ggml-org/llama.cpp

Pull Requests Commits

models : make deepstack graphs (e.g. Qwen3 VL) have constant topology

ggerganov committed 6 days ago

Verified 9f8a79ce

ggml : add ggml_build_forward_select

ggerganov committed 6 days ago

Verified 3ff34899

cont : reserve only when changes occur + timing

ggerganov committed 6 days ago

Verified c92df391

cont : fix reserve

ggerganov committed 6 days ago

Verified 16ba37de

ggerganov committed 6 days ago

Verified b8a223da

context : reserve new scheduler when graph topology changes

ggerganov committed 6 days ago

Verified 69a30dfa

common/grammar : replace problematic backtracking regex `[\s\S]*` (#18342)

aldehir committed 7 days ago

Verified cef1d23c

graph : fix graph reuse logic when `n_pos_per_embd > 1` (#18566)

ggerganov committed 7 days ago

Verified c69c7ebc

ggml-cuda: fixes for concurrent streams (#18496)

am17an committed 7 days ago

Verified e57f5233

context : fix reserve token padding to n_seqs (#18536)

ggerganov committed 7 days ago

Verified a554a1ec

CUDA: only allocate FA tmp buffer if needed (#18564)

JohannesGaessler committed 7 days ago

Verified 0f2e42ca

(Bugfix, ggml-cuda) Pool alloc count fix + small size computation type adjustment (#18559)

pl752 committed 7 days ago

Verified 9dba9f53

ggml-hexagon: optimize activation function (#18393)

joeldushouyu committed 7 days ago

Verified bcfc8c3c

vulkan: Optimize GGML_OP_CUMSUM (#18417)

jeffbolznv committed 8 days ago

Verified 18ddaea2

vulkan: Implement mmvq for iq1_s/iq1_m (#18450)

jeffbolznv committed 8 days ago

Verified 706e3f93

model : Maincoder-1B support (#18534)

maincode-prabod committed 8 days ago

Verified 5755e52d

metal : adjust extra size for FA buffer to avoid reallocations (#18545)

ggerganov committed 8 days ago

Verified f38de163

graph : reduce topology branching (#18548)

ggerganov committed 8 days ago

Verified af1e8e1a

vocab : reduce debug logs about non-EOG control tokens (#18541)

ggerganov committed 8 days ago

Verified d84a6a98

rpc : use unordered_map::reserve and emplace (#18513)

struct committed 8 days ago

Verified c6f0e832

cuda : fix copy of large tensors (ggml_nbytes <= INT_MAX assertion) (#18433)

Meet91721 committed 8 days ago

Verified e86f3c22

model : remove modern-bert iswa template (#18529)

CISC committed 8 days ago

Verified 169ee68f

model: support youtu-vl model (#18479)

f291400 committed 9 days ago

Verified ced765be

Add conversion support for IQuestCoderForCausalLM (#18524)

pwilkin committed 9 days ago

Verified 3ccccc83

model : add support for JinaBertModel with non-gated ffn (#18475)

o7si committed 9 days ago

Verified d0a6a314

convert : fix encoding of WPM vocab for BERT models (#18500)

o7si committed 9 days ago

Verified 2b2afade

model: add Solar Open model (#18511)

HelloKS committed 9 days ago

Verified f4f50192

webui: fix code copy stripping XML/HTML tags (#18518)

Anri-Lombard committed 9 days ago

Verified d5574c91

ggml-cuda: remove unneccesary prints on ggml_cuda_init (#18502)

am17an committed 9 days ago

Verified 26831bde

vulkan: extend topk_moe to handle sigmoid w/exp_probs_b for nemotron (#18295)

jeffbolznv committed 9 days ago

Verified be47fb92

Older