gpu-computing

It would be great to have FBeta, F2, or F0.5 metrics to be implemented without the need for a custom metric class defined by user.

catboost version: 0.26

Heston model has accurate density approximations for European option prices, which are of interest.

The module implementing this method should live under tf_quant_finance/volatility/heston_approximation.py. It should support both European option puts and calls approximations. Tests should be in heston_approximation_test.py in the same folder.

The standard accelerate test suite, used by all the backends, can be quite slow. Several of the tests are significantly slower than the others, for example segmented folds and scans, which I believe is because the reference implementations are very inefficient. Writing some more efficient reference implementations (e.g. using Data.Vector.Unboxed) should help speed things up.

Hi,

one could and should experiment with Interprocedural optimization (IPO) also known as link-time optimization (LTO), especially on the host side for smaller binaries and potentially faster code. It's supported by GCC, Clang, and ICC, among others, which are our typical go-to compilers in HPC.

It's very easy to implement as well

Open issue to openly discuss potential ideas or improvements, whether on documentation, interfaces, examples, bug fixes, etc.

Just an FYI whilst I was trawling through the ROCm GitHub page:

https://rocmdocs.amd.com/en/latest/Programming_Guides/Programming-Guides.html#

The problem is that the OpenCL types in https://github.com/triSYCL/triSYCL/blob/master/include/triSYCL/opencl_types.hpp are defined on the host according to the x86-64 Linux ABI which depends on the CPU & OS instead of using the description from https://www.khronos.org/registry/OpenCL/specs/2.2/html/OpenCL_C.html#built-in-scalar-data-types

Note that the system-wide cl_size_t has been removed

See https://forums.luxcorerender.org/viewtopic.php?f=5&p=31330#p31330
Should only require very small changes to nodes/textures/imagemap.py