ℹ️
注意:这是一个 latest 标签镜像

latest 并不代表最新版本,本站同步时间存在延迟,无法保证此镜像与上游最新版本一致
生产环境建议使用明确的版本号(如 v1.2.3),避免因版本不一致导致问题。 了解更多 →

logo
ghcr.io/aeon-7/aeon-vllm-ultimate:latest
linux/arm64 ghcr.io 请确认架构匹配

该镜像集成了vllm框架的终极版本,用于大语言模型的高效推理与部署,支持快速处理大模型推理请求,包含优化的配置及依赖,可帮助用户快速启动大语言模型服务。

14
浏览次数
52.30GB
镜像大小
国内镜像
swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64
源镜像
ghcr.io/aeon-7/aeon-vllm-ultimate:latest
镜像ID
sha256:715cf21dee93ebc45f6f37978b355ed2bd52231d9cf9bd6d9e8dc13b55195461
镜像 TAG
latest-linuxarm64
镜像大小
52.30GB
平台架构
linux/arm64
镜像源
ghcr.io
CMD
启动入口
/bin/bash
工作目录
/
OS/平台
linux/arm64
镜像创建
2026-09-11T06:32:06.171995256-05:00
同步时间
2026-09-18 06:20
浏览量
14 次
贡献者
⚙️ 环境变量 30
KeyValue
PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin 0
NVARCH=sbsa 1
NVIDIA_REQUIRE_CUDA=cuda>=13.0 brand=unknown,driver>=535,driver<536 brand=grid,driver>=535,driver<536 brand=tesla,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=vapps,driver>=535,driver<536 brand=vpc,driver>=535,driver<536 brand=vcs,driver>=535,driver<536 brand=vws,driver>=535,driver<536 brand=cloudgaming,driver>=535,driver<536 brand=unknown,driver>=550,driver<551 brand=grid,driver>=550,driver<551 brand=tesla,driver>=550,driver<551 brand=nvidia,driver>=550,driver<551 brand=quadro,driver>=550,driver<551 brand=quadrortx,driver>=550,driver<551 brand=nvidiartx,driver>=550,driver<551 brand=vapps,driver>=550,driver<551 brand=vpc,driver>=550,driver<551 brand=vcs,driver>=550,driver<551 brand=vws,driver>=550,driver<551 brand=cloudgaming,driver>=550,driver<551 brand=unknown,driver>=565,driver<566 brand=grid,driver>=565,driver<566 brand=tesla,driver>=565,driver<566 brand=nvidia,driver>=565,driver<566 brand=quadro,driver>=565,driver<566 brand=quadrortx,driver>=565,driver<566 brand=nvidiartx,driver>=565,driver<566 brand=vapps,driver>=565,driver<566 brand=vpc,driver>=565,driver<566 brand=vcs,driver>=565,driver<566 brand=vws,driver>=565,driver<566 brand=cloudgaming,driver>=565,driver<566 brand=unknown,driver>=570,driver<571 brand=grid,driver>=570,driver<571 brand=tesla,driver>=570,driver<571 brand=nvidia,driver>=570,driver<571 brand=quadro,driver>=570,driver<571 brand=quadrortx,driver>=570,driver<571 brand=nvidiartx,driver>=570,driver<571 brand=vapps,driver>=570,driver<571 brand=vpc,driver>=570,driver<571 brand=vcs,driver>=570,driver<571 brand=vws,driver>=570,driver<571 brand=cloudgaming,driver>=570,driver<571 brand=unknown,driver>=575,driver<576 brand=grid,driver>=575,driver<576 brand=tesla,driver>=575,driver<576 brand=nvidia,driver>=575,driver<576 brand=quadro,driver>=575,driver<576 brand=quadrortx,driver>=575,driver<576 brand=nvidiartx,driver>=575,driver<576 brand=vapps,driver>=575,driver<576 brand=vpc,driver>=575,driver<576 brand=vcs,driver>=575,driver<576 brand=vws,driver>=575,driver<576 brand=cloudgaming,driver>=575,driver<576 2
NV_CUDA_CUDART_VERSION=13.0.96-1 3
CUDA_VERSION=13.0.2 4
LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64 5
NVIDIA_VISIBLE_DEVICES=all 6
NVIDIA_DRIVER_CAPABILITIES=compute,utility 7
DEBIAN_FRONTEND=noninteractive 8
UV_HTTP_TIMEOUT=500 9
UV_INDEX_STRATEGY=unsafe-best-match 10
UV_LINK_MODE=copy 11
VLLM_ENABLE_CUDA_COMPATIBILITY=0 12
TORCH_CUDA_ARCH_LIST=12.1a 13
VLLM_USAGE_SOURCE=production-docker-image 14
VLLM_USE_FLASHINFER_SAMPLER=1 15
ENABLE_NVFP4_SM100=0 16
CCACHE_DISABLE=1 17
PIP_DEFAULT_TIMEOUT=180 18
PIP_RETRIES=5 19
CMAKE_BUILD_PARALLEL_LEVEL=8 20
MAX_JOBS=12 21
NVCC_THREADS=2 22
SETUPTOOLS_SCM_PRETEND_VERSION=0.29.0+aeon.sm121a.dspark 23
SETUPTOOLS_SCM_PRETEND_VERSION_FOR_VLLM=0.29.0+aeon.sm121a.dspark 24
TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas 25
VLLM_ALLREDUCE_USE_FLASHINFER=0 26
CC=gcc-12 27
CXX=g++-12 28
CUDAHOSTCXX=g++-12 29
🏷️ 镜像标签 17
KeyValue
z-lab/gemma-4-26B-A4B-it-DFlash ai.aeon.drafter
gemma4,qwen3.6,dflash,dflash-swa-mrv2,dspark,dspark-markov-heads,dspark-markov-quantized,dspark-topk,nvfp4,nvfp4-awq,nvfp4-kv,nvfp4-kv-nhd-safe,fp8-kv,spec-kv-dtype,hybrid-apc-fix,packed-kv-zeroer-fix,gdn-gate-align,flashinfer-0.6.18,torch-2.13.0,triton-3.7.1,cutlass-dsl-4.6.0,quack-0.6.1,tvm-ffi-0.1.11,tilelang-0.1.12,nccl-2.30.7,torchcodec-0.16.0,uma-clamp,thread-local-capture,tp2-ready,turboquant,tool-calling,mrv2-default-routing ai.aeon.features
NVIDIA DGX Spark GB10 SM121 ai.aeon.hardware
fleet: Gemma-4-26B-A4B, Gemma-4-31B-DECKARD (NVFP4_AWQ), Qwen3.6-27B, Qwen3.6-35B-A3B ai.aeon.model
vLLM 0.29.0 (from-source, sm_121a 3-way merge + 8 cherry-picks) ai.aeon.vllm_base
NVIDIA CORPORATION <cudatools@nvidia.com> maintainer
2026-05-06T19:21:45Z org.opencontainers.image.created
AEON vLLM Ultimate — vLLM 0.29.0 built from source for DGX Spark / Blackwell (sm_121a/GB10). DSpark Markov heads (quantized, #50424) + top-k projection (#49969), DFlash SWA on MRv2 (upstream #47914/#48113) with V1 carries as fallback, Triton NVFP4-KV (NHD-safe slicing), NVFP4_AWQ, packed-KV zeroer fix (#50276), GDN spec-token gate fix (#51812), hybrid-APC fix (#51843), DSpark bugfix set, thread_local capture at all graph sites (TP=2 2x-Spark ready, eager-first cross-node), torch 2.13.0 + Triton 3.7.1 + FlashInfer 0.6.18 trio + NCCL 2.30.7. org.opencontainers.image.description
https://github.com/AEON-7/vllm-ultimate-dgx-spark org.opencontainers.image.documentation
Apache-2.0 org.opencontainers.image.licenses
ubuntu org.opencontainers.image.ref.name
06e292d0ce7e0ddc4f84bd200c3bdf55c7875eb7 org.opencontainers.image.revision
https://github.com/AEON-7/vllm-ultimate-dgx-spark org.opencontainers.image.source
AEON-Gemma-4-26B-A4B-DFLASH-vLLM0201 org.opencontainers.image.title
https://huggingface.co/AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4 org.opencontainers.image.url
AEON-7 org.opencontainers.image.vendor
22.04 org.opencontainers.image.version

Docker拉取命令

docker pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64
docker tag  swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64  ghcr.io/aeon-7/aeon-vllm-ultimate:latest

Containerd拉取命令

ctr images pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64
ctr images tag  swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64  ghcr.io/aeon-7/aeon-vllm-ultimate:latest

