This repository has been archived by the owner on Jul 1, 2023. It is now read-only.
Accurate Tensor.device for TFEager backends
#1077
Closed
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
texasmichelle
commented
Sep 11, 2020
pschuh
approved these changes
Sep 11, 2020
| // Parse type and ordinal from a string with the expected syntax: | ||
| // /job:localhost/replica:0/task:0/device:CPU:0 | ||
| let pattern = ".+device:(.+):(\\d+)$" | ||
| let regex = try! NSRegularExpression(pattern: pattern) |
There was a problem hiding this comment.
This string parsing looks expensive. Might want to double check that this doesn't happen too much.
texasmichelle
changed the title
Device.defaultTFEagerTensor.device for TFEager backends
pschuh
reviewed
Sep 11, 2020
|
|
||
| // Parse type and ordinal from a string with the expected syntax: | ||
| // /job:localhost/replica:0/task:0/device:CPU:0 | ||
| let pattern = ".+device:(.+):(\\d+)$" |
There was a problem hiding this comment.
Maybe break this String -> Device out as a separate function?
Also, I'm concerned that the string parsing will be expensive. This function is called a lot. (whenever there is a scalar constant or anything like that). I think it would be best to see if you can add your own TFE_TensorHandleDevice_Type and TFE_TensorHandleDevice_Id. Some benchmarking results might work instead.
|
Different approach coming. Closing this PR |
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Outdated suggestions cannot be applied.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.
This modifies
Tensor.devicefor TFEager backends to retrieve device details from the underlying eager libraries. This is more accurate than returningDevice.defaultTFEagerand reflects the actual location of eager operations.This was tested by building a CUDA toolchain and verifying that
TFE_TensorHandleDeviceName()returns:and a GPU device is returned by
Tensor.device:On a device without GPU,
TFE_TensorHandleDeviceName()returns:and a CPU device is returned by
Tensor.device:Fixes tensorflow/swift#524.