ggml-org/llama.cpp

Pull Requests Commits

vulkan: Remove splitting for mul_mat_id (#15568)

jeffbolznv committed 1 year ago

Verified 34bdbbd7

CUDA: Accelerate MXFP4 table lookup using `__byte_perm` (#15451)

Qeeweew committed 1 year ago

Verified 74f52f77

opencl: fix support ops condition for `rms_norm` (#15560)

lhez committed 1 year ago

Verified f7207b04

vulkan: fix min subgroup 16 condition for mmid subgroup optimization (#15565)

0cc4m committed 1 year ago

Verified 4d917cd4

tests: Generate unique input values for count_equal (#15487)

jeffbolznv committed 1 year ago

Verified 886b97a5

metal: fix regression when no metal devices are present (#15531)

booxter committed 1 year ago

Verified 111f8d06

CUDA: MoE helper in device code, better tile sizes (#15525)

JohannesGaessler committed 1 year ago

Verified 5eff6ec9

model-conversion : set pooling type to none in logits.cpp (#15564)

danbev committed 1 year ago

Verified dfd9b5f6

model-conversion : add model card template for embeddings [no ci] (#15557)

danbev committed 1 year ago

Verified 5a6bc6b1

batched-bench : fix unified KV cache handling + pp timing (#15562)

ggerganov committed 1 year ago

Verified 6b64f74b

convert : update Ernie 4.5 dense architecture name (#15555)

ownia committed 1 year ago

Verified 0d5a4702

metal : add FA kernels for HS=40 (#15559)

ggerganov committed 1 year ago

Verified b0ba31f5

convert : support interns1-mini (#15412)

RunningLeon committed 1 year ago

Verified 7da9fed0

CANN: ROPE cache sin/cos repeat (#15501)

noemotiovon committed 1 year ago

Verified c247d06f

vulkan: apply MUL_MAT_ID subgroup optimization to non-coopmat devices (#15524)

0cc4m committed 1 year ago

Verified 043fb27d

kv-cache : support layer reuse (#15504)

ggerganov committed 1 year ago

Verified b730706a

vulkan: Support FA with any multiple of 8 head sizes (#15537)

jeffbolznv committed 1 year ago

Verified c9a24fb9

vulkan: enable Conv2D for Apple after MoltenVK fixed the bug (#15526)

0cc4m committed 1 year ago

Verified a9c6ffcb

vulkan: workaround MoltenVK compile failure in multi_add (#15506)

jeffbolznv committed 1 year ago

Verified e78cf0d4

CUDA: fix half2 -> half conversion for HIP (#15529)

JohannesGaessler committed 1 year ago

Verified 710dfc46

vulkan: optimize rms_norm, and allow the work to spread across multiple SMs (#15281)

jeffbolznv committed 1 year ago

Verified 611f419c

model : add support for Seed-OSS (#15490)

pwilkin committed 1 year ago

Verified b1afcab8

scripts: fix compare-llama-bench.py (#15521)

JohannesGaessler committed 1 year ago

Verified 9ef53690

chat : fix debug build assertion in trim function (#15520)

LaffeyNyaa committed 1 year ago

Verified 21dc4dda

vulkan: Rewrite synchronization to allow some overlap between nodes (#15489)

jeffbolznv committed 1 year ago

Verified 289bf411

vulkan.Dockerfile: install vulkan SDK using tarball (#15282)

yeahdongcn committed 1 year ago

Verified b55f06e1

vulkan : support ggml_mean (#15393)

Acly committed 1 year ago

Verified 0a9b43e5

vulkan: optimize mul_mat_id loading row ids into shared memory (#15427)

jeffbolznv committed 1 year ago

Verified 330c3d2d

test-opt: allow slight inprecision (#15503)

JohannesGaessler committed 1 year ago

Verified e92734d5

ggml WebGPU: add support for quantization types (#15440)

reeselevine committed 1 year ago

Verified 45363632

Newer Older