Shell快速替换命令

sed -i 's#ghcr.io/aeon-7/aeon-vllm-ultimate:latest#swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64#' deployment.yaml

Ansible快速分发-Docker

#ansible k8s -m shell -a 'docker pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64 && docker tag  swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64  ghcr.io/aeon-7/aeon-vllm-ultimate:latest'

Ansible快速分发-Containerd

#ansible k8s -m shell -a 'ctr images pull swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64 && ctr images tag  swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64  ghcr.io/aeon-7/aeon-vllm-ultimate:latest'

镜像构建历史


# 2026-09-11 19:32:06  0.00B 配置容器启动时运行的命令
ENTRYPOINT ["/bin/bash"]
                        
# 2026-09-11 19:32:06  0.00B 添加元数据标签
LABEL ai.aeon.vllm_base=vLLM 0.29.0 (from-source, sm_121a 3-way merge + 8 cherry-picks) ai.aeon.model=fleet: Gemma-4-26B-A4B, Gemma-4-31B-DECKARD (NVFP4_AWQ), Qwen3.6-27B, Qwen3.6-35B-A3B ai.aeon.hardware=NVIDIA DGX Spark GB10 SM121 ai.aeon.features=gemma4,qwen3.6,dflash,dflash-swa-mrv2,dspark,dspark-markov-heads,dspark-markov-quantized,dspark-topk,nvfp4,nvfp4-awq,nvfp4-kv,nvfp4-kv-nhd-safe,fp8-kv,spec-kv-dtype,hybrid-apc-fix,packed-kv-zeroer-fix,gdn-gate-align,flashinfer-0.6.18,torch-2.13.0,triton-3.7.1,cutlass-dsl-4.6.0,quack-0.6.1,tvm-ffi-0.1.11,tilelang-0.1.12,nccl-2.30.7,torchcodec-0.16.0,uma-clamp,thread-local-capture,tp2-ready,turboquant,tool-calling,mrv2-default-routing org.opencontainers.image.description=AEON vLLM Ultimate — vLLM 0.29.0 built from source for DGX Spark / Blackwell (sm_121a/GB10). DSpark Markov heads (quantized, #50424) + top-k projection (#49969), DFlash SWA on MRv2 (upstream #47914/#48113) with V1 carries as fallback, Triton NVFP4-KV (NHD-safe slicing), NVFP4_AWQ, packed-KV zeroer fix (#50276), GDN spec-token gate fix (#51812), hybrid-APC fix (#51843), DSpark bugfix set, thread_local capture at all graph sites (TP=2 2x-Spark ready, eager-first cross-node), torch 2.13.0 + Triton 3.7.1 + FlashInfer 0.6.18 trio + NCCL 2.30.7. org.opencontainers.image.documentation=https://github.com/AEON-7/vllm-ultimate-dgx-spark org.opencontainers.image.source=https://github.com/AEON-7/vllm-ultimate-dgx-spark org.opencontainers.image.licenses=Apache-2.0
                        
# 2026-09-11 19:32:06  0.00B 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c rm -rf /build /root/.cache/pip # buildkit
                        
# 2026-09-11 19:31:59  39.15MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir --no-deps "vllm-omni==0.27.0rc1" 2>&1 | tail -5     || pip install --no-cache-dir --no-deps "vllm-omni" 2>&1 | tail -5     || echo "[WARN] vllm-omni install failed — speech-out families may be absent; audio-in/vision still gated above" # buildkit
                        
# 2026-09-11 19:31:56  0.00B 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c python3 -c "import soundfile,librosa,torchcodec,cv2,PIL,torchaudio; from torchcodec.decoders import VideoDecoder,AudioDecoder; print('multimodal organs OK')" || echo "[WARN] some MM optional imports failed" # buildkit
                        
# 2026-09-11 19:31:54  0.00B 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c python3 -c "import numpy as np, soundfile as sf, io; from vllm.multimodal.media.audio import load_audio; buf=io.BytesIO(); sig=np.sin(np.arange(24000*2)*0.02).astype('float32'); sf.write(buf, sig, 24000, format='WAV'); buf.seek(0); y,sr=load_audio(buf, sr=16000); assert sr==16000 and len(y)>15000, (sr, len(y)); print('audio resample gate OK: 24kHz ->', sr, 'Hz,', len(y), 'samples')" # buildkit
                        
# 2026-09-11 19:31:49  208.57MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir av soundfile soxr librosa "mistral_common[audio]>=1.11.6" 2>&1 | tail -2     && python3 -c "import av, soundfile, soxr, librosa, mistral_common; print('audio libs OK', av.__version__, mistral_common.__version__)" # buildkit
                        
# 2026-09-11 19:31:37  0.00B 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c python3 /tmp/nvfp4_kv_gate.py && rm /tmp/nvfp4_kv_gate.py # buildkit
                        
# 2026-09-11 19:31:34  2.47KB 复制新文件或目录到容器中
COPY nvfp4_kv_gate.py /tmp/nvfp4_kv_gate.py # buildkit
                        
# 2026-09-11 19:31:34  252.63KB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c mkdir -p /tmp/cuda-stub &&     ln -s /usr/local/cuda-13.0/targets/sbsa-linux/lib/stubs/libcuda.so /tmp/cuda-stub/libcuda.so.1 &&     LD_LIBRARY_PATH=/tmp/cuda-stub:$LD_LIBRARY_PATH python3 -c "import vllm._C_stable_libtorch; import vllm._moe_C_stable_libtorch; import vllm, inspect; assert vllm.__version__.startswith('0.29.0'), vllm.__version__; from vllm import LLM, SamplingParams; assert SamplingParams().reasoning_eos_policy=='force_end'; import vllm.model_executor.models.qwen3_dflash as q; assert hasattr(q, 'dflash_has_any_non_causal'), 'dflash_has_any_non_causal missing'; import vllm.v1.spec_decode.dflash as d; assert 'dflash-blocktable-unpad' in inspect.getsource(d), 'blocktable slice lost'; import vllm.config.compilation as cc; assert 'AEON widened gate' in inspect.getsource(cc), 'cudagraph align widening lost'; _uma=any('uma-negative-cudagraph-estimate-clamp' in inspect.getsource(__import__(n, fromlist=['x'])) for n in ('vllm.v1.worker.gpu_model_runner','vllm.v1.worker.gpu.model_runner')); assert _uma, 'UMA clamp lost'; import vllm.envs as e; assert e.VLLM_USE_V2_MODEL_RUNNER is None, 'V2 runner pin unexpectedly baked'; import vllm.model_executor.models.qwen3_dspark as qds; assert 'quant_config' in inspect.signature(qds.DSparkMarkovHead.__init__).parameters, '#50424 quantized Markov heads missing'; import vllm.utils.torch_utils as tu; assert hasattr(tu, 'nvfp4_kv_cache_split_views'), 'nvfp4 split helper missing'; assert hasattr(tu, 'nvfp4_split_data_scale'), 'upstream-name nvfp4 alias missing'; assert 'as_strided' not in tu._nvfp4_split_data_scale.__code__.co_names, 'nvfp4 splitter reverted to as_strided (NHD-unsafe)'; import vllm.v1.attention.ops.triton_reshape_and_cache_flash as _trc; assert 'nvfp4' in inspect.getsource(_trc).lower(), 'NVFP4-KV path lost'; import vllm.compilation.cuda_graph as cg; assert 'thread_local' in inspect.getsource(cg), '#48053 thread_local lost (V1 graphs)'; import vllm.v1.worker.gpu.cudagraph_utils as cgu; assert 'thread_local' in inspect.getsource(cgu), '#48053 thread_local lost (MRv2 graphs)'; import vllm.model_executor.layers.quantization.modelopt as mo; src=inspect.getsource(mo); assert 'modelopt_mixed' in src or 'MIXED_PRECISION' in src; import vllm.multimodal.video as vid; print('vllm', vllm.__version__, '+aeon 0.29 hard-keeps OK')" &&     rm -rf /tmp/cuda-stub # buildkit
                        
# 2026-09-11 19:31:29  0.00B 设置工作目录为/
WORKDIR /
                        
# 2026-09-11 19:31:29  2.40MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c python3 /tmp/verify.py && rm /tmp/verify.py # buildkit
                        
# 2026-09-11 19:31:24  1.31KB 复制新文件或目录到容器中
COPY verify.py /tmp/verify.py # buildkit
                        
# 2026-09-11 19:31:24  15.05MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c python3 -c "from vllm.sampling_params import SamplingParams, validate_reasoning_eos_policy; assert validate_reasoning_eos_policy(None)=='force_end'; assert SamplingParams().reasoning_eos_policy=='force_end'; print('reasoning_eos smoke OK')" # buildkit
                        
