onnxruntime
Matmul_nbits kernel for mlas sqnbits to support Fp16 inputs
#21807
Merged
Merged
Matmul_nbits kernel for mlas sqnbits to support Fp16 inputs #21807
liqunfu
merged
24 commits into
main
from
liqun/mlas-sqnbit-kernel-fp16
matmul_nbits kernel for mlas sqnbits to support Fp16 inputs
edef19c4
fix fp16 for bias and zp
2e9e84fe
-mf16c
f85f72f6
unused args
4e8e284b
change tol, lint
27ba1bfd
lint and tol
07242a2a
Float16Cuda, lint, ARM64 compile
e4264586
lint
fc2c7b71
size_t
bb1f3d67
PREFast check
eb2439d1
dispatch fp32-16 conversion
a79d6eef
ConvertFp32ToFp16Avx
4e415499
x86 cpu failure due to dispatch == nullptr
67c6bbc0
default conversion
89d88e59
Merge branch 'main' into liqun/mlas-sqnbit-kernel-fp16
729dfd3d
yufenglee
commented
on 2024-09-10
yufenglee
commented
on 2024-09-10
New test skip Cuda EP
0bb7df03
not to template the kernel class
0b464077
Merge branch 'main' into liqun/mlas-sqnbit-kernel-fp16
e77933cc
undo emsdk
56fa3ee5
cast error
d32b4aee
refactor with existing halftofloat\
a5ce5dca
x86 build
f63c4741
kernel doc and dml
4b8a0f57
remove unused code
cd96fbbb
yufenglee
approved these changes
on 2024-09-13
fajin-corp
commented
on 2024-09-13
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub