Grouped conv_fwd_bias_bnorm_clamp instances and tests

[ApoorvaKalyani]

Proposed changes

Please describe the motivation behind the pull request, whether it enables a new feature or fixes a bug. If there are associated pull requests or issues, please link them to the pull request.

Checklist

Please put an x into the boxes that apply. You can also fill these out after creating the PR. If you're not sure, please don't hesitate to ask.

• I have added tests relevant to the introduced functionality, and the unit tests are passing locally
• I have added the test to REGRESSION_TESTS list defined at the top of CMakeLists.txt in tests/CMakeLists.txt, IF the test takes more than 30 seconds to run.
• I have added inline documentation which enables the maintainers with understanding the motivation
• I have removed the stale documentation which is no longer relevant after this pull request
• (If this change is user-facing) I have added release notes which provide the end users with a brief summary of the improvement from this pull request
• I have run clang-format on all changed files
• Any dependent changes have been merged

Discussion

If this is a relatively large or complex change, feel free to start a discussion by explaining why you chose the solution you did and what alternatives you considered

[krithalith]

I had a look at the test failures myself and found the following tolerances to be sufficient for BF16, F16,iInt initialization, and float initialization:

INTEGER INITIALIZATION:

BF16
double rtol = 1e-1;
double atol = 5e-3;

F16
double rtol = 1.5e-3;
double atol = 7e-3;

FLOAT INITIALIZATION:

BF16
double rtol = 1e-1;
double atol = 7e-3;

F16
double rtol = 2e-3;
double atol = 1e-2;

I think the bias bnorm clamp operation at the end just fundamentally magnifies small errors in the gemm, so the tolerances simply need to be higher for this op

Also a 1e-3 relative error for f16 (default value in check_err) is very low since that is pretty much exactly a single f16 epsilon. For BF16 this is suddenly a lot more lenient (1e-1 even though the epsilon is only 8 times as large). Also check_err() adds up the relative and absolute tolerance errors, which is dubious.