ggml-org/llama.cpp

Pull Requests Commits

kv-cache : fix k-shift for multiple streams

ggerganov committed 58 days ago

Verified 05baa62a

llama : reuse compute graphs (#14482)

ggerganov committed 58 days ago

Verified 01612b74

llama : fix parallel processing for lfm2 (#14705)

tdakhran committed 58 days ago

Verified 086cf81e

kv-cache : opt mask set input (#14600)

ggerganov committed 58 days ago

Verified d9b69108

batch : fix uninitialized has_cpl flag (#14733)

ggerganov committed 58 days ago

Verified ad57d3ed

ci : disable failing vulkan crossbuilds (#14723)

CISC committed 59 days ago

Verified 1ba45d49

convert : make hf token optional (#14717)

CISC committed 59 days ago

Verified 19e5943d

llama : fix parameter order for hybrid memory initialization (#14725)

dinerburger committed 59 days ago

Verified 496957e1

ggml: Add initial WebGPU backend (#14521)

reeselevine committed 59 days ago

Verified 21c02174

model : support output bias for qwen2 (#14711)

tempstudio committed 59 days ago

Verified b0f0ecc3

llama : add high-throughput mode (#14363)

ggerganov committed 59 days ago

Verified 225e7a14

Support diffusion models: Add Dream 7B (#14644)

am17an committed 59 days ago

Verified ab140198

ggml : add asserts (#14720)

ggerganov committed 59 days ago

Verified 64978340

server : pre-calculate EOG logit biases (#14721)

ggerganov committed 59 days ago

Verified 6ffd4e9c

llama : fix parallel processing for plamo2 (#14716)

mitmul committed 59 days ago

Verified e4841d24

server : fix handling of the ignore_eos flag (#14710)

ggerganov committed 59 days ago

Verified 538cc77f

scripts: synthetic prompt mode for server-bench.py (#14695)

JohannesGaessler committed 59 days ago

Verified 5cae7665

convert : only check for tokenizer folder if we need it (#14704)

CISC committed 59 days ago

Verified 4b91d6f7

convert : add pre-computed hashes first to prevent order mishaps (#14701)

CISC committed 59 days ago

Verified cf91f217

llama: add LLAMA_API to deprecated llama_kv_self_seq_div (#14708)

Min-Hua committed 60 days ago

Verified 79e0b68c

gguf-py : dump bpw per layer and model in markdown mode (#14703)

EAddario committed 60 days ago

Verified c81f4192

model : add Kimi-K2 support (#14654)

gabriellarson committed 60 days ago

Verified 4a4f4269

vulkan: fix noncontig check for mat_mul_id splitting (#14683)

jeffbolznv committed 60 days ago

Verified ba1ceb34

vulkan: add RTE variants for glu/add/sub/mul/div (#14653)

jeffbolznv committed 60 days ago

Verified 10a0351a

model : add PLaMo-2 support (#14560)

mitmul committed 60 days ago

Verified 68e37a61

cuda: fix build warnings in set-rows.cu (unused variable) (#14687)

yeahdongcn committed 60 days ago

Verified cbc68be5

sycl: Hotfix for non dnnl codepath (#14677)

ShanoToni committed 61 days ago

Verified bdca3837

ggml : refactor llamafile_sgemm PPC code (#14673)

shalinib-ibm committed 61 days ago

Verified 55c509da

llama-context: add ability to get logits (#14672)

am17an committed 61 days ago

Verified 9c9e4fc6

scripts: benchmark for HTTP server throughput (#14668)

JohannesGaessler committed 61 days ago

Verified 494c5899

Older