huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-16 10:38:54 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add the recip op + use it in stable-diffusion. (#331)

Browse Source 
* Add the recip unary op.

* Fix the cuda kernel.

* Use the recip op in sigmoid.
This commit is contained in:
Laurent Mazare
2023-08-06 22:14:52 +02:00
committed by GitHub
parent 1c062bf06b
commit 166bfd5847
5 changed files with 26 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
candle-kernels/src/unary.cu
View File

@ -80,6 +80,7 @@ extern "C" __global__ void FN_NAME( \
#if __CUDA_ARCH__ >= 800
UNARY_OP(__nv_bfloat16, ucopy_bf16, x)
UNARY_OP(__nv_bfloat16, uneg_bf16, -x)
UNARY_OP(__nv_bfloat16, urecip_bf16, recipg(x))
UNARY_OP(__nv_bfloat16, uexp_bf16, expg(x))
UNARY_OP(__nv_bfloat16, ulog_bf16, logg(x))
UNARY_OP(__nv_bfloat16, usin_bf16, sing(x))
@ -95,6 +96,7 @@ UNARY_OP1(__nv_bfloat16, uelu_bf16, elu_fwd(x, param))
#if __CUDA_ARCH__ >= 530
UNARY_OP(__half, ucopy_f16, x)
UNARY_OP(__half, uneg_f16, -x)
UNARY_OP(__half, urecip_f16, recipg(x))
UNARY_OP(__half, uexp_f16, expg(x))
UNARY_OP(__half, ulog_f16, logg(x))
UNARY_OP(__half, usin_f16, sing(x))
@ -113,6 +115,8 @@ UNARY_OP(float, ucopy_f32, x)
UNARY_OP(double, ucopy_f64, x)
UNARY_OP(float, uneg_f32, -x)
UNARY_OP(double, uneg_f64, -x)
UNARY_OP(float, urecip_f32, recipg(x))
UNARY_OP(double, urecip_f64, recipg(x))
UNARY_OP(float, uexp_f32, expg(x))
UNARY_OP(double, uexp_f64, expg(x))
UNARY_OP(float, ulog_f32, logg(x))