scitrera/coldsnap-ds4f-0731-capsules:sparkrun_a4b987a048c6c409_65bf0e4d7330-capture-e71c22a8226d-n580-unit-unit-1

Manifest digest

sha256:e54a66f574fa4a68125097b7dbdd4fe61ecc799ae40f3d92f346d876f33fc383

OS/ARCH

linux/arm64

Compressed size

23.32 GB

Last pushed

about 1 month by scitrera

Type

Image

Manifest digest

sha256:e54a66f574fa4a68125097b7dbdd4fe61ecc799ae40f3d92f346d876f33fc383

Image Layers

Image Layers

1ARG RELEASE0 B

2ARG LAUNCHPAD_BUILD_ARCH0 B

3LABEL org.opencontainers.image.ref.name=ubuntu0 B

4LABEL org.opencontainers.image.version=24.040 B

5ADD file ... in / 28.55 MB

6CMD ["/bin/bash"]0 B

7ENV NVARCH=sbsa0 B

8

ENV NVIDIA_REQUIRE_CUDA=cuda>=13.0 brand=unknown,driver>=535,driver<536 brand=grid,driver>=535,driver<536 brand=tesla,driver>=535,driver<536

0 B

9ENV NV_CUDA_CUDART_VERSION=13.0.96-10 B

10ARG TARGETARCH0 B

11LABEL maintainer=NVIDIA CORPORATION <[email protected]>0 B

12RUN |1 TARGETARCH=arm64 /bin/sh -c4.32 MB

13ENV CUDA_VERSION=13.0.20 B

14RUN |1 TARGETARCH=arm64 /bin/sh -c95.11 MB

15RUN |1 TARGETARCH=arm64 /bin/sh -c186 B

16ENV PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin0 B

17ENV LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib640 B

18COPY NGC-DL-CONTAINER-LICENSE / # buildkit6.72 KB

19ENV NVIDIA_VISIBLE_DEVICES=all0 B

20ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility0 B

21ENV NV_CUDA_LIB_VERSION=13.0.2-10 B

22ENV NV_NVTX_VERSION=13.0.85-10 B

23ENV NV_LIBNPP_VERSION=13.0.1.2-10 B

24ENV NV_LIBNPP_PACKAGE=libnpp-13-0=13.0.1.2-10 B

25ENV NV_LIBCUSPARSE_VERSION=12.6.3.3-10 B

26ENV NV_LIBCUBLAS_PACKAGE_NAME=libcublas-13-00 B

27ENV NV_LIBCUBLAS_VERSION=13.1.0.3-10 B

28ENV NV_LIBCUBLAS_PACKAGE=libcublas-13-0=13.1.0.3-10 B

29ENV NV_LIBNCCL_PACKAGE_NAME=libnccl20 B

30ENV NV_LIBNCCL_PACKAGE_VERSION=2.28.3-10 B

31ENV NCCL_VERSION=2.28.3-10 B

32ENV NV_LIBNCCL_PACKAGE=libnccl2=2.28.3-1+cuda13.00 B

33ARG TARGETARCH0 B

34LABEL maintainer=NVIDIA CORPORATION <[email protected]>0 B

35RUN |1 TARGETARCH=arm64 /bin/sh -c1.52 GB

36RUN |1 TARGETARCH=arm64 /bin/sh -c58.21 KB

37COPY entrypoint.d/ /opt/nvidia/entrypoint.d/ # buildkit1.65 KB

38COPY nvidia_entrypoint.sh /opt/nvidia/ # buildkit1.49 KB

39ENV NVIDIA_PRODUCT_NAME=CUDA0 B

40ENTRYPOINT ["/opt/nvidia/nvidia_entrypoint.sh"]0 B

41ENV NV_CUDA_LIB_VERSION=13.0.2-10 B

42ENV NV_CUDA_CUDART_DEV_VERSION=13.0.96-10 B

43ENV NV_NVML_DEV_VERSION=13.0.87-10 B

