Fix model layer for resnet v1. #4230
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
The final BN and ReLU layer is only need for v2 model since it was doing preactivation in each block.
|
How was this missed? Or, to put that another way, why didn't this have any effect in training, etc.? Does this fix the loss scale issue? |
|
Sadly this does not resolve the divergent fp16 v1 training. |
|
@karmel, the extra ReLU shouldn't impact the correctness of the model. The BN will affect the number to let the final number scale down further. The fp16 problem in v1 is because of overflow of the number on the top few layer of the model. Removing the final BN and ReLU won't fix it. |
|
have the pretrained model of resnet-v1 been replaced ? |
|
@HiKapok, the pretrained model has not been updated yet. Will do in later this week. |
omegafragger
pushed a commit
to omegafragger/models
that referenced
this pull request
May 15, 2018
The final BN and ReLU layer is only need for v2 model since it was doing preactivation in each block.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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Outdated suggestions cannot be applied.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.
The final BN and ReLU layer is only need for v2 model since it was
doing preactivation in each block.