# 2026-09-11 19:31:19  55.73KB 复制新文件或目录到容器中
COPY sampling_params.aeon029.py /usr/local/lib/python3.12/site-packages/vllm/sampling_params.py # buildkit
                        
# 2026-09-11 19:31:19  27.13KB 复制新文件或目录到容器中
COPY hotfixes/completion_protocol.py /usr/local/lib/python3.12/site-packages/vllm/entrypoints/openai/completion/protocol.py # buildkit
                        
# 2026-09-11 19:31:19  48.04KB 复制新文件或目录到容器中
COPY hotfixes/chat_protocol.py /usr/local/lib/python3.12/site-packages/vllm/entrypoints/openai/chat_completion/protocol.py # buildkit
                        
# 2026-09-11 19:31:19  24.58KB 复制新文件或目录到容器中
COPY hotfixes/thinking_budget_state.py /usr/local/lib/python3.12/site-packages/vllm/v1/sample/thinking_budget_state.py # buildkit
                        
# 2026-09-11 19:31:19  23.65KB 复制新文件或目录到容器中
COPY hotfixes/input_processor.py /usr/local/lib/python3.12/site-packages/vllm/v1/engine/input_processor.py # buildkit
                        
# 2026-09-11 19:31:19  0.00B 设置工作目录为/
WORKDIR /
                        
# 2026-09-11 19:31:19  10.82KB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir /tmp/humming-stub && rm -rf /tmp/humming-stub &&     python3 -c "from humming.dtypes import DataType; print('humming-stub: importable')" # buildkit
                        
# 2026-09-11 19:31:17  10.04KB 复制新文件或目录到容器中
COPY humming-stub/ /tmp/humming-stub/ # buildkit
                        
# 2026-09-11 19:17:57  100.16MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir --upgrade "transformers==5.14.1" 2>&1 | tail -3 &&     python3 -c "import transformers; print('transformers:', transformers.__version__)" # buildkit
                        
# 2026-09-11 19:17:50  148.83MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir "scipy>=1.11" 2>&1 | tail -3 &&     pip install --no-cache-dir --no-deps       "turboquant @ git+https://github.com/AEON-7/turboquant.git@fix/cuda-graph-safe-qjl-powers"       2>&1 | tail -3 ||     echo "[WARN] turboquant install attempted; check logs above if needed" # buildkit
                        
# 2026-09-11 19:17:42  9.25MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir "xgrammar>=0.2.1,<1.0.0" 2>&1 | tail -2 &&     python3 -c "from xgrammar import normalize_tool_choice; print('xgrammar: normalize_tool_choice OK')" # buildkit
                        
# 2026-09-11 19:17:38  20.28MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir "torchcodec==0.16.0" 2>&1 | tail -2 &&     python3 -c "import torchcodec; from torchcodec.decoders import VideoDecoder; print('torchcodec', torchcodec.__version__, 'loads')" ||     (pip uninstall -y torchcodec 2>&1 | tail -1 && echo "[WARN] torchcodec 0.16.0 broken on this stack — removed; video decode degrades gracefully") # buildkit
                        
# 2026-09-11 19:17:34  8.54GB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip uninstall -y flashinfer-jit-cache flashinfer-cubin 2>&1 | tail -2 || true;     pip install --no-cache-dir         --extra-index-url https://flashinfer.ai/whl/         --extra-index-url https://flashinfer.ai/whl/cu130/         "flashinfer-python==0.6.18" "flashinfer-cubin==0.6.18"         "flashinfer-jit-cache==0.6.18" 2>&1 | tail -4 &&     python3 -c "import flashinfer, importlib.metadata as m;       v=flashinfer.__version__; assert v.startswith('0.6.18'), v;       c=m.version('flashinfer-cubin'); assert c.startswith('0.6.18'), c;       j=m.version('flashinfer-jit-cache'); assert j.startswith('0.6.18'), j;       print('flashinfer trio matched:', v, '| cubin', c, '| jit-cache', j)" # buildkit
                        
# 2026-09-11 19:08:53  894.56MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir "nvidia-cutlass-dsl[cu13]==4.6.2" "quack-kernels==0.6.4"         "apache-tvm-ffi==0.1.11" "tilelang==0.1.12" 2>&1 | tail -3 &&     python3 -c "import importlib.metadata as m;       c=m.version('nvidia-cutlass-dsl'); q=m.version('quack-kernels'); t=m.version('apache-tvm-ffi');       print('cutlass-dsl:', c, '| quack:', q, '| tvm-ffi:', t, '| tilelang:', m.version('tilelang'));       assert c.startswith('4.6'), c; assert q=='0.6.4', q; assert t=='0.1.11', t;       import inspect as _i; from tvm_ffi.utils.kwargs_wrapper import make_kwargs_wrapper as _mkw;       assert 'map_dataclass_to_tuple' in _i.signature(_mkw).parameters, 'tvm-ffi too old for cutlass 4.6 cute.compile';       from quack import layout_utils; print('quack/cutlass/tvm-ffi ABI compatible')" # buildkit
                        
# 2026-09-11 19:07:21  11.87GB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c CUDA_HOME=/usr/local/cuda-13.0 PATH=/usr/local/cuda-13.0/bin:$PATH     pip install --no-deps --no-build-isolation -v . 2>&1 | tee /tmp/vllm-install.log | tail -120 &&     echo "[vllm pip install] exit=$?" # buildkit
                        
# 2026-09-11 18:18:14  74.29MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir "cmake>=3.26.1" "setuptools-rust>=1.9.0" wheel 2>&1 | tail -2 &&     python3 use_existing_torch.py 2>&1 | tail -5 # buildkit
                        
# 2026-09-11 18:18:05  0.00B 设置工作目录为/build/vllm-src
WORKDIR /build/vllm-src
                        
# 2026-09-11 18:18:05  0.00B 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip uninstall -y vllm 2>&1 | tail -3 # buildkit
                        
# 2026-09-11 18:18:03  109.59MB 复制新文件或目录到容器中
COPY vllm-src/ /build/vllm-src/ # buildkit
                        
# 2026-09-11 18:18:02  254.90MB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir "nvidia-nccl-cu13==2.30.7" 2>&1 | tail -2 &&     python3 -c "import importlib.metadata as m; v=m.version('nvidia-nccl-cu13'); assert v=='2.30.7', v; print('nccl', v)" # buildkit
                        
# 2026-09-11 18:17:19  3.44GB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c pip install --no-cache-dir "torch==2.13.0" "torchvision==0.28.0"         --index-url https://download.pytorch.org/whl/cu130 2>&1 | tail -4 &&     python3 -c "import torch, triton, torchaudio;       assert torch.__version__.startswith('2.13.0'), torch.__version__;       assert torch.version.cuda.startswith('13.0'), torch.version.cuda;       assert triton.__version__.startswith('3.7'), triton.__version__;       print('torch', torch.__version__, '| triton', triton.__version__, '| torchaudio', torchaudio.__version__)" # buildkit
                        
# 2026-09-11 18:14:21  9.19KB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c update-alternatives --set cuda /usr/local/cuda-13.0 2>/dev/null || true;     export CUDA_HOME=/usr/local/cuda-13.0 PATH=/usr/local/cuda-13.0/bin:$PATH;     nvcc --version | sed -n 's/.*release \([0-9.]\+\).*/nvcc release \1/p' | head -1;     test "$(nvcc --version | sed -n 's/.*release \([0-9.]\+\).*/\1/p' | head -1)" = "13.0" # buildkit
                        
# 2026-09-11 18:14:21  0.00B 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c CUDA_LIB=/usr/local/cuda-13.0/targets/sbsa-linux/lib &&     if [ ! -f $CUDA_LIB/libnvrtc.so ] && ls $CUDA_LIB/libnvrtc.so.13.0.* >/dev/null 2>&1; then         ln -sf $(ls $CUDA_LIB/libnvrtc.so.13.0.* | head -1) $CUDA_LIB/libnvrtc.so;         echo "Created libnvrtc.so symlink";     fi &&     ls -la $CUDA_LIB/libnvrtc.so* | head -5 # buildkit
                        