44ENV NV_LIBCUSPARSE_DEV_VERSION=12.6.3.3-10 B

45ENV NV_LIBNPP_DEV_VERSION=13.0.1.2-10 B

46ENV NV_LIBNPP_DEV_PACKAGE=libnpp-dev-13-0=13.0.1.2-10 B

47ENV NV_LIBCUBLAS_DEV_PACKAGE_NAME=libcublas-dev-13-00 B

48ENV NV_LIBCUBLAS_DEV_VERSION=13.1.0.3-10 B

49ENV NV_LIBCUBLAS_DEV_PACKAGE=libcublas-dev-13-0=13.1.0.3-10 B

50ENV NV_CUDA_NSIGHT_COMPUTE_VERSION=13.0.2-10 B

51ENV NV_CUDA_NSIGHT_COMPUTE_DEV_PACKAGE=cuda-nsight-compute-13-0=13.0.2-10 B

52ENV NV_LIBNCCL_DEV_PACKAGE_NAME=libnccl-dev0 B

53ENV NV_LIBNCCL_DEV_PACKAGE_VERSION=2.28.3-10 B

54ENV NCCL_VERSION=2.28.3-10 B

55ENV NV_LIBNCCL_DEV_PACKAGE=libnccl-dev=2.28.3-1+cuda13.00 B

56ARG TARGETARCH0 B

57LABEL maintainer=NVIDIA CORPORATION <[email protected]>0 B

58RUN |1 TARGETARCH=arm64 /bin/sh -c2.01 GB

59RUN |1 TARGETARCH=arm64 /bin/sh -c87.71 KB

60ENV LIBRARY_PATH=/usr/local/cuda/lib64/stubs0 B

61ARG TORCH_VERSION0 B

62ARG TORCHVISION_VERSION0 B

63ARG TORCHAUDIO_VERSION0 B

64ARG CUTLASS_DSL_VERSION0 B

65ARG B12X_REPO0 B

66ARG B12X_REF0 B

67ARG B12X_CACHEBUST0 B

68ARG BUILD_JOBS0 B

69ENV MAX_JOBS=160 B

70ENV CMAKE_BUILD_PARALLEL_LEVEL=160 B

71ENV NINJAFLAGS=-j160 B

72ENV MAKEFLAGS=-j160 B

73ENV DG_JIT_USE_NVRTC=00 B

74ENV USE_CUDNN=10 B

75ENV DEBIAN_FRONTEND=noninteractive0 B

76ENV PIP_BREAK_SYSTEM_PACKAGES=10 B

77ENV VLLM_BASE_DIR=/workspace/vllm0 B

78ENV PIP_CACHE_DIR=/root/.cache/pip0 B

79ENV UV_CACHE_DIR=/root/.cache/uv0 B

80ENV UV_SYSTEM_PYTHON=10 B

81ENV UV_BREAK_SYSTEM_PACKAGES=10 B

82ENV UV_LINK_MODE=copy0 B

83RUN |8 TORCH_VERSION=2.13.0 TORCHVISION_VERSION=0.28.0 TORCHAUDIO_VERSION=2.11.0760.54 MB

84WORKDIR /workspace/vllm118 B

85RUN |8 TORCH_VERSION=2.13.0 TORCHVISION_VERSION=0.28.0 TORCHAUDIO_VERSION=2.11.02.39 MB

86ARG PRE_TRANSFORMERS=00 B

87RUN |9 TORCH_VERSION=2.13.0 TORCHVISION_VERSION=0.28.0 TORCHAUDIO_VERSION=2.11.02.89 GB

88RUN |9 TORCH_VERSION=2.13.0 TORCHVISION_VERSION=0.28.0 TORCHAUDIO_VERSION=2.11.02.95 GB

89ARG TORCH_CUDA_ARCH_LIST=12.1a0 B

90ENV TORCH_CUDA_ARCH_LIST=12.1a0 B

91ARG FLASHINFER_CUDA_ARCH_LIST=12.1a0 B

92ENV FLASHINFER_CUDA_ARCH_LIST=12.1a0 B

93ENV TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas0 B

94ENV TIKTOKEN_ENCODINGS_BASE=/workspace/vllm/tiktoken_encodings0 B

95

ENV PATH=/workspace/vllm:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin

0 B

96RUN |11 TORCH_VERSION=2.13.0 TORCHVISION_VERSION=0.28.0 TORCHAUDIO_VERSION=2.11.0301.66 MB

