b5038

fix MUSA compiler warning (#12704)

* fix MUSA compiler warning

* replace (void) with GGML_UNUSED

b5037

CANN: Support operator SIN COS ARGMAX (#12709)

* [CANN]support sin cos argmax

Signed-off-by: noemotiovon <[email protected]>

* [CANN]codestyle adjustment

Signed-off-by: noemotiovon <[email protected]>

* [CANN]Remove redundant code

Signed-off-by: noemotiovon <[email protected]>

---------

Signed-off-by: noemotiovon <[email protected]>
Co-authored-by: noemotiovon <[email protected]>

b5036

Simplify and improve CUDA graphs through use of indirect copy pointer…

b5035

CANN: Fix failed test cases (#12708)

* CANN: Fix memory waste in aclnn_tensor

* CANN: fix backend ops fail

* CANN: fix acl_tensor memory alloc.

* CANN: format

* CANN: remove trailing whitespace

b5034

opencl: use `max_alloc_size` in backend ctx instead of querying again…

b5033

vulkan: Implement split_k for coopmat2 flash attention. (#12627)

When using group query attention, we have one workgroup per KV batch and this
can be very few workgroups (e.g. just 8 in some models). Enable split_k to
spread the work across SMs. This helps a lot when the KV cache is large.

b5032

cmake: remove caching from vulkan coopmat checks (#12719)

b5031

vulkan: Implement grouped query attention in the coopmat2 FA shader (…

b5030

Vulkan: Fix mmq int dot float cache size (#12722)

b5029

model : print tensor size during load (#12711)

* model : print tensor size during load

* cont : fix units MB -> MiB

Co-authored-by: Diego Devesa <[email protected]>

---------

Co-authored-by: Diego Devesa <[email protected]>