CANN: implement the SSM_CONV operator (#17737)

• CANN: implement SSM_CONV operator

Co-authored-by: Aleksei Lobanov, zeromarblectm@gmail.com
Co-authored-by: Sujin Kang, waterjin326@gmail.com

• CANN: remove custom error limit for SSM_CONV

• CANN: merge SSM_CONV tensor shape/strides into one line

Co-authored-by: Sujin Kang, waterjin326@gmail.com

ggml-cuda: fix regex for arch list (#18371)

• ggml-cuda: fix regex for arch list

• make regex exact

cuda: optimize cumsum cub path (#18362)

• cuda: optimize cumsum cub path

• remove heavy perf test

ggml-cuda: fix blackwell native builds (#18361)

• ggml-cuda: fix blackwell native builds

Replace 12x in native architectures by 12xa

• replace for GGML_NATIVE=OFF too

• only replace for native

• remove 120f-virtual for default compilation

Co-authored-by: Aman Gupta

model : support for LlamaBidirectionalModel architecture (#18220)

• model: llama-embed-nemotron

• minor: python lint

• changed arch-name

• templated llm_build_llama to be used for both llama and llama-embed arch

vulkan: fix command buffer corruption in ggml_backend_vk_event_wait (#18302)

CANN : refactor ACL graph cache (#17752)

Move the graph property checking code into methods of LRU cache.

Signed-off-by: Wang Weixuan wangweixvan@gmail.com

vulkan: use fewer FA rows for small cache runs (#18280)

CANN: Uses yarn_ramp cache in ROPE (#17725)

common: add LLAMA_ARG_OVERRIDE_TENSOR env var for -ot arg (#18267)