit
that referenced
this pull request
Apr 24, 2023
Add directionality to SummaryActivationStatsCollector to allow collection of statistics on incoming and outgoing activations/feature-maps; instead of just outgoing activations. Also includes some code refactoring.
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
Patch adds support to collecting IFM sparsity, as well as multiple IFMs. If multiple IFMs are present, they are merged with
torch.cat, which is a parameter to the collector.IFM sparsity will be collected whenever activation sparsity is collected with
create_activation_stats_collectors.Renamed some of the fields/filenames.
Tested against Python3.6. Passed
excluding the post-training tests.(merged @guyjacob bugfix)Note the use of typing library for type hinting. Not sure about the state of support for this in older Python 3.5 environments. It's possible that this patch requires Python3.6 OR Python3.5+updates, but not sure how to check that.