[GPU/OpenCL] Initial version of Addition Layer with OpenCL ops #2606
Conversation
yashSingh0723
requested review from
myungjoo,
jijoongmoon,
again4you,
jaeyun-jung,
leemgs,
wooksong,
helloahn,
kparichay,
gichan-jang,
anyj0527,
zhoonit,
lhs8928,
songgot,
jihochu,
DonghakPark,
SeoHyungjun,
baek2sm,
skykongkong8,
djeong20,
EunjuYang and
a team
as code owners
|
📝 TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2606. Please a submit 1commit/1PR (one commit per one PR) policy to get comments quickly from reviewers. Your PR must pass all verificiation processes of cibot before starting a review process from reviewers. If you are new member to join this project, please read manuals in documentation folder and wiki page. In order to monitor a progress status of your PR in more detail, visit http://ci.nnstreamer.ai/. |
There was a problem hiding this comment.
@yashSingh0723, 💯 All CI checkers are successfully verified. Thanks.
| addition_cl(data, rdata, size, context); | ||
|
|
||
| } else | ||
| throw std::invalid_argument("Error: OpenCL fp16 is not supported yet."); |
There was a problem hiding this comment.
I think it is because we are not implementing FP16 yet. isn't it? If it is, then I think supporting FP16 is more important because the LLM is working based on FP16.
There was a problem hiding this comment.
Added FP16 support for Addition Layer in the latest commit.
| input_step_dim.batch(1); | ||
| input_step_dim.height(to - from); | ||
|
|
||
| Tensor input_step = input_.getSharedDataTensor( |
There was a problem hiding this comment.
I wonder if we do like this way, we need to data transfer from CPU to GPU. Doesn't it affect the latency?
There was a problem hiding this comment.
@jijoongmoon This optimisation will be a part of future PRs.
yashSingh0723
force-pushed
the
opencl_layer_addition
branch
from
5de81ff
to
a8f089f
Compare
There was a problem hiding this comment.
@yashSingh0723, 💯 All CI checkers are successfully verified. Thanks.
There was a problem hiding this comment.
@yashSingh0723, 💯 All CI checkers are successfully verified. Thanks.
yashSingh0723
force-pushed
the
opencl_layer_addition
branch
from
9976775
to
a98c244
Compare
There was a problem hiding this comment.
@yashSingh0723, 💯 All CI checkers are successfully verified. Thanks.
There was a problem hiding this comment.
@yashSingh0723, 💯 All CI checkers are successfully verified. Thanks.
yashSingh0723
force-pushed
the
opencl_layer_addition
branch
from
cf43132
to
39bd0d9
Compare
There was a problem hiding this comment.
@yashSingh0723, 💯 All CI checkers are successfully verified. Thanks.
There was a problem hiding this comment.
@yashSingh0723 LGTM. However, please address Jijoong's concern before merging.
Added naive version of OpenCL implementation for Addition Layer. Incorporated kernel for ops used. Added unit test for addition_layer_cl. Signed-off-by: yash.singh <[email protected]>
Added addition kernel to enhance reusability of the common blas kernels. Used AdditionLayer interface for both CPU and GPU calls. Signed-off-by: yash.singh <[email protected]> [GPU/OpenCL] Initial version of Addition Layer with OpenCL ops Added naive version of OpenCL implementation for Addition Layer. Incorporated kernel for ops used. Added unit test for addition_layer_cl. Signed-off-by: yash.singh <[email protected]> [GPU/OpenCL] Addition Kernel added in reusable blas OpenCL kernels Added addition kernel to enhance reusability of the common blas kernels. Used AdditionLayer interface for both CPU and GPU calls. Signed-off-by: yash.singh <[email protected]>
yashSingh0723
force-pushed
the
opencl_layer_addition
branch
from
39bd0d9
to
3a17d5f
Compare
|
|
cibot: Added fp16 support for Addition layer Added unit tests for fp16 support Updated the Layer Semantics for GPU Signed-off-by: yash.singh <[email protected]>
yashSingh0723
force-pushed
the
opencl_layer_addition
branch
from
3a17d5f
to
5ad3e53
Compare
There was a problem hiding this comment.
@yashSingh0723, 💯 All CI checkers are successfully verified. Thanks.
Added initial version of Addition Layer for GPU. This is a basic implementation using naive kernel.
Changes added with this PR:
addition_layer_cl.cppadded containing the newAdditionLayerCLclass for OpenCL implementation.LayerKernelenum insidelayer_context.h.unittest_layers_addition_cl.cppto test Addition Layer on GPU.Signed-off-by: Yash Singh [email protected]