# 2026-09-11 18:14:21  0.00B 设置环境变量 CUDAHOSTCXX
ENV CUDAHOSTCXX=g++-12
                        
# 2026-09-11 18:14:21  0.00B 设置环境变量 CXX
ENV CXX=g++-12
                        
# 2026-09-11 18:14:21  0.00B 设置环境变量 CC
ENV CC=gcc-12
                        
# 2026-09-11 18:14:21  2.56GB 执行命令并创建新的镜像层
RUN /bin/bash -o pipefail -c apt-get update &&     apt-get install -y --no-install-recommends         git ca-certificates ffmpeg         gcc-12 g++-12         cuda-nvrtc-dev-13-0         libcusparse-dev-13-0         libcublas-dev-13-0         libcusolver-dev-13-0         libcufft-dev-13-0         libcurand-dev-13-0         libnvjitlink-dev-13-0 &&     apt-get clean && rm -rf /var/lib/apt/lists/* # buildkit
                        
# 2026-09-11 18:12:07  0.00B 设置工作目录为/build
WORKDIR /build
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 VLLM_ALLREDUCE_USE_FLASHINFER
ENV VLLM_ALLREDUCE_USE_FLASHINFER=0
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 TRITON_PTXAS_PATH
ENV TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 SETUPTOOLS_SCM_PRETEND_VERSION_FOR_VLLM
ENV SETUPTOOLS_SCM_PRETEND_VERSION_FOR_VLLM=0.29.0+aeon.sm121a.dspark
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 SETUPTOOLS_SCM_PRETEND_VERSION
ENV SETUPTOOLS_SCM_PRETEND_VERSION=0.29.0+aeon.sm121a.dspark
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 NVCC_THREADS
ENV NVCC_THREADS=2
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 MAX_JOBS
ENV MAX_JOBS=12
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 CMAKE_BUILD_PARALLEL_LEVEL
ENV CMAKE_BUILD_PARALLEL_LEVEL=8
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 PIP_RETRIES
ENV PIP_RETRIES=5
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 PIP_DEFAULT_TIMEOUT
ENV PIP_DEFAULT_TIMEOUT=180
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 CCACHE_DISABLE
ENV CCACHE_DISABLE=1
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 ENABLE_NVFP4_SM100
ENV ENABLE_NVFP4_SM100=0
                        
# 2026-09-11 18:12:07  0.00B 设置环境变量 TORCH_CUDA_ARCH_LIST
ENV TORCH_CUDA_ARCH_LIST=12.1a
                        
# 2026-09-11 18:12:07  0.00B 
SHELL [/bin/bash -o pipefail -c]
                        
# 2026-05-07 03:21:50  0.00B 添加元数据标签
LABEL org.opencontainers.image.title=AEON-Gemma-4-26B-A4B-DFLASH-vLLM0201 org.opencontainers.image.description=Official vLLM 0.20.1 runtime with AEON DFlash Python overlay for Gemma 4 26B A4B NVFP4 on DGX Spark/GB10. org.opencontainers.image.vendor=AEON-7 org.opencontainers.image.url=https://huggingface.co/AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4 org.opencontainers.image.documentation=https://github.com/AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4 org.opencontainers.image.source=https://github.com/AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4 org.opencontainers.image.created=2026-05-06T19:21:45Z org.opencontainers.image.revision=06e292d0ce7e0ddc4f84bd200c3bdf55c7875eb7 ai.aeon.model=AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4 ai.aeon.drafter=z-lab/gemma-4-26B-A4B-it-DFlash ai.aeon.hardware=NVIDIA DGX Spark GB10 SM121 ai.aeon.vllm_base=vllm/vllm-openai:latest vLLM 0.20.1 ai.aeon.features=gemma4,dflash,nvfp4,flashinfer-cutlass,vllm-cutlass,tool-calling
                        
# 2026-05-07 03:21:50  0.00B 设置默认要执行的命令
CMD []
                        
# 2026-05-07 03:21:50  0.00B 配置容器启动时运行的命令
ENTRYPOINT []
                        
# 2026-05-07 03:21:50  0.00B 设置环境变量 VLLM_USE_FLASHINFER_SAMPLER
ENV VLLM_USE_FLASHINFER_SAMPLER=1
                        
# 2026-05-07 03:21:50  27.23MB 执行命令并创建新的镜像层
RUN |3 VLLM_DFLASH_PATCH=vllm0201-dflash-python.patch BUILD_DATE=2026-05-06T19:21:45Z VLLM_PATCH_COMMIT=06e292d0ce7e0ddc4f84bd200c3bdf55c7875eb7 /bin/sh -c set -eux;     command -v patch >/dev/null || (apt-get update && apt-get install -y --no-install-recommends patch && rm -rf /var/lib/apt/lists/*);     site_dir="$(python3 -c 'import pathlib, vllm; print(pathlib.Path(vllm.__file__).resolve().parents[1])')";     cd "$site_dir";     patch --dry-run -p1 < /tmp/vllm0201-dflash-python.patch;     patch -p1 < /tmp/vllm0201-dflash-python.patch;     python3 -m py_compile       vllm/config/attention.py       vllm/config/speculative.py       vllm/v1/attention/selector.py       vllm/v1/spec_decode/dflash.py       vllm/v1/spec_decode/eagle.py       vllm/v1/spec_decode/utils.py       vllm/v1/worker/gpu_model_runner.py;     rm /tmp/vllm0201-dflash-python.patch # buildkit
                        
# 2026-05-07 02:11:55  29.54KB 复制新文件或目录到容器中
COPY vllm0201-dflash-python.patch /tmp/vllm0201-dflash-python.patch # buildkit
                        
# 2026-05-07 02:11:55  0.00B 定义构建参数
ARG VLLM_PATCH_COMMIT=unknown
                        
# 2026-05-07 02:11:55  0.00B 定义构建参数
ARG BUILD_DATE=unknown
                        
# 2026-05-07 02:11:55  0.00B 定义构建参数
ARG VLLM_DFLASH_PATCH=vllm0201-dflash-python.patch
                        
# 2026-05-04 17:25:24  0.00B 配置容器启动时运行的命令
ENTRYPOINT ["vllm" "serve"]
                        
# 2026-05-04 17:25:24  0.00B 设置环境变量 VLLM_USAGE_SOURCE
ENV VLLM_USAGE_SOURCE=production-docker-image
                        
# 2026-05-04 17:25:24  1.71GB 执行命令并创建新的镜像层
RUN |8 TARGETPLATFORM=linux/arm64 INSTALL_KV_CONNECTORS=true CUDA_VERSION=13.0.2 PIP_INDEX_URL= UV_INDEX_URL= PIP_EXTRA_INDEX_URL= UV_EXTRA_INDEX_URL= torch_cuda_arch_list=8.0 8.7 8.9 9.0 10.0 11.0 12.0+PTX /bin/sh -c CUDA_MAJOR="${CUDA_VERSION%%.*}";     CUDA_VERSION_DASH=$(echo $CUDA_VERSION | cut -d. -f1,2 | tr '.' '-');     CUDA_HOME=/usr/local/cuda;     BUILD_PKGS="libcusparse-dev-${CUDA_VERSION_DASH}                 libcublas-dev-${CUDA_VERSION_DASH}                 libcusolver-dev-${CUDA_VERSION_DASH}";     if [ "$INSTALL_KV_CONNECTORS" = "true" ]; then         if [ "$CUDA_MAJOR" -ge 13 ]; then             uv pip install --system nixl-cu13;         fi;         uv pip install --system -r /tmp/kv_connectors.txt --no-build || (             apt-get update -y &&             apt-get install -y --no-install-recommends --allow-change-held-packages ${BUILD_PKGS} &&             uv pip install --system -r /tmp/kv_connectors.txt --no-build-isolation &&             apt-get purge -y ${BUILD_PKGS} &&             rm -rf /var/lib/apt/lists/*         );     fi # buildkit
                        
