Delete FP8 Scaling Factors in GEMM Rewriter #16841
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
reedwm
requested changes
Sep 26, 2024
Comment on lines
1089
to
1090
| one = instr->AddInstruction( | ||
| HloInstruction::CreateConstant(LiteralUtil::One(F32))); |
There was a problem hiding this comment.
Why not unconditionally create this, as was done before? Even before, it's possible this was unused. I worry this may be accidentally used without being initialized in the future.
philipphack
force-pushed
the
u_fp8_scales_xla
branch
from
2ef615e
to
fd9750f
Compare
reedwm
approved these changes
Sep 27, 2024
copybara-service bot
pushed a commit
that referenced
this pull request
Sep 27, 2024
Imported from GitHub PR #16841 Removes the scaling factors of C and D (matrix bias and result) from FP8 Custom Calls created in the GEMM rewriter when their data types are not FP8. See #15795. Copybara import of the project: -- fd9750f by Philipp Hack <[email protected]>: Removes superfluous FP8 scaling factors in GEMM rewriter. Merging this change closes #16841 FUTURE_COPYBARA_INTEGRATE_REVIEW=#16841 from philipphack:u_fp8_scales_xla fd9750f PiperOrigin-RevId: 679766659
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Sep 27, 2024
Imported from GitHub PR openxla/xla#16841 Removes the scaling factors of C and D (matrix bias and result) from FP8 Custom Calls created in the GEMM rewriter when their data types are not FP8. See openxla/xla#15795. Copybara import of the project: -- fd9750fa8474fe72fe641c7b3bc005ff30396e0a by Philipp Hack <[email protected]>: Removes superfluous FP8 scaling factors in GEMM rewriter. Merging this change closes #16841 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#16841 from philipphack:u_fp8_scales_xla fd9750fa8474fe72fe641c7b3bc005ff30396e0a PiperOrigin-RevId: 679766659
copybara-service bot
pushed a commit
that referenced
this pull request
Sep 28, 2024
Imported from GitHub PR #16841 Removes the scaling factors of C and D (matrix bias and result) from FP8 Custom Calls created in the GEMM rewriter when their data types are not FP8. See #15795. Copybara import of the project: -- fd9750f by Philipp Hack <[email protected]>: Removes superfluous FP8 scaling factors in GEMM rewriter. Merging this change closes #16841 FUTURE_COPYBARA_INTEGRATE_REVIEW=#16841 from philipphack:u_fp8_scales_xla fd9750f PiperOrigin-RevId: 679766659
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Sep 28, 2024
Imported from GitHub PR openxla/xla#16841 Removes the scaling factors of C and D (matrix bias and result) from FP8 Custom Calls created in the GEMM rewriter when their data types are not FP8. See openxla/xla#15795. Copybara import of the project: -- fd9750fa8474fe72fe641c7b3bc005ff30396e0a by Philipp Hack <[email protected]>: Removes superfluous FP8 scaling factors in GEMM rewriter. Merging this change closes #16841 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#16841 from philipphack:u_fp8_scales_xla fd9750fa8474fe72fe641c7b3bc005ff30396e0a PiperOrigin-RevId: 679766659
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Sep 28, 2024
Imported from GitHub PR openxla/xla#16841 Removes the scaling factors of C and D (matrix bias and result) from FP8 Custom Calls created in the GEMM rewriter when their data types are not FP8. See openxla/xla#15795. Copybara import of the project: -- fd9750fa8474fe72fe641c7b3bc005ff30396e0a by Philipp Hack <[email protected]>: Removes superfluous FP8 scaling factors in GEMM rewriter. Merging this change closes #16841 PiperOrigin-RevId: 679784586
hsharsha
added a commit
to ROCm/xla
that referenced
this pull request
Oct 8, 2024
openxla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test
hsharsha
added a commit
to ROCm/xla
that referenced
this pull request
Oct 8, 2024
openxla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 14, 2024
Imported from GitHub PR #18062 #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b PiperOrigin-RevId: 685632239
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Oct 14, 2024
Imported from GitHub PR openxla/xla#18062 openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b8de0785d43e18dbdb0773307870084e32 by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b8de0785d43e18dbdb0773307870084e32 PiperOrigin-RevId: 685632239
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 14, 2024
Imported from GitHub PR #18062 #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b PiperOrigin-RevId: 685632239
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Oct 14, 2024
Imported from GitHub PR openxla/xla#18062 openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b8de0785d43e18dbdb0773307870084e32 by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b8de0785d43e18dbdb0773307870084e32 PiperOrigin-RevId: 685632239
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 14, 2024
Imported from GitHub PR #18062 #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b PiperOrigin-RevId: 685632239
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 14, 2024
Imported from GitHub PR #18062 #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b PiperOrigin-RevId: 685632239
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Oct 14, 2024
Imported from GitHub PR openxla/xla#18062 openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b8de0785d43e18dbdb0773307870084e32 by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b8de0785d43e18dbdb0773307870084e32 PiperOrigin-RevId: 685632239
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 21, 2024
Imported from GitHub PR #18062 #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b PiperOrigin-RevId: 688022433
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Oct 21, 2024
Imported from GitHub PR openxla/xla#18062 openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b8de0785d43e18dbdb0773307870084e32 by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b8de0785d43e18dbdb0773307870084e32 PiperOrigin-RevId: 688022433
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 21, 2024
Imported from GitHub PR #18062 #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b PiperOrigin-RevId: 688022433
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Oct 21, 2024
Imported from GitHub PR openxla/xla#18062 openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b8de0785d43e18dbdb0773307870084e32 by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b8de0785d43e18dbdb0773307870084e32 PiperOrigin-RevId: 688022433
copybara-service bot
pushed a commit
that referenced
this pull request
Oct 21, 2024
Imported from GitHub PR #18062 #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch #16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 COPYBARA_INTEGRATE_REVIEW=#18062 from ROCm:ci_fix_gemm_rewriter_fp8_tests_20241008 be4da5b PiperOrigin-RevId: 688034088
copybara-service bot
pushed a commit
to tensorflow/tensorflow
that referenced
this pull request
Oct 21, 2024
Imported from GitHub PR openxla/xla#18062 openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Copybara import of the project: -- be4da5b8de0785d43e18dbdb0773307870084e32 by Harsha HS <[email protected]>: [ROCm] Fix gemm_rewriter_test for AMD GCN Arch openxla/xla#16841 removes scaling factor constants in gemm_rewriter for FP8 data types. This patch address the same in the gemm_rewriter_test Merging this change closes #18062 PiperOrigin-RevId: 688034088
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.
Removes the scaling factors of C and D (matrix bias and result) from FP8 Custom Calls created in the GEMM rewriter when their data types are not FP8. See #15795.