ggerganov/llama.cpp

Pull Requests Commits

tests : consolidate the context state tests into test-llama-context

ggerganov committed 26 days ago

65b4dc5b

llama : report compute graph reallocations during inference (#30329)

ggerganov committed 26 days ago

Verified 972b5fcc

ui: fix text attachment preview overflow (#29898)

shamnadxyz committed 26 days ago

Verified 2f160ba5

sycl: guard DMMV reads at row tails (#29338)

newjordan committed 26 days ago

Verified f9a95aeb

vulkan: use ternary expressions to avoid switch lowering for CM2 quant block size (#30164)

SG-Amadeus committed 26 days ago

Verified 08ffba21

ggml: Change default visibility of symbols to hidden (#30014)

ckastner committed 26 days ago

Verified a05e58c9

vulkan: fix buffer logging on 32-bit platforms after source split (#30326)

miyanyan committed 26 days ago

Verified b2f036c0

hexagon: optimize q2_k bpw and activation (#30035)

jhen0409 committed 26 days ago

Verified e937b17a

hexagon: add IM2COL_3D (#30305)

kurquhar committed 26 days ago

Verified b13a93a3

opencl: improve fa, allow dk512 for gemma-4, improve dk64 (#30266)

lhez committed 27 days ago

Verified f2cc7282

opencl: fix image limit for q4_k dense bin kernels, q4_0 bcast and rms_norm (#30310)

lhez committed 27 days ago

Verified c83f3058

rpc : turn GGML_RPC_DEBUG into a verbosity level and add logs (#29544)

ggerganov committed 27 days ago

Verified 5c8854fd

model : support for Prism Bonsai 2 27B (#29600)

bri-prism committed 27 days ago

Verified a9d27ac6

cpu: fix AVX2 CPUs not using F16C when available (#30270)

xinitrcn1 committed 27 days ago

Verified 1cf77677

vulkan: handle mul_mat_id duplicates in the prepass rather than looping (#29998)

jeffbolznv committed 27 days ago

Verified 966baae7

server: add component tests (#30312)

ngxson committed 27 days ago

Verified 62a6f74b

server: leave a busy slot untouched when a request pins it (#30295)

ServeurpersoCom committed 28 days ago

Verified 23b0202a

model : support MiniCPM-V 4.7 (#29416)

tc-mb committed 28 days ago

Verified 69f201a2

cmake(s390x): disable z17 target for unsupported compilers (#30297)

taronaeo committed 28 days ago

Verified abee0c84

args: add LLAMA_ARG_SLOT_SAVE_PATH env for --slot-save-path (#30272)

kucharskim committed 28 days ago

Verified aa94f208

spec: properly handle mtmd input for mtp (#30257)

ngxson committed 28 days ago

Verified 781dbc5a

mtmd: add build_inp_attn_mask (#30259)

ngxson committed 28 days ago

Verified 0fd868cb

cuda: always use MMVQ for MUL_MAT_ID on sm_60 (#27828)

dfriehs committed 28 days ago

Verified 1623d8ce

CUDA/HIP: fix race in flash_attn_ext_f16_process_tile when nbatch_combine != DKQ/2 (#30103)

IMbackK committed 28 days ago

Verified 1bb2b9fc

ggml-cpu: vectorize fp32 to fp16 conversion (#30157)

taronaeo committed 28 days ago

Verified 2bbca8f2

vulkan: use 4 rows for NVIDIA MUL_MAT_ID MMVQ (#29274)

SG-Amadeus committed 28 days ago

Verified 404f557b

ggml: fix s390x all cpu build (#30140)

taronaeo committed 28 days ago

Verified b797c82c

opencl: add bin kernels kernel_gemm_moe_q4_k_q8_1_dp4a_bin, kernel_gemm_moe_q6_k_q8_1_dp4a_bin (#30187)

shawngu-quic committed 28 days ago

Verified f2918cab

sycl : accelerate MXFP4 MoE with arithmetic decoding and weight reordering (#29809)

Captain-Tripps committed 28 days ago

Verified 1e6f04a7

vendor: apply deep nested json patch from upstream (#30253)

ngxson committed 28 days ago

Verified 10a60cf3

Older