-
Notifications
You must be signed in to change notification settings - Fork 498
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Support expanding ragged all-to-all dims similar to all-to-alls.
PiperOrigin-RevId: 715831514
- Loading branch information
1 parent
10daba6
commit 370a76e
Showing
4 changed files
with
203 additions
and
0 deletions.
There are no files selected for viewing
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
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
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
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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,111 @@ | ||
| /* Copyright 2025 The OpenXLA Authors. | ||
| Licensed under the Apache License, Version 2.0 (the "License"); | ||
| you may not use this file except in compliance with the License. | ||
| You may obtain a copy of the License at | ||
| http://www.apache.org/licenses/LICENSE-2.0 | ||
| Unless required by applicable law or agreed to in writing, software | ||
| distributed under the License is distributed on an "AS IS" BASIS, | ||
| WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| See the License for the specific language governing permissions and | ||
| limitations under the License. | ||
| ==============================================================================*/ | ||
|
|
||
| #include "xla/service/all_to_all_decomposer.h" | ||
|
|
||
| #include <memory> | ||
| #include <string> | ||
| #include <vector> | ||
|
|
||
| #include <gmock/gmock.h> | ||
| #include <gtest/gtest.h> | ||
| #include "xla/hlo/ir/hlo_opcode.h" | ||
| #include "xla/hlo/parser/hlo_parser.h" | ||
| #include "xla/hlo/utils/hlo_matchers.h" | ||
| #include "xla/tests/hlo_test_base.h" | ||
| #include "xla/tsl/platform/statusor.h" | ||
| #include "xla/util.h" | ||
|
|
||
| namespace xla { | ||
| namespace { | ||
|
|
||
| using AllToAllDecomposerTest = HloTestBase; | ||
| using ::testing::_; | ||
| namespace op = xla::testing::opcode_matchers; | ||
|
|
||
| TEST_F(AllToAllDecomposerTest, RaggedAllToAllRank1) { | ||
| const std::string module_str = | ||
| R"(HloModule RaggedAllToAll | ||
| ENTRY AllToAll { | ||
| p0 = s32[8]{0} parameter(0) | ||
| c0 = s32[] constant(0) | ||
| output = s32[8]{0} broadcast(c0), dimensions={} | ||
| p1 = s32[4]{0} parameter(1) | ||
| p2 = s32[4]{0} parameter(2) | ||
| p3 = s32[4]{0} parameter(3) | ||
| p4 = s32[4]{0} parameter(4) | ||
| input = s32[8]{0} copy(p0) | ||
| input_offsets = s32[4]{0} copy(p1) | ||
| send_sizes = s32[4]{0} copy(p2) | ||
| output_offsets = s32[4]{0} copy(p3) | ||
| recv_sizes = s32[4]{0} copy(p4) | ||
| ra2a = s32[8]{0} ragged-all-to-all(input, output, input_offsets, send_sizes, output_offsets, recv_sizes), replica_groups={{0,1,2,3}} | ||
| ROOT copy = s32[8]{0} copy(ra2a) | ||
| })"; | ||
|
|
||
| TF_ASSERT_OK_AND_ASSIGN(std::unique_ptr<HloModule> module, | ||
| ParseAndReturnUnverifiedModule((module_str))); | ||
| AllToAllDecomposer decomposer(true, 3); | ||
| TF_ASSERT_OK_AND_ASSIGN(bool changed, decomposer.Run(module.get())); | ||
| EXPECT_TRUE(changed); | ||
| EXPECT_THAT(module->entry_computation()->root_instruction(), | ||
| op::Copy(op::Reshape(op::RaggedAllToAll( | ||
| op::Reshape(op::Copy(op::Parameter(0))), | ||
| op::Reshape(op::Broadcast(op::Constant())), _, _, _, _)))); | ||
| std::vector<HloInstruction*> reshapes; | ||
| std::vector<HloInstruction*> ragged_all_to_alls; | ||
| for (HloInstruction* instruction : | ||
| module->entry_computation()->instructions()) { | ||
| if (instruction->opcode() == HloOpcode::kReshape) { | ||
| reshapes.push_back(instruction); | ||
| } | ||
| if (instruction->opcode() == HloOpcode::kRaggedAllToAll) { | ||
| ragged_all_to_alls.push_back(instruction); | ||
| } | ||
| } | ||
| EXPECT_EQ(reshapes.size(), 3); | ||
| EXPECT_EQ(ragged_all_to_alls.size(), 1); | ||
| EXPECT_EQ(ragged_all_to_alls[0]->shape().rank(), 3); | ||
| } | ||
|
|
||
| TEST_F(AllToAllDecomposerTest, RaggedAllToAllRank3) { | ||
| const std::string module_str = | ||
| R"(HloModule RaggedAllToAll | ||
| ENTRY AllToAll { | ||
| p0 = s32[8,16,256]{2,1,0} parameter(0) | ||
| c0 = s32[] constant(0) | ||
| output = s32[8,16,256]{2,1,0} broadcast(c0), dimensions={} | ||
| p1 = s32[4]{0} parameter(1) | ||
| p2 = s32[4]{0} parameter(2) | ||
| p3 = s32[4]{0} parameter(3) | ||
| p4 = s32[4]{0} parameter(4) | ||
| input = s32[8,16,256]{2,1,0} copy(p0) | ||
| input_offsets = s32[4]{0} copy(p1) | ||
| send_sizes = s32[4]{0} copy(p2) | ||
| output_offsets = s32[4]{0} copy(p3) | ||
| recv_sizes = s32[4]{0} copy(p4) | ||
| ra2a = s32[8,16,256]{2,1,0} ragged-all-to-all(input, output, input_offsets, send_sizes, output_offsets, recv_sizes), replica_groups={{0,1,2,3}} | ||
| ROOT copy = s32[8,16,256]{2,1,0} copy(ra2a) | ||
| })"; | ||
|
|
||
| TF_ASSERT_OK_AND_ASSIGN(std::unique_ptr<HloModule> module, | ||
| ParseAndReturnUnverifiedModule((module_str))); | ||
| AllToAllDecomposer decomposer(true, 3); | ||
| TF_ASSERT_OK_AND_ASSIGN(bool changed, decomposer.Run(module.get())); | ||
| EXPECT_FALSE(changed); | ||
| } | ||
|
|
||
| } // namespace | ||
| } // namespace xla |