Search results

Misc. bug: llama-finetune won't work even with 17M parameters arch:llama

bugSomething isn't workingSomething isn't working

help wantedNeeds help from the communityNeeds help from the community

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

Status: Open.

#18499 In ggml-org/llama.cpp;

· yousefabdel1727 opened on Dec 30, 2025

mtmd: plan to add video input support

enhancementNew feature or requestNew feature or request

hotSomething that is hotSomething that is hot

mtmdRelated to multimodal functionality (video/image/audio)Related to multimodal functionality (video/image/audio)

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

Status: Open.

#18389 In ggml-org/llama.cpp;

· ngxson opened on Dec 26, 2025

Feature Request: support the multi-modal Bagel model

enhancementNew feature or requestNew feature or request

help wantedNeeds help from the communityNeeds help from the community

modelModel specificModel specific

mtmdRelated to multimodal functionality (video/image/audio)Related to multimodal functionality (video/image/audio)

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

Status: Open.

#17861 In ggml-org/llama.cpp;

· ghost opened on Dec 8, 2025

Feature Request: support Qwen3-omni, Qwen3-TTS and MiniCPM-o

enhancementNew feature or requestNew feature or request

good first issueGood for newcomersGood for newcomers

help wantedNeeds help from the communityNeeds help from the community

mtmdRelated to multimodal functionality (video/image/audio)Related to multimodal functionality (video/image/audio)

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

ttsText-to-speechText-to-speech

Status: Open.

Feature

#17634 In ggml-org/llama.cpp;

· ghost opened on Dec 1, 2025

ggml-cuda : add TQ2_0 kernels, for ternary inference on GPU

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

enhancementNew feature or requestNew feature or request

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

Nvidia GPUIssues specific to Nvidia GPUsIssues specific to Nvidia GPUs

performanceSpeed related topicsSpeed related topics

pythonpython script changespython script changes

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

testingEverything test relatedEverything test related

Status: Open (in progress).

ggml-org/llama.cppnumber 11183

#11183 In ggml-org/llama.cpp;

· compilade opened on Jan 10, 2025

ggml: skip excess iteration for pair whose vars same element when i2 == i1

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

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

Status: Open (in progress).

ggml-org/llama.cppnumber 9177

#9177 In ggml-org/llama.cpp;

· GermanAizek opened on Aug 25, 2024

ggml : make GeLU faster and more accurate on CPU

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

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

Status: Open (in progress).

ggml-org/llama.cppnumber 8878

#8878 In ggml-org/llama.cpp;

· jart opened on Aug 5, 2024

Add support for loongarch backend in sgemm.cpp

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

Status: Open (in progress).

ggml-org/llama.cppnumber 8726

#8726 In ggml-org/llama.cpp;

· Tianzhengshuyuan opened on Jul 27, 2024

Introduce Q8_0 and Q4_0 with Bf16 delta values

examples

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

pythonpython script changespython script changes

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

Tensor Encoding Schemehttps://github.com/ggerganov/llama.cpp/wiki/Tensor-Encoding-Schemeshttps://github.com/ggerganov/llama.cpp/wiki/Tensor-Encoding-Schemes

Status: Open (in progress).

ggml-org/llama.cppnumber 7497

#7497 In ggml-org/llama.cpp;

· Srihari-mcw opened on May 23, 2024

Add token healing to main and server

enhancementNew feature or requestNew feature or request

examples

help wantedNeeds help from the communityNeeds help from the community

need feedbackTesting and feedback with results are neededTesting and feedback with results are needed

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

server

Status: Open (in progress).

ggml-org/llama.cppnumber 7187

#7187 In ggml-org/llama.cpp;

· mare5x opened on May 9, 2024

support MiniCPM-V-2

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

enhancementNew feature or requestNew feature or request

examples

pythonpython script changespython script changes

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

Status: Open (in progress).

ggml-org/llama.cppnumber 6919

#6919 In ggml-org/llama.cpp;

· Achazwl opened on Apr 26, 2024

cuda : use amd wave sharing intrinsics for warp_reduce functions

performanceSpeed related topicsSpeed related topics

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

Status: Open (in progress).

ggml-org/llama.cppnumber 6522

#6522 In ggml-org/llama.cpp;

· Engininja2 opened on Apr 7, 2024