Search results

rpc : fix alloc size logic

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Status: Open (in progress).

ggml-org/llama.cppnumber 17116

#17116 In ggml-org/llama.cpp;

· ggerganov opened on Nov 9, 2025

graph : reuse SSM graphs

Status: Open (in progress).

ggml-org/llama.cppnumber 16490

#16490 In ggml-org/llama.cpp;

· ggerganov opened on Oct 9, 2025

batch : add <code>pad_equal</code> [RFC]

Status: Draft (not ready).

ggml-org/llama.cppnumber 15636

#15636 In ggml-org/llama.cpp;

· ggerganov opened on Aug 28, 2025

repack : optimize mul_mat_id path

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Status: Draft (not ready).

ggml-org/llama.cppnumber 14918

#14918 In ggml-org/llama.cpp;

· ggerganov opened on Jul 28, 2025

metal : reuse graphs

Apple Metalhttps://en.wikipedia.org/wiki/Metal_(API)https://en.wikipedia.org/wiki/Metal_(API)

demoDemonstrate some concept or idea, not intended to be mergedDemonstrate some concept or idea, not intended to be merged

ggmlchanges relating to the ggml tensor library for machine learningchanges relating to the ggml tensor library for machine learning

Status: Draft (not ready).

ggml-org/llama.cppnumber 14570

#14570 In ggml-org/llama.cpp;

· ggerganov opened on Jul 7, 2025

server : improve infill stop criteria

examples

server

Status: Draft (not ready).

ggml-org/llama.cppnumber 12333

#12333 In ggml-org/llama.cpp;

· ggerganov opened on Mar 11, 2025

server : chunked prefill support

examples

server

Status: Draft (not ready).

ggml-org/llama.cppnumber 10718

#10718 In ggml-org/llama.cpp;

· ggerganov opened on Dec 8, 2024

speculative : add infill mode

examples

server

Status: Draft (not ready).

ggml-org/llama.cppnumber 10510

#10510 In ggml-org/llama.cpp;

· ggerganov opened on Nov 26, 2024

metal : use F16 math in mul_mat kernels

Status: Draft (not ready).

ggml-org/llama.cppnumber 10220

#10220 In ggml-org/llama.cpp;

· ggerganov opened on Nov 8, 2024

llama : compute BERT graph with F16 K, V

demoDemonstrate some concept or idea, not intended to be mergedDemonstrate some concept or idea, not intended to be merged

Review Complexity : HighGenerally require indepth knowledge of LLMs or GPUsGenerally require indepth knowledge of LLMs or GPUs

Status: Draft (not ready).

ggml-org/llama.cppnumber 5891

#5891 In ggml-org/llama.cpp;

· ggerganov opened on Mar 5, 2024

llama : add llama_kv_cache_compress

demoDemonstrate some concept or idea, not intended to be mergedDemonstrate some concept or idea, not intended to be merged

Status: Draft (not ready).

ggml-org/llama.cppnumber 5719

#5719 In ggml-org/llama.cpp;

· ggerganov opened on Feb 25, 2024

llama : switch to floating-point token positions

demoDemonstrate some concept or idea, not intended to be mergedDemonstrate some concept or idea, not intended to be merged

refactoringRefactoringRefactoring

Review Complexity : HighGenerally require indepth knowledge of LLMs or GPUsGenerally require indepth knowledge of LLMs or GPUs

Status: Draft (not ready).

ggml-org/llama.cppnumber 5679

#5679 In ggml-org/llama.cpp;

· ggerganov opened on Feb 23, 2024