# 2026-05-04 17:21:32  0.00B 设置环境变量 TORCH_CUDA_ARCH_LIST
ENV TORCH_CUDA_ARCH_LIST=8.0 8.7 8.9 9.0 10.0 11.0 12.0+PTX
                        
# 2026-05-04 17:21:32  0.00B 定义构建参数
ARG torch_cuda_arch_list=7.5 8.0 8.6 8.9 9.0 10.0 11.0 12.0+PTX
                        
# 2026-05-04 17:21:32  0.00B 设置环境变量 UV_HTTP_TIMEOUT
ENV UV_HTTP_TIMEOUT=500
                        
# 2026-05-04 17:21:32  0.00B 定义构建参数
ARG PIP_EXTRA_INDEX_URL UV_EXTRA_INDEX_URL
                        
# 2026-05-04 17:21:32  0.00B 定义构建参数
ARG PIP_INDEX_URL UV_INDEX_URL
                        
# 2026-05-04 17:21:32  0.00B 定义构建参数
ARG CUDA_VERSION
                        
# 2026-05-04 17:21:32  0.00B 定义构建参数
ARG INSTALL_KV_CONNECTORS=false
                        
# 2026-05-04 17:21:32  0.00B 定义构建参数
ARG TARGETPLATFORM
                        
# 2026-05-04 17:21:32  35.09KB 复制新文件或目录到容器中
COPY ./vllm/collect_env.py . # buildkit
                        
# 2026-05-04 17:21:32  986.54KB 复制新文件或目录到容器中
COPY benchmarks benchmarks # buildkit
                        
# 2026-05-04 17:21:32  1.25MB 复制新文件或目录到容器中
COPY examples examples # buildkit
                        
# 2026-05-04 17:21:32  0.00B 设置环境变量 LD_LIBRARY_PATH
ENV LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64
                        
# 2026-05-04 17:21:32  29.58MB 执行命令并创建新的镜像层
RUN |22 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl FLASHINFER_VERSION=0.6.8.post1 GDRCOPY_CUDA_VERSION=12.8 GDRCOPY_OS_VERSION=Ubuntu22_04 TARGETPLATFORM=linux/arm64 BITSANDBYTES_VERSION_X86=0.46.1 BITSANDBYTES_VERSION_ARM64=0.42.0 TIMM_VERSION=>=1.0.17 RUNAI_MODEL_STREAMER_VERSION=>=0.15.7 PIP_INDEX_URL= UV_INDEX_URL= PIP_EXTRA_INDEX_URL= UV_EXTRA_INDEX_URL= PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl PIP_KEYRING_PROVIDER=disabled UV_KEYRING_PROVIDER=disabled PYTORCH_NIGHTLY= /bin/sh -c uv pip install --system ep_kernels/dist/*.whl --verbose         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.') # buildkit
                        
# 2026-05-04 17:21:31  0.00B 设置环境变量 LD_LIBRARY_PATH
ENV LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64
                        
# 2026-05-04 17:21:31  0.00B 执行命令并创建新的镜像层
RUN |22 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl FLASHINFER_VERSION=0.6.8.post1 GDRCOPY_CUDA_VERSION=12.8 GDRCOPY_OS_VERSION=Ubuntu22_04 TARGETPLATFORM=linux/arm64 BITSANDBYTES_VERSION_X86=0.46.1 BITSANDBYTES_VERSION_ARM64=0.42.0 TIMM_VERSION=>=1.0.17 RUNAI_MODEL_STREAMER_VERSION=>=0.15.7 PIP_INDEX_URL= UV_INDEX_URL= PIP_EXTRA_INDEX_URL= UV_EXTRA_INDEX_URL= PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl PIP_KEYRING_PROVIDER=disabled UV_KEYRING_PROVIDER=disabled PYTORCH_NIGHTLY= /bin/sh -c . /etc/environment && uv pip list # buildkit
                        
# 2026-05-04 17:21:31  797.31MB 执行命令并创建新的镜像层
RUN |22 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl FLASHINFER_VERSION=0.6.8.post1 GDRCOPY_CUDA_VERSION=12.8 GDRCOPY_OS_VERSION=Ubuntu22_04 TARGETPLATFORM=linux/arm64 BITSANDBYTES_VERSION_X86=0.46.1 BITSANDBYTES_VERSION_ARM64=0.42.0 TIMM_VERSION=>=1.0.17 RUNAI_MODEL_STREAMER_VERSION=>=0.15.7 PIP_INDEX_URL= UV_INDEX_URL= PIP_EXTRA_INDEX_URL= UV_EXTRA_INDEX_URL= PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl PIP_KEYRING_PROVIDER=disabled UV_KEYRING_PROVIDER=disabled PYTORCH_NIGHTLY= /bin/sh -c if [ "${PYTORCH_NIGHTLY}" = "1" ]; then         echo "Installing torch nightly..."         && uv pip install --system $(cat torch_lib_versions.txt | xargs) --pre         --index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/nightly/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')         && echo "Installing vLLM..."         && uv pip install --system dist/*.whl --verbose         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/nightly/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.');     else         echo "Installing vLLM..."         && uv pip install --system dist/*.whl --verbose         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.');     fi # buildkit
                        
# 2026-05-04 17:12:54  71.00B 复制新文件或目录到容器中
COPY /workspace/torch_lib_versions.txt torch_lib_versions.txt # buildkit
                        
# 2026-05-04 17:12:54  0.00B 定义构建参数
ARG PYTORCH_NIGHTLY
                        
# 2026-05-04 17:12:54  0.00B 定义构建参数
ARG PIP_KEYRING_PROVIDER UV_KEYRING_PROVIDER
                        
# 2026-05-04 17:12:54  0.00B 定义构建参数
ARG PYTORCH_CUDA_INDEX_BASE_URL
                        
# 2026-05-04 17:12:54  0.00B 定义构建参数
ARG PIP_EXTRA_INDEX_URL UV_EXTRA_INDEX_URL
                        
# 2026-05-04 17:12:54  0.00B 定义构建参数
ARG PIP_INDEX_URL UV_INDEX_URL
                        
# 2026-05-04 17:12:54  576.01MB 执行命令并创建新的镜像层
RUN |14 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl FLASHINFER_VERSION=0.6.8.post1 GDRCOPY_CUDA_VERSION=12.8 GDRCOPY_OS_VERSION=Ubuntu22_04 TARGETPLATFORM=linux/arm64 BITSANDBYTES_VERSION_X86=0.46.1 BITSANDBYTES_VERSION_ARM64=0.42.0 TIMM_VERSION=>=1.0.17 RUNAI_MODEL_STREAMER_VERSION=>=0.15.7 /bin/sh -c if [ "$TARGETPLATFORM" = "linux/arm64" ]; then         BITSANDBYTES_VERSION="${BITSANDBYTES_VERSION_ARM64}";     else         BITSANDBYTES_VERSION="${BITSANDBYTES_VERSION_X86}";     fi;     uv pip install --system accelerate modelscope         "bitsandbytes>=${BITSANDBYTES_VERSION}" "timm${TIMM_VERSION}" "runai-model-streamer[s3,gcs,azure]${RUNAI_MODEL_STREAMER_VERSION}" # buildkit
                        
# 2026-05-04 17:12:50  0.00B 定义构建参数
ARG RUNAI_MODEL_STREAMER_VERSION=>=0.15.7
                        
# 2026-05-04 17:12:50  0.00B 定义构建参数
ARG TIMM_VERSION=>=1.0.17
                        
# 2026-05-04 17:12:50  0.00B 定义构建参数
ARG BITSANDBYTES_VERSION_ARM64=0.42.0
                        
# 2026-05-04 17:12:50  0.00B 定义构建参数
ARG BITSANDBYTES_VERSION_X86=0.46.1
                        
# 2026-05-04 17:12:50  2.30MB 执行命令并创建新的镜像层
RUN |10 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl FLASHINFER_VERSION=0.6.8.post1 GDRCOPY_CUDA_VERSION=12.8 GDRCOPY_OS_VERSION=Ubuntu22_04 TARGETPLATFORM=linux/arm64 /bin/sh -c set -eux;     case "${TARGETPLATFORM}" in       linux/arm64) UUARCH="aarch64" ;;       linux/amd64) UUARCH="x64" ;;       *) echo "Unsupported TARGETPLATFORM: ${TARGETPLATFORM}" >&2; exit 1 ;;     esac;     /tmp/install_gdrcopy.sh "${GDRCOPY_OS_VERSION}" "${GDRCOPY_CUDA_VERSION}" "${UUARCH}" &&     rm /tmp/install_gdrcopy.sh # buildkit
                        
# 2026-05-04 17:12:45  1.44KB 复制新文件或目录到容器中
COPY tools/install_gdrcopy.sh /tmp/install_gdrcopy.sh # buildkit
                        
# 2026-05-04 17:12:45  0.00B 定义构建参数
ARG TARGETPLATFORM
                        
# 2026-05-04 17:12:45  0.00B 定义构建参数
ARG GDRCOPY_OS_VERSION=Ubuntu22_04
                        
# 2026-05-04 17:12:45  0.00B 定义构建参数
ARG GDRCOPY_CUDA_VERSION=12.8
                        
# 2026-05-04 17:12:45  10.58GB 执行命令并创建新的镜像层
RUN |7 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl FLASHINFER_VERSION=0.6.8.post1 /bin/sh -c uv pip install --system flashinfer-jit-cache==${FLASHINFER_VERSION}         --extra-index-url https://flashinfer.ai/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.')     && flashinfer show-config     && flashinfer download-cubin # buildkit
                        
