ggml-org/llama.cpp

Pull Requests Commits

Revert "Change max context length for auto-fitting with unified KV (#28849)"

gaugarg-nv committed 30 days ago

Verified 9461c7e4

mtmd: fix mel preprocessor in LFM2 audio (#29403)

ykhrustalev committed 30 days ago

Verified fcc89154

opencl: add bin kernel `kernel_gemm_noshuffle_q5_k_f32_32b_trans_ila_a8_bin`, `kernel_gemm_noshuffle_q5_k_q8_1_dp4a_ila_a8_bin` (#29401)

shaofeiqi committed 30 days ago

Verified a25c9865

Fixing the vulkan build issue of legacy GLSLC version that has no cooperativeMatrix API support (https://github.com/ggml-org/llama.cpp/issues/29373) (#29409)

sliu39 committed 30 days ago

Verified e85e15cf

gguf-py : ByteLevel processing defaults bos/eos to False (#29422)

CISC committed 30 days ago

Verified b248f4a3

gguf-py : TemplateProcessing has final word on add_special_token (#29417)

CISC committed 30 days ago

Verified d81aef19

common : extract shared unicode path/string helpers (#29415)

angt committed 30 days ago

Verified 27b20ba8

metal: FWHT kernels for block widths above 512 (#29095)

bri-prism committed 30 days ago

Verified e351231c

metal : split fa kernels into per-dtype libraries (#29329)

ggerganov committed 30 days ago

Verified 5a75f14c

llama : add `llama_prec_policy` + model-driven W4A4 path (#24364)

ynankani committed 30 days ago

Verified e9f824d8

HIP: bump HIP_VERSION requried for fp8 to avoid missing __hip_fp8_e4m3 support in 6.2 (#29231)

IMbackK committed 30 days ago

Verified d028c697

rpc: include nb in the get_alloc_size cache key and floor the result at ggml_nbytes (#29283)

Jesssullivan committed 30 days ago

Verified 66963a8b

[SYCL] support sparse FA (#28796)

arthw committed 30 days ago

Verified cd74ef62

musa: fix PH1 (MTT S5000) operator failures and build issues (#29193)

yeahdongcn committed 30 days ago

Verified f9af9be2

CUDA: fuse RMS_NORM + SCALE into one kernel (#29393)

InflexCZE committed 30 days ago

Verified 1ab7e5ad

llama : fix tensor split for fused qkv with uneven K/V head sizes (#29294)

am17an committed 31 days ago

Verified f805c57a

hexagon: add q5_k quant type support (#29123)

jhen0409 committed 31 days ago

Verified 4de09265

hexagon: use DMA for contiguous dim1 CONCAT (#29404)

kurquhar committed 31 days ago

Verified ed319feb

metal : fix graph capture and handle empty graphs (#29390)

ggerganov committed 31 days ago

Verified 84e76d8a

metal : optimize sparse FA + clean-up (#29377)

ggerganov committed 31 days ago

Verified cdc06426

sync : ggml (#29396)

ggerganov committed 31 days ago

Verified bced4595

hexagon: handle multi-sequence in concat_2d (#29344)

jhen0409 committed 31 days ago

Verified a02c7f58

llama-grammar: fix numeric truncation for token_id parsing (#29382)

apach301 committed 31 days ago

Verified 5cf3a352

hexagon: dynamic quantizer improvements (#29395)

kurquhar committed 31 days ago

Verified 07fc586e

hexagon: support I32 CPY and CONT (#29379)

kurquhar committed 31 days ago

Verified 97a418bd

cuda : add F16 kernel support for CONV_2D_DW (#29064)

crion99 committed 31 days ago

Verified a72e04ab

test: flush status (#28352)

kurquhar committed 31 days ago

Verified 8212c780

ui : fix missing svg use and animation elements in preview and download (#28962)

nandan2003 committed 31 days ago

Verified 945064fc

llama: add llama_batch_ext (#24669)

ngxson committed 31 days ago

Verified fc343a84

server : change default pytest workers to 4 (#29376)

danbev committed 31 days ago

Verified 308883b3

Older