From 0b64d2139ca253251be7bd7b890606a853107055 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Tue, 23 Sep 2025 17:09:55 +0300 Subject: [PATCH 1/3] ci : disable AMD workflows + update NVIDIA workflows --- .github/workflows/build.yml | 97 ++++++++++++++----------------------- 1 file changed, 36 insertions(+), 61 deletions(-) diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml index 8f8f48607172b..d73452a1734d8 100644 --- a/.github/workflows/build.yml +++ b/.github/workflows/build.yml @@ -1302,8 +1302,8 @@ jobs: run: | GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - ggml-ci-x64-nvidia-v100-cuda: - runs-on: [self-hosted, Linux, X64, NVIDIA, V100] + ggml-ci-x64-nvidia-cuda: + runs-on: [self-hosted, Linux, X64, NVIDIA] steps: - name: Clone @@ -1316,8 +1316,8 @@ jobs: nvidia-smi GG_BUILD_CUDA=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - ggml-ci-x64-nvidia-v100-vulkan: - runs-on: [self-hosted, Linux, X64, NVIDIA, V100] + ggml-ci-x64-nvidia-vulkan: + runs-on: [self-hosted, Linux, X64, NVIDIA] steps: - name: Clone @@ -1327,11 +1327,11 @@ jobs: - name: Test id: ggml-ci run: | - vulkaninfo + vulkaninfo --summary GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - ggml-ci-x64-nvidia-t4-cuda: - runs-on: [self-hosted, Linux, X64, NVIDIA, T4] + ggml-ci-x64-nvidia-vulkan-coopmat1: + runs-on: [self-hosted, Linux, X64, NVIDIA] steps: - name: Clone @@ -1341,35 +1341,7 @@ jobs: - name: Test id: ggml-ci run: | - nvidia-smi - GG_BUILD_CUDA=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - - ggml-ci-x64-nvidia-t4-vulkan: - runs-on: [self-hosted, Linux, X64, NVIDIA, T4] - - steps: - - name: Clone - id: checkout - uses: actions/checkout@v4 - - - name: Test - id: ggml-ci - run: | - vulkaninfo - GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - - ggml-ci-x64-nvidia-t4-vulkan-coopmat1: - runs-on: [self-hosted, Linux, X64, NVIDIA, T4] - - steps: - - name: Clone - id: checkout - uses: actions/checkout@v4 - - - name: Test - id: ggml-ci - run: | - vulkaninfo + vulkaninfo --summary GG_BUILD_VULKAN=1 GGML_VK_DISABLE_COOPMAT2=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp ggml-ci-x64-cpu-amx: @@ -1385,31 +1357,33 @@ jobs: run: | bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - ggml-ci-x64-amd-v710-vulkan: - runs-on: [self-hosted, Linux, X64, AMD, V710] - - steps: - - name: Clone - id: checkout - uses: actions/checkout@v4 - - - name: Test - id: ggml-ci - run: | - GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - - ggml-ci-x64-amd-v710-rocm: - runs-on: [self-hosted, Linux, X64, AMD, V710] - - steps: - - name: Clone - id: checkout - uses: actions/checkout@v4 - - - name: Test - id: ggml-ci - run: | - GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp +# ggml-ci-x64-amd-vulkan: +# runs-on: [self-hosted, Linux, X64, AMD] +# +# steps: +# - name: Clone +# id: checkout +# uses: actions/checkout@v4 +# +# - name: Test +# id: ggml-ci +# run: | +# vulkaninfo --summary +# GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp +# +# ggml-ci-x64-amd-rocm: +# runs-on: [self-hosted, Linux, X64, AMD] +# +# steps: +# - name: Clone +# id: checkout +# uses: actions/checkout@v4 +# +# - name: Test +# id: ggml-ci +# run: | +# amd-smi +# GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp ggml-ci-mac-metal: runs-on: [self-hosted, macOS, ARM64] @@ -1435,4 +1409,5 @@ jobs: - name: Test id: ggml-ci run: | + vulkaninfo --summary GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp ~/mnt/llama.cpp From acca45225ea0eb1353ff32c36192e70f57483648 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Tue, 23 Sep 2025 17:30:47 +0300 Subject: [PATCH 2/3] cont : fixes --- .github/workflows/build.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml index d73452a1734d8..5c0bc4fdef078 100644 --- a/.github/workflows/build.yml +++ b/.github/workflows/build.yml @@ -1331,7 +1331,7 @@ jobs: GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp ggml-ci-x64-nvidia-vulkan-coopmat1: - runs-on: [self-hosted, Linux, X64, NVIDIA] + runs-on: [self-hosted, Linux, X64, NVIDIA, COOPMAT2] steps: - name: Clone @@ -1382,7 +1382,7 @@ jobs: # - name: Test # id: ggml-ci # run: | -# amd-smi +# amd-smi static # GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp ggml-ci-mac-metal: From ba238466e2f68547316172b05f6f6ccb6fa60ec8 Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Tue, 23 Sep 2025 19:47:37 +0300 Subject: [PATCH 3/3] cont : update nvidia vulkan workflows --- .github/workflows/build.yml | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml index 5c0bc4fdef078..a9ab68c3e8588 100644 --- a/.github/workflows/build.yml +++ b/.github/workflows/build.yml @@ -1316,7 +1316,7 @@ jobs: nvidia-smi GG_BUILD_CUDA=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - ggml-ci-x64-nvidia-vulkan: + ggml-ci-x64-nvidia-vulkan-cm: runs-on: [self-hosted, Linux, X64, NVIDIA] steps: @@ -1328,9 +1328,9 @@ jobs: id: ggml-ci run: | vulkaninfo --summary - GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp + GG_BUILD_VULKAN=1 GGML_VK_DISABLE_COOPMAT2=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp - ggml-ci-x64-nvidia-vulkan-coopmat1: + ggml-ci-x64-nvidia-vulkan-cm2: runs-on: [self-hosted, Linux, X64, NVIDIA, COOPMAT2] steps: @@ -1342,7 +1342,7 @@ jobs: id: ggml-ci run: | vulkaninfo --summary - GG_BUILD_VULKAN=1 GGML_VK_DISABLE_COOPMAT2=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp + GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp ggml-ci-x64-cpu-amx: runs-on: [self-hosted, Linux, X64, CPU, AMX]