# 2026-05-04 17:06:53  0.00B 定义构建参数
ARG FLASHINFER_VERSION=0.6.8.post1
                        
# 2026-05-04 17:06:53  7.01GB 执行命令并创建新的镜像层
RUN |6 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py PYTORCH_CUDA_INDEX_BASE_URL=https://download.pytorch.org/whl /bin/sh -c uv pip install --system -r /tmp/requirements-cuda.txt         --extra-index-url ${PYTORCH_CUDA_INDEX_BASE_URL}/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d '.') &&     rm /tmp/requirements-cuda.txt /tmp/common.txt # buildkit
                        
# 2026-05-04 17:06:44  821.00B 复制新文件或目录到容器中
COPY requirements/cuda.txt /tmp/requirements-cuda.txt # buildkit
                        
# 2026-05-04 17:06:44  2.94KB 复制新文件或目录到容器中
COPY requirements/common.txt /tmp/common.txt # buildkit
                        
# 2026-05-04 17:06:44  0.00B 定义构建参数
ARG PYTORCH_CUDA_INDEX_BASE_URL
                        
# 2026-05-04 17:06:44  0.00B 设置环境变量 VLLM_ENABLE_CUDA_COMPATIBILITY
ENV VLLM_ENABLE_CUDA_COMPATIBILITY=0
                        
# 2026-05-04 17:06:44  0.00B 设置环境变量 UV_LINK_MODE
ENV UV_LINK_MODE=copy
                        
# 2026-05-04 17:06:44  0.00B 设置环境变量 UV_INDEX_STRATEGY
ENV UV_INDEX_STRATEGY=unsafe-best-match
                        
# 2026-05-04 17:06:44  0.00B 设置环境变量 UV_HTTP_TIMEOUT
ENV UV_HTTP_TIMEOUT=500
                        
# 2026-05-04 17:06:44  76.45MB 执行命令并创建新的镜像层
RUN |5 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py /bin/sh -c python3 -m pip install uv # buildkit
                        
# 2026-05-04 17:06:42  1.82GB 执行命令并创建新的镜像层
RUN |5 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py /bin/sh -c CUDA_VERSION_DASH=$(echo $CUDA_VERSION | cut -d. -f1,2 | tr '.' '-') &&     CUDA_VERSION_SHORT=$(echo $CUDA_VERSION | cut -d. -f1,2) &&     apt-get update -y &&     apt-get install -y --no-install-recommends --allow-change-held-packages         cuda-nvcc-${CUDA_VERSION_DASH}         cuda-cudart-${CUDA_VERSION_DASH}         cuda-nvrtc-${CUDA_VERSION_DASH}         cuda-cuobjdump-${CUDA_VERSION_DASH}         libcurand-dev-${CUDA_VERSION_DASH}         libcublas-${CUDA_VERSION_DASH}         libnuma-dev         numactl &&     NCCL_VER=$(apt-cache madison libnccl-dev | grep "+cuda${CUDA_VERSION_SHORT}" | head -1 | awk -F'|' '{gsub(/^ +| +$/, "", $2); print $2}') &&     apt-get install -y --no-install-recommends --allow-change-held-packages libnccl-dev=${NCCL_VER} libnccl2=${NCCL_VER} &&     rm -rf /var/lib/apt/lists/* # buildkit
                        
# 2026-05-04 17:06:02  1.01GB 执行命令并创建新的镜像层
RUN |5 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py /bin/sh -c apt-get update -y     && apt-get install -y --no-install-recommends         curl         sudo         ffmpeg         libsm6         libxext6         libgl1         libibverbs-dev         build-essential         libssl-dev         libffi-dev         zlib1g-dev         libbz2-dev         libreadline-dev         libsqlite3-dev         libncurses-dev         liblzma-dev         libgdbm-dev         uuid-dev         tk-dev     && PYTHON_MAJOR_MINOR=${PYTHON_VERSION}     && PYTHON_FULL_VERSION=$(curl -s https://www.python.org/ftp/python/         | grep -oE "${PYTHON_MAJOR_MINOR}\.[0-9]+"         | sort -t. -k3 -n | uniq         | tail -1)     && echo "Building Python ${PYTHON_FULL_VERSION} from source..."     && curl -fSL https://www.python.org/ftp/python/${PYTHON_FULL_VERSION}/Python-${PYTHON_FULL_VERSION}.tgz -o /tmp/python.tgz     && tar -xzf /tmp/python.tgz -C /tmp     && cd /tmp/Python-${PYTHON_FULL_VERSION}     && ./configure --enable-optimizations --with-ensurepip=install --prefix=/usr/local     && make -j$(nproc)     && make install     && cd /     && rm -rf /tmp/python.tgz /tmp/Python-${PYTHON_FULL_VERSION}     && ln -sf /usr/local/bin/python${PYTHON_MAJOR_MINOR} /usr/bin/python3     && ln -sf /usr/local/bin/python${PYTHON_MAJOR_MINOR}-config /usr/bin/python3-config     && ln -sf /usr/local/bin/pip${PYTHON_MAJOR_MINOR} /usr/bin/pip     && rm -rf /var/lib/apt/lists/*     && python3 --version && python3 -m pip --version # buildkit
                        
# 2026-05-04 17:02:58  136.00B 执行命令并创建新的镜像层
RUN |5 CUDA_VERSION=13.0.2 PYTHON_VERSION=3.12 DEADSNAKES_MIRROR_URL= DEADSNAKES_GPGKEY_URL= GET_PIP_URL=https://bootstrap.pypa.io/get-pip.py /bin/sh -c PYTHON_VERSION_STR=$(echo ${PYTHON_VERSION} | sed 's/\.//g') &&     echo "export PYTHON_VERSION_STR=${PYTHON_VERSION_STR}" >> /etc/environment # buildkit
                        
# 2026-05-04 17:02:57  0.00B 设置工作目录为/vllm-workspace
WORKDIR /vllm-workspace
                        
# 2026-05-04 17:02:57  0.00B 设置环境变量 DEBIAN_FRONTEND
ENV DEBIAN_FRONTEND=noninteractive
                        
# 2026-05-04 17:02:57  0.00B 定义构建参数
ARG GET_PIP_URL
                        
# 2026-05-04 17:02:57  0.00B 定义构建参数
ARG DEADSNAKES_GPGKEY_URL
                        
# 2026-05-04 17:02:57  0.00B 定义构建参数
ARG DEADSNAKES_MIRROR_URL
                        
# 2026-05-04 17:02:57  0.00B 定义构建参数
ARG PYTHON_VERSION
                        
# 2026-05-04 17:02:57  0.00B 定义构建参数
ARG CUDA_VERSION
                        
# 2025-10-11 00:51:55  0.00B 设置环境变量 NVIDIA_DRIVER_CAPABILITIES
ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
                        
# 2025-10-11 00:51:55  0.00B 设置环境变量 NVIDIA_VISIBLE_DEVICES
ENV NVIDIA_VISIBLE_DEVICES=all
                        
# 2025-10-11 00:51:55  17.29KB 复制新文件或目录到容器中
COPY NGC-DL-CONTAINER-LICENSE / # buildkit
                        
# 2025-10-11 00:51:55  0.00B 设置环境变量 LD_LIBRARY_PATH
ENV LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64
                        
# 2025-10-11 00:51:55  0.00B 设置环境变量 PATH
ENV PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
                        
# 2025-10-11 00:51:55  22.00B 执行命令并创建新的镜像层
RUN |1 TARGETARCH=arm64 /bin/sh -c echo "/usr/local/cuda/lib64" >> /etc/ld.so.conf.d/nvidia.conf # buildkit
                        
# 2025-10-11 00:51:55  303.51MB 执行命令并创建新的镜像层
RUN |1 TARGETARCH=arm64 /bin/sh -c apt-get update && apt-get install -y --no-install-recommends     cuda-cudart-13-0=${NV_CUDA_CUDART_VERSION}     cuda-compat-13-0     && rm -rf /var/lib/apt/lists/* # buildkit
                        
# 2025-10-11 00:50:55  0.00B 设置环境变量 CUDA_VERSION
ENV CUDA_VERSION=13.0.2
                        
# 2025-10-11 00:50:55  10.21MB 执行命令并创建新的镜像层
RUN |1 TARGETARCH=arm64 /bin/sh -c apt-get update && apt-get install -y --no-install-recommends     gnupg2 curl ca-certificates &&     curl -fsSLO https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/${NVARCH}/cuda-keyring_1.1-1_all.deb &&     dpkg -i cuda-keyring_1.1-1_all.deb &&     apt-get purge --autoremove -y curl     && rm -rf /var/lib/apt/lists/* # buildkit
                        
