ggerganov/llama.cpp

Pull Requests Commits

ggml-cuda: hip: small nwarps GCN fix on Q3_K(ncols=2,3), Q6_K(ncols=4)

thelittlefireman committed 1 day ago

6e9b69aa

CUDA: faster MMVQ for low-bit types

am17an committed 3 days ago

c869d1e3

ggml : add note about mul_mat_id + disable mul_mat_id_dup tests (#30501)

ggerganov committed 3 days ago

Verified 8c2134a8

model: split layers evenly when no device reports free memory (#29677)

ServeurpersoCom committed 3 days ago

Verified df18bee1

vendor : update cpp-httplib to 0.61.0 (#30494)

angt committed 3 days ago

Verified 3eb10136

ggml : improve RPC tensor transfer logs (#30481)

ggerganov committed 3 days ago

Verified af0579c7

cuda : use SIMT implicit GEMM conv2d on AMD without WMMA (#30325)

aisk committed 3 days ago

Verified 62bacfa0

jinja : write floats in tojson as Python does (#30362)

ykhrustalev committed 3 days ago

Verified 77390baa

opencl: use buf+offset instead of subbuffer for adreno moe expert sort (#30418)

lhez committed 3 days ago

Verified c5a99762

opencl: fix q8_0 dp4a mul_mat_id and mark PREC_F32 mul_mat_id unsupported for adreno (#30365)

lhez committed 3 days ago

Verified 21a19cd3

models: drop the t_embd expand of the embedding graphs (#30461)

ServeurpersoCom committed 3 days ago

Verified 9cf967a0

HIP: use 64-row MMQ tiles for some quant types on RDNA3.5 (#28195)

liangliangchang committed 3 days ago

Verified 4dd6dd4b

vulkan: fix compile error (#30495)

0cc4m committed 3 days ago

Verified 14310190

llama : build the k-pool selection in the causal mask layout (#30480)

ggerganov committed 3 days ago

Verified e7268bf6

vulkan: Add workaround for device free memory > total memory on Intel Windows (#29835)

rillomas committed 3 days ago

Verified cdce056a

vulkan: Define the shader count for Intel Arc Pro B65 + B70 (#30431)

TheBlueMatt committed 3 days ago

Verified d524bb66

vulkan: support mulmat bf16 activations (#29069)

0cc4m committed 3 days ago

Verified 7ec420bc

unicode : add opt-in NFC normalization for Unigram tokenizers (#30390)

sandrohanea committed 3 days ago

Verified 10da8c46

convert : export YaRN truncation config (#30288)

tokinasin committed 3 days ago

Verified eca5458c

Enable OpenSSL in FP32/FP16 SYCL builds to restore HF HTTPS downloads (#30330)

raffaeleedidonna committed 3 days ago

Verified 091b6a7f

webgpu: (Fix CI) fix mul_mat_id with duplicate expert ids (#30451)

yomaytk committed 3 days ago

Verified fcac1907

server : escape the special tokens without std::regex (#30243)

ykhrustalev committed 3 days ago

Verified 04f7226a

chat: replace output string with parts (#30389)

ngxson committed 3 days ago

Verified ff026f8c

ggml : simplify backend search path error handling (#28367)

angt committed 3 days ago

Verified 1b70350c

docs : update zDNN supported ops (#30472)

taronaeo committed 3 days ago

Verified d71e324a

kleidiai: add SME2 Q4_K and Q6_K matmul support (#29219)

chaxu01 committed 3 days ago

Verified 45451c36

vulkan: Optimize pool_2d when the pooling window is large (#30119)

jeffbolznv committed 4 days ago

Verified 77c4540c

meta: recreate all subgraphs when the arena is reallocated (#30453)

Xanewok committed 4 days ago

Verified ecdd29e8

hexagon: add configurable HTP DCVS/HMX power mode (#30423)

pdhinaka committed 4 days ago

Verified ec0bfac1

graph: pad the KQ bias of the no-cache attention (#30446)

ServeurpersoCom committed 4 days ago

Verified ace20f67

Older