ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-14 02:02:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

NHWC conv 2d: fwd bfp16/int8, Device level tuning and host API (#73)

Browse Source 
* add fwd bf16 conv

* change tunning parametor

* add int8 for conv fwd

* remove comments

* change tunning parametor for int8

* change init int8 example

* add test for conv2d fwd

* change device operation file pos because merge develop

* fwd int8 use reference

* test_conv_fwd use reference

* add braket for if statement

* rename fwd example name

* remove StaticBufferOfVectorTypeV2

* tweak example

Co-authored-by: ltqin <letaoqin@amd.com>
Co-authored-by: Chao Liu <chao.liu2@amd.com>

[ROCm/composable_kernel commit: 880fbee957]
This commit is contained in:
ltqin
2022-02-12 10:06:40 +08:00
committed by GitHub
parent a76cc6deed
commit 32c128bcc5
16 changed files with 960 additions and 130 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
reference_operation/include/reference_conv_fwd.hpp
View File

@@ -86,10 +86,9 @@ struct ReferenceConvFwd : public device::BaseOperator
float v_wei;
arg.in_element_op_(
v_in,
static_cast<const float>(arg.in_n_c_hi_wi_(n, c, hi, wi)));
v_in, ck::type_convert<float>(arg.in_n_c_hi_wi_(n, c, hi, wi)));
arg.wei_element_op_(
v_wei, static_cast<const float>(arg.wei_k_c_y_x_(k, c, y, x)));
v_wei, ck::type_convert<float>(arg.wei_k_c_y_x_(k, c, y, x)));
v_acc += v_in * v_wei;
}
@@ -101,7 +100,7 @@ struct ReferenceConvFwd : public device::BaseOperator
arg.out_element_op_(v_out, v_acc);
arg.out_n_k_ho_wo_(n, k, ho, wo) = v_out;
arg.out_n_k_ho_wo_(n, k, ho, wo) = ck::type_convert<OutDataType>(v_out);
};
make_ParallelTensorFunctor(f_nchw,