This repository was archived by the owner on Apr 28, 2023. It is now read-only.
Add ReLU + masked convolution #571
Merged
Conversation
This file contains hidden or 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
python/examples/masked_conv.py
Outdated
| .devices(args.tuner_devices)) | ||
|
|
||
| # This function is used for reinforcing tuning | ||
| # 1. make_idx is small and does not get tuned or saved, just using naive |
Contributor
There was a problem hiding this comment.
These comments are actually from tum.py
python/examples/masked_conv.py
Outdated
| output.cpu(), | ||
| Output.cpu(), | ||
| input_data.cpu(), conv.weight.cpu(), mask.cpu(), | ||
| precision=1e-5) |
Contributor
There was a problem hiding this comment.
So this is going to be 1e-5 relative precision.
Maybe it is better to write:
tc.assert_almost_equal(
output.cpu(),
Output.cpu(),
input_data.cpu(), conv.weight.cpu(), mask.cpu(),
operations= C * K * K,
precision=1e-7)
This means that we have 1e-7 relative precision (i.e. machine precision) modulo C * K * K FP representation error (the number of accumulations in a single memory element).
At training time, CondenseNets use a convolution in which certain input-output connections (that is, the entire kernel) are masked out. This example implements a combination of that masked convolution with a ReLU activations. The goal is to further work out this example into increasingly large chunks of the CondenseNet. Tested this internally: tests pass.
nicolasvasilache
approved these changes
Jul 16, 2018
Contributor
nicolasvasilache
left a comment
nicolasvasilache
left a comment
There was a problem hiding this comment.
Cool, thanks for your contribution!
Please wait for CI to be green before merging
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
3 participants
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.
At training time, CondenseNets use a convolution in which certain input-output connections (that is, the entire kernel) are masked out. This example implements a combination of that masked convolution with a ReLU activations.
The goal is to further work out this example into increasingly large chunks of the CondenseNet.
Tested this internally: tests pass.