Commits

Breadcrumbs

History for

onmain

User selector

All users

All time

Commit History

Commits on Dec 9, 2025

• Attempting to fix the AVX2 workflow: Missing GGML_FMA and GGML_F16C may cause an OSError: [WinError -1073741795] Windows Error 0xc000001d error on processors that support AVX2 instructions.

  JamePeng

  committed
  fcfde24

  Copy full SHA for fcfde24

Commits on Dec 7, 2025

• workflow: Fixed nvcc fatal : Unsupported gpu architecture 'compute_101'

  JamePeng

  committed
  4189786

  Copy full SHA for 4189786

• workflow: Added workflows for compiling with CUDA 13.0.2 on Windows and Linux.

  JamePeng

  committed
  2b97b7e

  Copy full SHA for 2b97b7e

• Remove outdated CUDA workflows

  JamePeng

  committed
  61e2346

  Copy full SHA for 61e2346

Commits on Dec 2, 2025

• Fixed Basic workflow typos

  JamePeng

  committed
  5d49287

  Copy full SHA for 5d49287

• Add new workflow with Basic options for Linux platform (CUDA 12.4, 12.6, 12.8)

  JamePeng

  committed
  22034f5

  Copy full SHA for 22034f5

Commits on Nov 21, 2025

• Sync cmake : add option to build and link BoringSSL and OpenSSL

  JamePeng

  committed
  01ed775

  Copy full SHA for 01ed775

Commits on Nov 17, 2025

• Add a Basic workflow for cu128 windows wheels

  JamePeng

  committed
  cfb4f8f

  Copy full SHA for cfb4f8f

Commits on Nov 12, 2025

• Use httplib to download model from an URL when the libcurl is disabled

  Show description for 4e901b7
  JamePeng

  committed
  4e901b7

  Copy full SHA for 4e901b7

Commits on Nov 10, 2025

• Refining the AVX instruction compilation workflow

  JamePeng

  committed
  dd38ec6

  Copy full SHA for dd38ec6

Commits on Nov 9, 2025

• Adding workflow compilation options for platforms without AVX instructions or AVXVNNI instruction can be achieved by setting the releasetag parameter above to Basic/AVXVNNI.

  Show description for 5abcf51
  JamePeng

  committed
  5abcf51

  Copy full SHA for 5abcf51

Commits on Nov 3, 2025

• Optimize Windows CUDA Wheel Build Workflow

  Show description for c008d9e
  JamePeng

  committed
  c008d9e

  Copy full SHA for c008d9e

Commits on Oct 23, 2025

• Separate cu128 compilation from the windows github action and supplement sm_87 sm101 compilation

  JamePeng

  committed
  4f98b28

  Copy full SHA for 4f98b28

• Use the N-Storm/cuda-toolkit instead of Jimver/cuda-toolkit

  JamePeng

  committed
  6ec4e64

  Copy full SHA for 6ec4e64

Commits on Sep 24, 2025

• Adjust test.yaml to macos14

  JamePeng

  committed
  94e8eb5

  Copy full SHA for 94e8eb5

• Fixed the huggingface_hub command error

  JamePeng

  committed
  128e23c

  Copy full SHA for 128e23c

Commits on Sep 13, 2025

• Simplify the code structure of test.yaml

  JamePeng

  committed
  44e5534

  Copy full SHA for 44e5534

Commits on Aug 21, 2025

• Try to add python3.13 version compilation

  JamePeng

  committed
  62bea58

  Copy full SHA for 62bea58

Commits on Aug 1, 2025

• Remove sm_70 in cuda12.8.1 action flow

  JamePeng

  committed
  fcbb32e

  Copy full SHA for fcbb32e

• Try to compile with cuda12.8.1 for Blackwell architecture (sm_100 and sm_120) computing cards

  JamePeng

  committed
  a617e31

  Copy full SHA for a617e31

Commits on Jul 8, 2025

• Sync abetlen/llama-cpp-python 0.3.12 code

  JamePeng

  committed
  c7a1e13

  Copy full SHA for c7a1e13

Commits on Jul 6, 2025

• Sync abetlen/llama-cpp-python 0.3.11 code

  JamePeng

  committed
  0701c73

  Copy full SHA for 0701c73

Commits on Jun 20, 2025

• modify setup-msbuild@v2 remove vs-version

  JamePeng

  committed
  40f3403

  Copy full SHA for 40f3403

Commits on Jun 10, 2025

• pin gh-release action to v2.2.2

  JamePeng

  committed
  3e0bb66

  Copy full SHA for 3e0bb66

Commits on Jun 3, 2025

• Fixed Jimver/cuda-toolkit last normal version 0.2.23

  JamePeng

  committed
  5b63721

  Copy full SHA for 5b63721

Commits on May 20, 2025

• build-wheels-linux -DCMAKE_CUDA_ARCHITECTURES add 70-real;75-real

  JamePeng

  committed
  c785a54

  Copy full SHA for c785a54

Commits on May 4, 2025

• Update Submodule vendor/llama.cpp e1e8e09..8ae5ebc

  JamePeng

  committed
  2e89e3f

  Copy full SHA for 2e89e3f

Commits on May 3, 2025

• Split cu124 and cu126 build-wheels-linux.yml:

  Show description for 623f7ed
  JamePeng

  committed
  623f7ed

  Copy full SHA for 623f7ed

• Try to enable LLAVA_BUILD in build-wheels-cuda-linux.yml

  JamePeng

  committed
  c108044

  Copy full SHA for c108044

Commits on May 2, 2025

• Upload build-wheels-cuda-linux.yml

  JamePeng

  committed
  786d15c

  Copy full SHA for 786d15c

Commits on May 1, 2025

• Update build-wheels-cuda-win.yml: use astral-sh/setup-uv@v6 for setup-uv option

  JamePeng

  committed
  b0ea709

  Copy full SHA for b0ea709

Commits on Apr 8, 2025

• add -DLLAMA_CURL=OFF cmake arg avoid the the error "Could NOT find CURL"

  JamePeng

  committed
  a505718

  Copy full SHA for a505718

Commits on Mar 18, 2025

• Build Wheels (CUDA) for Windows workflow append build date tag

  JamePeng

  committed
  b750a5f

  Copy full SHA for b750a5f

Commits on Mar 13, 2025

• Add new Build Wheels (CUDA) for Windows workflow

  Show description for 37ae873
  JamePeng

  committed
  37ae873

  Copy full SHA for 37ae873

Commits on Jan 29, 2025

• fix(ci): Fix the CUDA workflow (#1894)

  oobabooga

  authored
  14879c7

  Copy full SHA for 14879c7