Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Develop upstream sync 230731 #2170

Merged
merged 453 commits into from
Aug 7, 2023
Merged

Develop upstream sync 230731 #2170

merged 453 commits into from
Aug 7, 2023

Conversation

weihanmines
Copy link

No description provided.

tensorflower-gardener and others added 30 commits July 26, 2023 09:08
@tensorflower-gardener
Added attributed to IFRT CallOp that specifies if an atom program is …
63abf15 
…in global or local view.

If the attribute is set on a CallOp, then verification logic converts the programs arguments and results from local view to global view to verify that local view shape + sharding is equivalent to the expected global view shape.

PiperOrigin-RevId: 551222813
@tensorflower-gardener
Add python and numpy headers to the local_config_python folder in the…
a034b3d 
… wheel.

PiperOrigin-RevId: 551224665
@tensorflower-gardener
Integrate LLVM at llvm/llvm-project@365d6eb1f7d8
daa9a34 
Updates LLVM usage to match
[365d6eb1f7d8](llvm/llvm-project@365d6eb1f7d8)

PiperOrigin-RevId: 551229328
@fergushenderson @tensorflower-gardener
Update sample_stable_delegate for promotion of experimental/accelerat…
f178576 
…ion/configuration out of experimental.

PiperOrigin-RevId: 551235514
@qlzh727 @tensorflower-gardener
Increase the memory limit for the dtensor GPU test.
0ef963c 
PiperOrigin-RevId: 551238120
@mihaimaruseac
Remove trigraph
0321ee1
@mihaimaruseac
Remove trigraph
cc5aa34
@fergushenderson @tensorflower-gardener
Remove unnecessary 'const' from pass-by-value function parameters.
7ba36c1 
Also fix typo in SetAllowBufferHandleOutput comment: false->true.
Also fix #include order to match style guide.

PiperOrigin-RevId: 551247708
@ddunl @tensorflower-gardener
Internal visibility change only.
d9ec8c5 
PiperOrigin-RevId: 551261650
@trisolaran @tensorflower-gardener
special allocations' aggregated metrics need to consider memory color.
cf4afb6 
PiperOrigin-RevId: 551275563
@trisolaran @tensorflower-gardener
deprecate instruction name, it is changed over 1 years ago.
7f8be6e 
PiperOrigin-RevId: 551276374
@LionOfJewdah @tensorflower-gardener
Add macros for working with TF_Status in C++ code
0cc2c30 
`TF_STATUS_ASSIGN_OR_RETURN` and `TF_STATUS_RETURN_IF_ERROR`

PiperOrigin-RevId: 551278625
@tensorflower-gardener
Correct the device assignment for tf._XlaCompile
3874ea2 
PiperOrigin-RevId: 551290442
@mpcallanan @tensorflower-gardener
#tf-data-service Graduate "data_transfer" experiment.
69541bb 
PiperOrigin-RevId: 551295403
@tensorflower-gardener
Merge pull request tensorflow#60026 from milpuz01:node_rewrite_to_mkl…
2d04d5c 
…_heuristics

PiperOrigin-RevId: 551297374
@tensorflower-gardener
Potential improvement opportunity to eliminate extra transpose op.
c2a9dd3 
This CL will add patterns to fold Transpose and FC to covert into a BMM, like below-

FC(lhs, Transpose(rhs)) -> BMM(lha, rhs, false, false)

The right thing to do in this pattern will be to apply the pattern only if keep_num_dims==True. Because, if the output rank is less-than the input rank, it means `keep_num_dims` has reduced the output. But checking for rank will improve the coverage. This pattern will now work

PiperOrigin-RevId: 551297769
@tensorflower-gardener
Merge pull request tensorflow#61176 from tensorflow:pjpratik-patch-6
f7e9b91 
PiperOrigin-RevId: 551313932
@junwhanahn @tensorflower-gardener
Remove one use of inlining in XlaCallModule shape refinement.
f9e2045 
To improve debuggability, we want the shape refinement to make as few changes as possible to the module. In this change we remove one use of inlining.

PiperOrigin-RevId: 551325242
@tensorflower-gardener
Merge consecutive Pad operators
fe33928 
PiperOrigin-RevId: 551347216
@tensorflower-gardener
update internal files for release
1c26b1c 
PiperOrigin-RevId: 551353292
@skye @tensorflower-gardener
[XLA:Python] Use int64_t instead of ssize_t (part 2) to fix another M…
0d22c4b 
…ac compiler error

Apparently ssize_t is only a long sometimes (at least 32-bit), instead
of long long (at least 64-bit). I don't have a mac so I can't repro
the failing build, but hopefully this fixes it based on the error
message.

PiperOrigin-RevId: 551376003
@tensorflower-gardener
Added a workaround for broadcast.
6021733 
PiperOrigin-RevId: 551401683
@cezheng @tensorflower-gardener
[PJRT] Add PjRtDevice::PoisonExecution.
54eeb2a 
PiperOrigin-RevId: 551408554
@tensorflower-gardener
[IFRT] Update ShardingParam to also support scalars.
333bb69 
PiperOrigin-RevId: 551410772
@tensorflower-gardener
Merge pull request tensorflow#61384 from elfringham:fewer_jobs
60574d4 
PiperOrigin-RevId: 551413305
@anlunx @tensorflower-gardener
[xla:gpu] Fix the BFS algorithm in dataflow analysis
b03ea08 
The BFS algorithm didn't have a visited set, therefore had a complexity of O(N*E).

