FP8 QuantDot operation #2506
Merged
FP8 QuantDot operation #2506
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
umangyadav
commented
Dec 7, 2023
test/verify/main.cpp
Outdated
| "quant_dot_3args_5<migraphx::fp8::fp8e4m3fnuz, float>"}); | ||
| rv.disable_test_for("gpu", | ||
| {"test_conv_bn_add", | ||
| // These passes on MI300 but fails on others, same issue as CPU. |
There was a problem hiding this comment.
Will have to raise tolerance for FP8. I can't see other way around.
umangyadav
force-pushed
the
quant_gemm_fp8
branch
from
4da1510 to
d4a6dbd
Compare
TedThemistokleous
approved these changes
Dec 7, 2023
pfultz2
reviewed
Dec 8, 2023
src/targets/ref/lowering.cpp
Outdated
| return static_cast<int32_t>(x); | ||
| }); | ||
| }); | ||
| }); |
There was a problem hiding this comment.
We shouldn't be converting the input here. gemm is already using a double accumulator. We could also convert the amat and bmat to double to avoid loss with the multiply as well, but this can be done inline in the function.
pfultz2
reviewed
Dec 8, 2023
src/targets/ref/lowering.cpp
Outdated
| }); | ||
| }); | ||
| }); | ||
| migemm(result, arg_0, arg_1, int32_t{1}, int32_t{0}); |
There was a problem hiding this comment.
This should just call our gemm function instead of migemm.
ahsan-ca
approved these changes
Dec 11, 2023
pfultz2
approved these changes
Dec 12, 2023
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Outdated suggestions cannot be applied.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.
this version of Dot/GEMM is different in the sense that, it uses FP8 dtype for the input args but computes fp32 output.
need to disable two of the verify test for the CPU backend because they are failing. Failure is because of lossy cast from (Float->fp8->float) inside the "ref" implementation while CPU backend optimizes outfloat -> fp8 -> floatconverts tono-op. Therefore results are not matching for "Ref" and "CPU".Depends on #2473