97COPY docker/pin_cutlass_dsl.py /tmp/pin_cutlass_dsl.py # buildkit1.04 KB

98RUN |11 TORCH_VERSION=2.13.0 TORCHVISION_VERSION=0.28.0 TORCHAUDIO_VERSION=2.11.01.32 MB

99RUN |11 TORCH_VERSION=2.13.0 TORCHVISION_VERSION=0.28.0 TORCHAUDIO_VERSION=2.11.0295 B

100COPY build-metadata.yaml /workspace/build-metadata.yaml # buildkit548 B

101ARG SPARKRUN_REPO_COMMIT=0e943a4470d65f4d632e80df9cb05843c891ba3b0 B

102ARG SPARKRUN_VLLM_HASH=b5f995e73e6b7fe27c9927477e277a151ebcc9e90 B

103ARG SPARKRUN_VLLM_VERSION=0.1.dev20133+gb5f995e73.d202608230 B

104ARG SPARKRUN_FLASHINFER_HASH=2f519edced8b027f8aee960bc622346e510e2cfa0 B

105ARG SPARKRUN_BUILD_FLAGS=0 B

106ARG SPARKRUN_SOURCE_IMAGE=docker.io/eugr/spark-vllm-b12x0 B

107ARG SPARKRUN_SOURCE_DIGEST=sha256:7dc02f162929943ba2e14514066ed2a04bb7e9ed3592d4eb460ebcbb1f8376bd0 B

108ARG SPARKRUN_SOURCE_TAG=nightly-202608230 B

109

LABEL dev.sparkrun.repo-commit=0e943a4470d65f4d632e80df9cb05843c891ba3b dev.sparkrun.vllm-hash=b5f995e73e6b7fe27c9927477e277a151ebcc9e9 dev.sparkrun.vllm-version=0.1.dev20133+gb5f995e73.d20260823 dev.sparkrun.flashinfer-hash=2f519edced8b027f8aee960bc622346e510e2cfa

0 B

110ARG VLLM_IMAGE0 B

111ARG CRIU_IMAGE0 B

112

LABEL io.sparksq.coldsnap.runtime=vllm-cuda-criu-v1 io.sparksq.coldsnap.base-image=ghcr.io/spark-arena/dgx-vllm-eugr-nightly-b12x@sha256:af9629c5bc9f5edc1b478a996cf4aa4304311524f31e61ab66e59fce8ff8fd02 io.sparksq.coldsnap.criu-image=ghcr.io/sparksq/criu@sha256:2ff53a61af48e7e676bd4d64747394ca7c7622840ef0c740e719e6ecadb0d07c io.sparksq.coldsnap.copyright-holders=Scitrera

0 B

113

RUN |2 VLLM_IMAGE=ghcr.io/spark-arena/dgx-vllm-eugr-nightly-b12x@sha256:af9629c5bc9f5edc1b478a996cf4aa4304311524f31e61ab66e59fce8ff8fd02 CRIU_IMAGE=ghcr.io/sparksq/criu@sha256:2ff53a61af48e7e676bd4d64747394ca7c7622840ef0c740e719e6ecadb0d07c /bin/sh

4.64 MB

114COPY /usr/sbin/criu /opt/coldsnap/criu/bin/criu # buildkit2.87 MB

115COPY LICENSE THIRD_PARTY_NOTICES.md /opt/coldsnap/licenses/coldsnap/ #13.26 KB

116COPY third_party/licenses/criu/COPYING /opt/coldsnap/licenses/third-party/criu/COPYING # buildkit12.14 KB

117COPY third_party/licenses/nccl/LICENSE.txt /opt/coldsnap/licenses/third-party/nccl/LICENSE.txt # buildkit1.21 KB

118COPY LICENSE /opt/coldsnap/licenses/third-party/go-criu/LICENSE # buildkit4.06 KB

119COPY LICENSE /opt/coldsnap/licenses/third-party/cuda-checkpoint/LICENSE # buildkit14.82 KB