# 2025-10-11 00:50:55  0.00B 添加元数据标签
LABEL maintainer=NVIDIA CORPORATION <cudatools@nvidia.com>
                        
# 2025-10-11 00:50:55  0.00B 定义构建参数
ARG TARGETARCH
                        
# 2025-10-11 00:50:55  0.00B 设置环境变量 NV_CUDA_CUDART_VERSION
ENV NV_CUDA_CUDART_VERSION=13.0.96-1
                        
# 2025-10-11 00:50:55  0.00B 设置环境变量 NVIDIA_REQUIRE_CUDA brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand brand
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 brand=nvidia,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=vapps,driver>=535,driver<536 brand=vpc,driver>=535,driver<536 brand=vcs,driver>=535,driver<536 brand=vws,driver>=535,driver<536 brand=cloudgaming,driver>=535,driver<536 brand=unknown,driver>=550,driver<551 brand=grid,driver>=550,driver<551 brand=tesla,driver>=550,driver<551 brand=nvidia,driver>=550,driver<551 brand=quadro,driver>=550,driver<551 brand=quadrortx,driver>=550,driver<551 brand=nvidiartx,driver>=550,driver<551 brand=vapps,driver>=550,driver<551 brand=vpc,driver>=550,driver<551 brand=vcs,driver>=550,driver<551 brand=vws,driver>=550,driver<551 brand=cloudgaming,driver>=550,driver<551 brand=unknown,driver>=565,driver<566 brand=grid,driver>=565,driver<566 brand=tesla,driver>=565,driver<566 brand=nvidia,driver>=565,driver<566 brand=quadro,driver>=565,driver<566 brand=quadrortx,driver>=565,driver<566 brand=nvidiartx,driver>=565,driver<566 brand=vapps,driver>=565,driver<566 brand=vpc,driver>=565,driver<566 brand=vcs,driver>=565,driver<566 brand=vws,driver>=565,driver<566 brand=cloudgaming,driver>=565,driver<566 brand=unknown,driver>=570,driver<571 brand=grid,driver>=570,driver<571 brand=tesla,driver>=570,driver<571 brand=nvidia,driver>=570,driver<571 brand=quadro,driver>=570,driver<571 brand=quadrortx,driver>=570,driver<571 brand=nvidiartx,driver>=570,driver<571 brand=vapps,driver>=570,driver<571 brand=vpc,driver>=570,driver<571 brand=vcs,driver>=570,driver<571 brand=vws,driver>=570,driver<571 brand=cloudgaming,driver>=570,driver<571 brand=unknown,driver>=575,driver<576 brand=grid,driver>=575,driver<576 brand=tesla,driver>=575,driver<576 brand=nvidia,driver>=575,driver<576 brand=quadro,driver>=575,driver<576 brand=quadrortx,driver>=575,driver<576 brand=nvidiartx,driver>=575,driver<576 brand=vapps,driver>=575,driver<576 brand=vpc,driver>=575,driver<576 brand=vcs,driver>=575,driver<576 brand=vws,driver>=575,driver<576 brand=cloudgaming,driver>=575,driver<576
                        
# 2025-10-11 00:50:55  0.00B 设置环境变量 NVARCH
ENV NVARCH=sbsa
                        
# 2025-10-01 15:16:12  0.00B 
/bin/sh -c #(nop)  CMD ["/bin/bash"]
                        
# 2025-10-01 15:16:12  69.35MB 
/bin/sh -c #(nop) ADD file:7a71c1d52054f8e04c815eaec639d14adaaa62346860f4003201834430b7ff18 in / 
                        
# 2025-10-01 15:16:10  0.00B 
/bin/sh -c #(nop)  LABEL org.opencontainers.image.version=22.04
                        
# 2025-10-01 15:16:10  0.00B 
/bin/sh -c #(nop)  LABEL org.opencontainers.image.ref.name=ubuntu
                        
# 2025-10-01 15:16:10  0.00B 
/bin/sh -c #(nop)  ARG LAUNCHPAD_BUILD_ARCH
                        
# 2025-10-01 15:16:10  0.00B 
/bin/sh -c #(nop)  ARG RELEASE
                        
                    

