Add feed support for ParallelExecutor #9637
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
chengduoZH
reviewed
Apr 4, 2018
| @@ -135,7 +144,9 @@ def bottleneck_block(input, num_filters, stride, cardinality, reduction_ratio): | |||
| return fluid.layers.elementwise_add(x=short, y=scale, act='relu') | |||
|
|
|||
|
|
|||
| def SE_ResNeXt152Small(batch_size=2): | |||
| def SE_ResNeXt152Small(batch_size=2, use_feed=False): | |||
There was a problem hiding this comment.
SE_ResNeXt152 can be replaced with SE_ResNeXt50, the example is here. SE_ResNeXt50 consumes smaller memory than SE_ResNeXt152. And also be faster.
There was a problem hiding this comment.
This change is not from this PR, let's leave this as a followup PR.
| or numpy array. | ||
| :return: fetched value list. | ||
| """ | ||
| feed_tensor_dict = {} |
There was a problem hiding this comment.
Please check the type of feed_dict.
| feed_tensor = feed_dict[feed_name] | ||
| if not isinstance(feed_tensor, core.LoDTensor): | ||
| feed_tensor = core.LoDTensor() | ||
| feed_tensor.set(feed_dict[feed_name], self._act_places[0]) |
There was a problem hiding this comment.
There are two ways that feeding data:
- All the tensors are transferred to GPU(0) and then transferring sub data to other GPUs by P2P.
- All the tensors are in CPU side and then transferring sub data to GPUs by CPU->GPU
Which is better?
There was a problem hiding this comment.
We can adopt the first, parallel_do also use this way.
There was a problem hiding this comment.
Yes, I believe this is the current solution. set() place the tensor in GPU0, SplitLoDTensor transfer them to each GPUs.
|
I think is ok, how about you @qingqing01 ? |
chengduoZH
approved these changes
Apr 4, 2018
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.
TODO: Need to work with reader when both reader and
feed are enabled. Normally feed replaces reader input
or var produced from upstream ops.