Add BF16 in FP8 quantize ops #1961
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
✅ Deploy Preview for pytorch-fbgemm-docs canceled.
|
|
This pull request was exported from Phabricator. Differential Revision: D47904459 |
|
This pull request was exported from Phabricator. Differential Revision: D47904459 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Aug 21, 2023
Summary: Pull Request resolved: pytorch#1961 - Added output_dtype for half, bfloat16 and float as output in dequantization functions; currently it's an integer value defined by Sparse_dtype (float:0, half:1, bfloat16:5) - Added type conversion in quant and dequant kernels by using native CUDA/HIP functions for half to float conversion and writing everything explicitly. Differential Revision: D47904459 fbshipit-source-id: 41d3f0c50365d0482aab912c202f458a787419d8
sryap
force-pushed
the
export-D47904459
branch
from
8e44a0d
to
facb7ed
Compare
|
This pull request was exported from Phabricator. Differential Revision: D47904459 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Aug 23, 2023
Summary: Pull Request resolved: pytorch#1961 - Added output_dtype for half, bfloat16 and float as output in dequantization functions; currently it's an integer value defined by Sparse_dtype (float:0, half:1, bfloat16:5) - Added type conversion in quant and dequant kernels by using native CUDA/HIP functions for half to float conversion and writing everything explicitly. Reviewed By: jianyuh Differential Revision: D47904459 fbshipit-source-id: f608d7da5dcf05ff78a6e0eb13d985ed99207d1a
sryap
force-pushed
the
export-D47904459
branch
from
facb7ed
to
56e870d
Compare
|
This pull request was exported from Phabricator. Differential Revision: D47904459 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Aug 24, 2023
Summary: Pull Request resolved: pytorch#1961 - Added output_dtype for half, bfloat16 and float as output in dequantization functions; currently it's an integer value defined by Sparse_dtype (float:0, half:1, bfloat16:5) - Added type conversion in quant and dequant kernels by using native CUDA/HIP functions for half to float conversion and writing everything explicitly. Reviewed By: jianyuh Differential Revision: D47904459 fbshipit-source-id: 3fdca310b5262c249e7dc552070e27a569c9af23
sryap
force-pushed
the
export-D47904459
branch
from
56e870d
to
7d5b278
Compare
|
This pull request was exported from Phabricator. Differential Revision: D47904459 |
sryap
added a commit
to sryap/FBGEMM
that referenced
this pull request
Aug 24, 2023
Summary: Pull Request resolved: pytorch#1961 - Added output_dtype for half, bfloat16 and float as output in dequantization functions; currently it's an integer value defined by Sparse_dtype (float:0, half:1, bfloat16:5) - Added type conversion in quant and dequant kernels by using native CUDA/HIP functions for half to float conversion and writing everything explicitly. Reviewed By: jianyuh Differential Revision: D47904459 fbshipit-source-id: 4f5bbf71cb3c5f0ec2f4ef0048f30c6cdf48cd2e
sryap
force-pushed
the
export-D47904459
branch
from
7d5b278
to
a6ee85a
Compare
Summary: Pull Request resolved: pytorch#1961 - Added output_dtype for half, bfloat16 and float as output in dequantization functions; currently it's an integer value defined by Sparse_dtype (float:0, half:1, bfloat16:5) - Added type conversion in quant and dequant kernels by using native CUDA/HIP functions for half to float conversion and writing everything explicitly. Reviewed By: jianyuh Differential Revision: D47904459 fbshipit-source-id: d48da0fc7b0b158c46628952a7c7ec8e1aa502df
|
This pull request was exported from Phabricator. Differential Revision: D47904459 |
sryap
force-pushed
the
export-D47904459
branch
from
a6ee85a
to
19fb8e1
Compare
|
This pull request has been merged in 4920770. |
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.
Summary:
dequantization functions; currently it's an integer value defined by
Sparse_dtype (float:0, half:1, bfloat16:5)
CUDA/HIP functions for half to float conversion and writing
everything explicitly.
Differential Revision: D47904459