ggml-org/llama.cpp

Pull Requests Commits

vocab : validate plamo2 byte tokens (#26511)

CISC committed 6 days ago

Verified cf06ad7d

convert : import bytes_to_unicode from convert_slow_tokenizer (#26217)

SolshineCode committed 6 days ago

Verified b06fbc96

model : allow reshape of tensors during load (#26531)

ggerganov committed 6 days ago

Verified 1269cb1f

llama : move n_vocab from llama_sampler_data to penalty_sampler (#26520)

ggerganov committed 6 days ago

Verified 935cad64

ci: fix vulkan llvmpipe runs (#26533)

netrunnereve committed 6 days ago

Verified 22dc605c

sycl: parallelize the non-contiguous concat kernel (#25852)

Titaniumtown committed 6 days ago

Verified 6c8dcaa7

Extended SYCL oneDNN SDPA to non-FP16 KV caches (Q4_0–Q8_0 and FP32) (#25874)

johnkarlhill committed 6 days ago

Verified 66fa168a

chat : add new template for DeepSeek V4 Flash 0731 (#26398)

tarruda committed 6 days ago

Verified 0ef6e55e

vendor : update cpp-httplib to 0.52.0 (#26485)

cabelo committed 7 days ago

Verified 94bc47f2

vendor : update BoringSSL to 0.20260803.0 (#26523)

cabelo committed 7 days ago

Verified fe2adf0e

model : support MTP in GLM-4.7-Flash (#24868)

jacekpoplawski committed 7 days ago

Verified 57c09213

tests: add model resolution test on synthetic repo listings (#26172)

ServeurpersoCom committed 7 days ago

Verified ee0445c9

server: add get_info tool (#26522)

ngxson committed 7 days ago

Verified 99111b19

vocab : validate default special token ids (#26506)

CISC committed 7 days ago

Verified e8e06f78

ggml: use dynamic allocation for split graph inputs (#22789)

AgoraPete committed 7 days ago

Verified dbadb68e

opencl: route large q6_K lm_head to the flat GEMV (#26427)

wanghqc committed 7 days ago

Verified 39eab74a

graph : fix unused input tensors in minimax m3 graph (#26519)

ggerganov committed 7 days ago

Verified c50b34a1

model: M3: Move MSA into a new memory implementation (#26338)

timkhronos committed 7 days ago

Verified 67d5978b

llama : allocate indexer cache only in "full" indexer layers (#26474)

fairydreaming committed 7 days ago

Verified 563dec81

CUDA: Add backend sampler for penalties sampler (#25262)

kmorennv committed 7 days ago

Verified 96278e39

CUDA: Fix data-races when reusing SMEM in block_reduce (#26385)

ORippler committed 7 days ago

Verified 9bd4c09e

server: add notice for upcoming default port change 8080 --> 9931 (#26508)

ngxson committed 7 days ago

Verified 0b14b87d

server: (tools) add x-tool-cwd header (#26420)

ngxson committed 7 days ago

Verified f2b52a87

model: MTP support for Qwen3-Next (#25589)

yomaytk committed 7 days ago

Verified 4ed2b13f

llama : MTP support for DeepSeek V3.2 (#26457)

fairydreaming committed 7 days ago

Verified 2b63e061

metal: implement DSv4 Lightning Indexer (#25893)

tarruda committed 7 days ago

Verified 1464c62d

metal : add SILU_BACK (#25982)

Blackcyan30 committed 8 days ago

Verified 221f0f63

metal : add F16 support for bin ops (#26465)

ggerganov committed 8 days ago

Verified 9d21b57f

opencl: limit local workgroup size for GLU operation (#26383)

mgroeber9110 committed 8 days ago

Verified 0ab9d6fe

metal: implement DeepSeek V4 hyper-connections (#26459)

ggerganov committed 8 days ago

Verified fffbcbdb

Newer Older