120

RUN |2 VLLM_IMAGE=ghcr.io/spark-arena/dgx-vllm-eugr-nightly-b12x@sha256:af9629c5bc9f5edc1b478a996cf4aa4304311524f31e61ab66e59fce8ff8fd02 CRIU_IMAGE=ghcr.io/sparksq/criu@sha256:2ff53a61af48e7e676bd4d64747394ca7c7622840ef0c740e719e6ecadb0d07c /bin/sh

110.39 KB

121COPY /out/coldsnap /usr/local/bin/coldsnap # buildkit3.25 MB

122COPY /out/coldsnap-coordinator /usr/local/bin/coldsnap-coordinator # buildkit1.87 MB

123COPY /out/coldsnap-criu-rpc /usr/local/bin/coldsnap-criu-rpc # buildkit3.91 MB

124COPY bin /tmp/cuda-checkpoint-bin # buildkit33.36 KB

125COPY /out/libcoldsnap_hydration.so /opt/coldsnap/native/libcoldsnap_hydration.so # buildkit45.71 KB

126COPY /out/libcoldsnap_cuda_epoch.so /opt/coldsnap/native/libcoldsnap_cuda_epoch.so # buildkit17.92 KB

127

COPY /usr/lib/criu/coldsnap_nvidia_reset_plugin.so /opt/coldsnap/criu/plugins/libcoldsnap_criu_nvidia_reset.so # buildkit

6.53 KB

128COPY /out/libcoldsnap_nvidia_mmap_shim.so /opt/coldsnap/native/libcoldsnap_nvidia_mmap_shim.so # buildkit6.3 KB

129COPY /out/libcoldsnap_nvml_dlopen_shim.so /opt/coldsnap/native/libcoldsnap_nvml_dlopen_shim.so # buildkit4.98 KB

130COPY . /tmp/coldsnap-nccl-provider/ # buildkit34.59 MB

131COPY /out/libcoldsnap-nccl-dlsym.so /tmp/libcoldsnap-nccl-dlsym.so # buildkit3.65 KB

132COPY /out/plugin /opt/coldsnap/plugin # buildkit185.66 KB

133COPY runtime/shared/coldsnap_coord.py /opt/coldsnap/runtime/coldsnap_coord.py # buildkit2.33 KB

134COPY runtime/engine/coldsnap_cuda_criu.py /opt/coldsnap/runtime/coldsnap_cuda_criu.py # buildkit7.11 KB

135COPY runtime/engine/coldsnap_engine_exec.py /opt/coldsnap/runtime/coldsnap-engine-exec.py # buildkit7.1 KB

136COPY runtime/engine/coldsnap_n580_criu.py /opt/coldsnap/runtime/coldsnap_n580_criu.py # buildkit7.36 KB

137

RUN |2 VLLM_IMAGE=ghcr.io/spark-arena/dgx-vllm-eugr-nightly-b12x@sha256:af9629c5bc9f5edc1b478a996cf4aa4304311524f31e61ab66e59fce8ff8fd02 CRIU_IMAGE=ghcr.io/sparksq/criu@sha256:2ff53a61af48e7e676bd4d64747394ca7c7622840ef0c740e719e6ecadb0d07c /bin/sh

43.72 MB

138

ENV PYTHONPATH=/opt/coldsnap/plugin:/opt/coldsnap/runtime VLLM_PLUGINS=coldsnap COLDSNAP_CUDA_EPOCH_NATIVE_LIBRARY=/opt/coldsnap/native/libcoldsnap_cuda_epoch.so COLDSNAP_NCCL_ACTIVE_RUNTIME_PATH=/opt/coldsnap/nccl/active.json

0 B

139ARG SOURCE_DATE_EPOCH=00 B

140

LABEL io.sparksq.coldsnap.kind=unit-capsule io.sparksq.coldsnap.capture=sparkrun_a4b987a048c6c409_65bf0e4d7330-capture-e71c22a8226d io.sparksq.coldsnap.unit=unit-1 io.sparksq.coldsnap.snapshot-driver=n580

0 B

141COPY --chown=0:0 / / #68.53 MB

142COPY --chown=0:0 . /opt/coldsnap/capsule/ #12.63 GB

143RUN |1 SOURCE_DATE_EPOCH=0 /bin/sh -c345 B

Command

 ARG RELEASE