PiperOrigin-RevId: 551414282
@dansuh17 @tensorflower-gardener
Add uniform quantized stablehlo.convolution -> tfl.conv_2d conver…
d5422e3 
…sion pattern.

This change implements a conversion pattern that converts stablehlo.convolution to tfl.conv_2d.
This is a minimal version that converts quantized `stablehlo.convolution` with certain assumptions like that the filter has the format of `[0, 1, i, o]`.

PiperOrigin-RevId: 551419638
@tensorflower-gardener
[XLA] Update PatternMatchUnmergeSharding to avoid illegal function ca…
7d50115 
…ll on SplitShardingDimension.

PiperOrigin-RevId: 551438695
@bchetioui @tensorflower-gardener
[XLA:GPU] Enable using region analysis in the CopyInsertion pass.
9b2016c 
This is in preparation for another change improving the state of copies
in while loops.

PiperOrigin-RevId: 551451818
@tensorflower-gardener
Add a tflite model containing tensors that store string data.
e5e0228 
PiperOrigin-RevId: 551452964
bixia1 and others added 12 commits July 31, 2023 16:21
@bixia1 @tensorflower-gardener
[xla] Make Send a control predecessor of Recv-done in the generated S…
bc68c98 
…end-Recv

sequence.

This is to prevent the latency hiding scheduler to interleave two Send-Recv
sequences.

PiperOrigin-RevId: 552621536
@IMMZEK @tensorflower-gardener
Fixed conflict in graph_execution_options wrapper.
cd1cdc8 
PiperOrigin-RevId: 552621643
@tensorflower-gardener
Remove setting up GCS in FindAndLoadTpuLibrary.
3c4627f 
InitializeCreateGcsFileSystemFnPtr is a temporary fix and it is no longer needed.

PiperOrigin-RevId: 552624923
@tensorflower-gardener
[XLA:GPU] Update intercept check of DUS and Copy in LiveRangeRegion A…
73a54a8 
…nalysis

PiperOrigin-RevId: 552625083
@tensorflower-gardener
Merge pull request tensorflow#61428 from elfringham:fix_xla_lit
0eed700 
PiperOrigin-RevId: 552626000
@hhb @tensorflower-gardener
Update Rendezvous API to not depend on :tf_status
b769fc4 
PiperOrigin-RevId: 552631765
@tensorflower-gardener
Add a device count cache to CudaPlatform.
67dbc78 
This removes some unnecessary `cuDeviceGetCount()` calls when custom ops are used.

PiperOrigin-RevId: 552634342
@tensorflower-gardener
Add input type inference, conversion and WeakTensor construction logi…
32803d8 
…c in tf.constant according to auto dtype conversion semantics.

WeakTensor is created if it satisfies both of the following conditions:
1. tf.constant is called with no dtype arg specified.
2. Input is a nested Python type.

PiperOrigin-RevId: 552634845
@chhe7 @tensorflower-gardener
Creates an optimization to fuse transpose and reshape into batch_mat…
9d0fea2 
…mul.

PiperOrigin-RevId: 552636662
@tensorflower-gardener
When sharding propagation does not return an input sharding for opera…
d1280be 
…nd 0 of a gather, assume that the sharding of that operand does not matter.

PiperOrigin-RevId: 552637713
@weihanmines
weekly sync 230731 before solving conflicts
e1de22b
@weihanmines
weekly sync 230731 after solving conflicts
da2cefb
@i-chaochen
Copy link

Since I did the cover here https://github.com/openxla/xla/pull/4603/files#diff-fc02eb6aea06ad0d72011e9a64da28e8c6fae000a9e0cab9b15b40b4e914af4aR956-R958

could you try to enable triton-softmax flag to have a go, please?
da2cefb#diff-04cb485c1774fda54f8346ece0ade9efbcd813235768905b234d596c22660cf6R170

we need to see are any other unit tests affected on that flag.

i-chaochen
i-chaochen reviewed Aug 1, 2023
View reviewed changes
Copy link

@i-chaochen i-chaochen left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Since gpu.graph PR is merged, you should enable xla_graph_level = 1

@weihanmines
Copy link
Author

Since gpu.graph PR is merged, you should enable xla_graph_level = 1

turn it on in the latest commit.

@weihanmines
Copy link
Author

Since I did the cover here https://github.com/openxla/xla/pull/4603/files#diff-fc02eb6aea06ad0d72011e9a64da28e8c6fae000a9e0cab9b15b40b4e914af4aR956-R958

could you try to enable triton-softmax flag to have a go, please? da2cefb#diff-04cb485c1774fda54f8346ece0ade9efbcd813235768905b234d596c22660cf6R170

we need to see are any other unit tests affected on that flag.

turn it on in the latest commit.

@weihanmines
Copy link
Author

Jenkins: retest Ubuntu-GPU-single please

i-chaochen
i-chaochen reviewed Aug 3, 2023
View reviewed changes
tensorflow/compiler/xla/service/gpu/BUILD
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is added in tensorflow@56f261b we need a ticket to track this test

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sure.

@weihanmines weihanmines merged commit 7a70b16 into develop-upstream Aug 7, 2023
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@i-chaochen i-chaochen i-chaochen left review comments

@jayfurmanek jayfurmanek jayfurmanek approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.