b5272

imatrix: fix oob writes if src1 is not contiguous (#13286)

b5271

clip : revert the change of BOI/EOI token for GLM-edge (⚠️ breaking c…

b5270

llama : Llama-3_1-Nemotron-Ultra-253B-v1 support (#12843)

b5269

llama : move end-user examples to tools directory (#13249)

* llama : move end-user examples to tools directory

---------

Co-authored-by: Xuan Son Nguyen <[email protected]>

b5267

context : fix reorder logic (#13267)

ggml-ci

b5266

ggml : Enable MMA for BF16 in llamafile_sgemm (#13148)

This patch upstreams llamafile's cpu matrix multiplication kernels for ppc64le using MMA builtins for BF16 data type.

This change results in 9x - 40x gains
in total speed S t/s (ie all tokens/total time), across various batch sizes tested using llama-batched-bench benchmark.

The patch is tested with Meta-Lllama-3-8B,
and Mistral-7B models (BF16 models generated by using llama-quantize from corresponding FP32 models) on an IBM POWER10 machine.

Signed-off-by: Shalini Salomi Bodapati <[email protected]>

b5265

llama-model : support Qwen2 embedding models and pooling_mode_lasttok…

b5261

llama : orion rope type is neox (#13261)

b5260

llama : plamo rope type is neox (#13260)

b5259

llama-chat : reset glmedge chat template (#13253)

* reset glmedge chat template

* fix glmedge chat template