Skip to content
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.

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

[GPU/OpenCL] Fused DotCL, Addition and RMS for optimization #2831

Draft
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

yashSingh0723
Copy link
Contributor

  1. Added the Fused Layer for all above operation. fused_fc_norm_cl.cpp
  2. Fused Process where existing kernels are getting used, Please refer the below files: blas_kernel_interface.cpp and blas_kernels.cpp.
  3. Wrote Related Unittest for the above
  4. Need to check for layer unittest as they are failing. Some issues with generating golden data, Output is correct but st some point not able to match it with generated golden data. Check file test/unittest/layers/unittest_layers_fused_fc_rms_cl.cpp and test/input_gen/gen_layer_tests.py. For Reference and Debugging check test/unittest/layers/layers_golden_tests.cpp.
  5. Added rms_norm_cl_kernel in blas_kernels directory for generalization.

TO-DO

  1. Add for FP16
  2. Add for dotcl for case M==1 and N==1.

Signed-off-by: Yash Singh [email protected]

Added the Fused Layer for all above operation
Wrote Related Unittest for the above

Signed-off-by: Yash Singh <[email protected]>
auto disable_bias = std::get<props::DisableBias>(*layer_impl_props);
bool disable_bias_value = disable_bias.empty() || disable_bias.get() == false;
const Tensor &bias = context.getWeight(weight_idx[FC_RMSParams::bias]);
// printf("\n*************************************************************************************************************************************\n");
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

do we need these comments?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants