Skip to content

[FRONTEND] fix matmul int8 overflow issue #2297

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

Merged
ptillet merged 1 commit into from
Sep 17, 2023
+4 −0
Merged

[FRONTEND] fix matmul int8 overflow issue #2297

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions python/triton/ops/matmul.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -155,6 +155,8 @@ def _call(a, b, dot_out_dtype, allow_tf32):
if a.dtype in [tl.float8e4nv, tl.float8e4b15, tl.float8e5] or\
b.dtype in [tl.float8e4nv, tl.float8e4b15, tl.float8e5]:
c_dtype = torch.float16
elif a.dtype in [torch.int8] or b.dtype in [torch.int8]:
c_dtype = torch.int32
else:
c_dtype = get_higher_dtype(a.dtype, b.dtype)
c = torch.empty((M, N), device=device, dtype=c_dtype)
Expand All @@ -174,6 +176,8 @@ def _call(a, b, dot_out_dtype, allow_tf32):
ab_dtype = True
if a.dtype in [tl.float8e4nv, tl.float8e5] and b.dtype in [tl.float8e4nv, tl.float8e5]:
ab_dtype = False
if a.dtype in [torch.int8] and b.dtype in [torch.int8]:
ab_dtype = False
# launch kernel
grid = lambda META: (cdiv(M, META['BLOCK_M']) * cdiv(N, META['BLOCK_N']), META['SPLIT_K'])
_kernel[grid](a, b, c, M, N, K,
Expand Down