镜像信息

{
    "Id": "sha256:715cf21dee93ebc45f6f37978b355ed2bd52231d9cf9bd6d9e8dc13b55195461",
    "RepoTags": [
        "ghcr.io/aeon-7/aeon-vllm-ultimate:latest",
        "swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate:latest-linuxarm64"
    ],
    "RepoDigests": [
        "ghcr.io/aeon-7/aeon-vllm-ultimate@sha256:2421bb1228a85370c1c50adb31f605c4361acf4d48d65282fcb919e74f34fae7",
        "swr.cn-north-4.myhuaweicloud.com/ddn-k8s/ghcr.io/aeon-7/aeon-vllm-ultimate@sha256:3cb1a439732aa38c7194b87a0956d276f2007ddd17497fd215cdf00f9271e28b"
    ],
    "Parent": "",
    "Comment": "buildkit.dockerfile.v0",
    "Created": "2026-09-11T06:32:06.171995256-05:00",
    "Container": "",
    "ContainerConfig": null,
    "DockerVersion": "",
    "Author": "",
    "Config": {
        "Hostname": "",
        "Domainname": "",
        "User": "",
        "AttachStdin": false,
        "AttachStdout": false,
        "AttachStderr": false,
        "Tty": false,
        "OpenStdin": false,
        "StdinOnce": false,
        "Env": [
            "PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin",
            "NVARCH=sbsa",
            "NVIDIA_REQUIRE_CUDA=cuda\u003e=13.0 brand=unknown,driver\u003e=535,driver\u003c536 brand=grid,driver\u003e=535,driver\u003c536 brand=tesla,driver\u003e=535,driver\u003c536 brand=nvidia,driver\u003e=535,driver\u003c536 brand=quadro,driver\u003e=535,driver\u003c536 brand=quadrortx,driver\u003e=535,driver\u003c536 brand=nvidiartx,driver\u003e=535,driver\u003c536 brand=vapps,driver\u003e=535,driver\u003c536 brand=vpc,driver\u003e=535,driver\u003c536 brand=vcs,driver\u003e=535,driver\u003c536 brand=vws,driver\u003e=535,driver\u003c536 brand=cloudgaming,driver\u003e=535,driver\u003c536 brand=unknown,driver\u003e=550,driver\u003c551 brand=grid,driver\u003e=550,driver\u003c551 brand=tesla,driver\u003e=550,driver\u003c551 brand=nvidia,driver\u003e=550,driver\u003c551 brand=quadro,driver\u003e=550,driver\u003c551 brand=quadrortx,driver\u003e=550,driver\u003c551 brand=nvidiartx,driver\u003e=550,driver\u003c551 brand=vapps,driver\u003e=550,driver\u003c551 brand=vpc,driver\u003e=550,driver\u003c551 brand=vcs,driver\u003e=550,driver\u003c551 brand=vws,driver\u003e=550,driver\u003c551 brand=cloudgaming,driver\u003e=550,driver\u003c551 brand=unknown,driver\u003e=565,driver\u003c566 brand=grid,driver\u003e=565,driver\u003c566 brand=tesla,driver\u003e=565,driver\u003c566 brand=nvidia,driver\u003e=565,driver\u003c566 brand=quadro,driver\u003e=565,driver\u003c566 brand=quadrortx,driver\u003e=565,driver\u003c566 brand=nvidiartx,driver\u003e=565,driver\u003c566 brand=vapps,driver\u003e=565,driver\u003c566 brand=vpc,driver\u003e=565,driver\u003c566 brand=vcs,driver\u003e=565,driver\u003c566 brand=vws,driver\u003e=565,driver\u003c566 brand=cloudgaming,driver\u003e=565,driver\u003c566 brand=unknown,driver\u003e=570,driver\u003c571 brand=grid,driver\u003e=570,driver\u003c571 brand=tesla,driver\u003e=570,driver\u003c571 brand=nvidia,driver\u003e=570,driver\u003c571 brand=quadro,driver\u003e=570,driver\u003c571 brand=quadrortx,driver\u003e=570,driver\u003c571 brand=nvidiartx,driver\u003e=570,driver\u003c571 brand=vapps,driver\u003e=570,driver\u003c571 brand=vpc,driver\u003e=570,driver\u003c571 brand=vcs,driver\u003e=570,driver\u003c571 brand=vws,driver\u003e=570,driver\u003c571 brand=cloudgaming,driver\u003e=570,driver\u003c571 brand=unknown,driver\u003e=575,driver\u003c576 brand=grid,driver\u003e=575,driver\u003c576 brand=tesla,driver\u003e=575,driver\u003c576 brand=nvidia,driver\u003e=575,driver\u003c576 brand=quadro,driver\u003e=575,driver\u003c576 brand=quadrortx,driver\u003e=575,driver\u003c576 brand=nvidiartx,driver\u003e=575,driver\u003c576 brand=vapps,driver\u003e=575,driver\u003c576 brand=vpc,driver\u003e=575,driver\u003c576 brand=vcs,driver\u003e=575,driver\u003c576 brand=vws,driver\u003e=575,driver\u003c576 brand=cloudgaming,driver\u003e=575,driver\u003c576",
            "NV_CUDA_CUDART_VERSION=13.0.96-1",
            "CUDA_VERSION=13.0.2",
            "LD_LIBRARY_PATH=/usr/local/nvidia/lib64:/usr/local/cuda/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64",
            "NVIDIA_VISIBLE_DEVICES=all",
            "NVIDIA_DRIVER_CAPABILITIES=compute,utility",
            "DEBIAN_FRONTEND=noninteractive",
            "UV_HTTP_TIMEOUT=500",
            "UV_INDEX_STRATEGY=unsafe-best-match",
            "UV_LINK_MODE=copy",
            "VLLM_ENABLE_CUDA_COMPATIBILITY=0",
            "TORCH_CUDA_ARCH_LIST=12.1a",
            "VLLM_USAGE_SOURCE=production-docker-image",
            "VLLM_USE_FLASHINFER_SAMPLER=1",
            "ENABLE_NVFP4_SM100=0",
            "CCACHE_DISABLE=1",
            "PIP_DEFAULT_TIMEOUT=180",
            "PIP_RETRIES=5",
            "CMAKE_BUILD_PARALLEL_LEVEL=8",
            "MAX_JOBS=12",
            "NVCC_THREADS=2",
            "SETUPTOOLS_SCM_PRETEND_VERSION=0.29.0+aeon.sm121a.dspark",
            "SETUPTOOLS_SCM_PRETEND_VERSION_FOR_VLLM=0.29.0+aeon.sm121a.dspark",
            "TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas",
            "VLLM_ALLREDUCE_USE_FLASHINFER=0",
            "CC=gcc-12",
            "CXX=g++-12",
            "CUDAHOSTCXX=g++-12"
        ],
        "Cmd": null,
        "ArgsEscaped": true,
        "Image": "",
        "Volumes": null,
        "WorkingDir": "/",
        "Entrypoint": [
            "/bin/bash"
        ],
        "OnBuild": null,
        "Labels": {
            "ai.aeon.drafter": "z-lab/gemma-4-26B-A4B-it-DFlash",
            "ai.aeon.features": "gemma4,qwen3.6,dflash,dflash-swa-mrv2,dspark,dspark-markov-heads,dspark-markov-quantized,dspark-topk,nvfp4,nvfp4-awq,nvfp4-kv,nvfp4-kv-nhd-safe,fp8-kv,spec-kv-dtype,hybrid-apc-fix,packed-kv-zeroer-fix,gdn-gate-align,flashinfer-0.6.18,torch-2.13.0,triton-3.7.1,cutlass-dsl-4.6.0,quack-0.6.1,tvm-ffi-0.1.11,tilelang-0.1.12,nccl-2.30.7,torchcodec-0.16.0,uma-clamp,thread-local-capture,tp2-ready,turboquant,tool-calling,mrv2-default-routing",
            "ai.aeon.hardware": "NVIDIA DGX Spark GB10 SM121",
            "ai.aeon.model": "fleet: Gemma-4-26B-A4B, Gemma-4-31B-DECKARD (NVFP4_AWQ), Qwen3.6-27B, Qwen3.6-35B-A3B",
            "ai.aeon.vllm_base": "vLLM 0.29.0 (from-source, sm_121a 3-way merge + 8 cherry-picks)",
            "maintainer": "NVIDIA CORPORATION \u003ccudatools@nvidia.com\u003e",
            "org.opencontainers.image.created": "2026-05-06T19:21:45Z",
            "org.opencontainers.image.description": "AEON vLLM Ultimate — vLLM 0.29.0 built from source for DGX Spark / Blackwell (sm_121a/GB10). DSpark Markov heads (quantized, #50424) + top-k projection (#49969), DFlash SWA on MRv2 (upstream #47914/#48113) with V1 carries as fallback, Triton NVFP4-KV (NHD-safe slicing), NVFP4_AWQ, packed-KV zeroer fix (#50276), GDN spec-token gate fix (#51812), hybrid-APC fix (#51843), DSpark bugfix set, thread_local capture at all graph sites (TP=2 2x-Spark ready, eager-first cross-node), torch 2.13.0 + Triton 3.7.1 + FlashInfer 0.6.18 trio + NCCL 2.30.7.",
            "org.opencontainers.image.documentation": "https://github.com/AEON-7/vllm-ultimate-dgx-spark",
            "org.opencontainers.image.licenses": "Apache-2.0",
            "org.opencontainers.image.ref.name": "ubuntu",
            "org.opencontainers.image.revision": "06e292d0ce7e0ddc4f84bd200c3bdf55c7875eb7",
            "org.opencontainers.image.source": "https://github.com/AEON-7/vllm-ultimate-dgx-spark",
            "org.opencontainers.image.title": "AEON-Gemma-4-26B-A4B-DFLASH-vLLM0201",
            "org.opencontainers.image.url": "https://huggingface.co/AEON-7/Gemma-4-26B-A4B-it-Uncensored-NVFP4",
            "org.opencontainers.image.vendor": "AEON-7",
            "org.opencontainers.image.version": "22.04"
        },
        "Shell": [
            "/bin/bash",
            "-o",
            "pipefail",
            "-c"
        ]
    },
    "Architecture": "arm64",
    "Os": "linux",
    "Size": 52296026666,
    "GraphDriver": {
        "Data": {
            "LowerDir": "/var/lib/docker/overlay2/42eb197bf8e3e0311545eaf1735f69afe9ab8e19515e09fa8f32b8ef695a3bc1/diff:/var/lib/docker/overlay2/3befd1898c379a0dc94a57f63be66fd30bd9fdf71cbba3c3fd95fca958c3f1d1/diff:/var/lib/docker/overlay2/ad2fdd375e1c87abc7cc903b3a6094d42b8abeead9e7df2486a028d3c30b4c58/diff:/var/lib/docker/overlay2/e71ab246448ef88cdba19ee6efce03da98ebc532748fb7bbdf6443aa62e62996/diff:/var/lib/docker/overlay2/9b17b4729b94896d10f6458688b2d551395f0f8e8ebe6de2ce9a1848cf1e339e/diff:/var/lib/docker/overlay2/3bf87b99200dd37a2ba030d8c5768a7d752a427814c2c28326706dcb6197a8c0/diff:/var/lib/docker/overlay2/bf815c8474c83d9ed5808b9a993be94d6b5d9a5f8687447db4137f92c1043303/diff:/var/lib/docker/overlay2/161555e8ed4342e6ce7edd7c24b84ca8952bd4f8902890a015506fdfc53924b0/diff:/var/lib/docker/overlay2/6f0caf038ca645f3d21d6014b9a0b9156239a2e4f553a2e3bca1cf6e9f0d9586/diff:/var/lib/docker/overlay2/97a8f753db26ef6ffd7a8f4e57f056a2791db542409e834294da82ce809a24d7/diff:/var/lib/docker/overlay2/3f9f5f17c39fda3f87bf1a4df90f14c5b03a99f6a2165ec675cbebc89c20f7dd/diff:/var/lib/docker/overlay2/a9fe04d811fa820e17bca110984cbd3137d42d9252df5d75d69af4728f9b497c/diff:/var/lib/docker/overlay2/a8b84b6fa94744d4d589a145082284c404f9a1d7ac84a8694a0db20719b98e2a/diff:/var/lib/docker/overlay2/afc2f3e2a0ee0cea23ca74a7426ef4a155d820eace5488bbc5674cb22e88d308/diff:/var/lib/docker/overlay2/2b2592d612db3b96f64cfbff0a6a2428cc671968fffac6e707ee251cf565451e/diff:/var/lib/docker/overlay2/30a69eb22d18107f0a8a75bddf3dce9ce80e08f6b16c0c33c93b17744bac88f1/diff:/var/lib/docker/overlay2/9faa49efcdcf7a994072663ab8808f9e9256f04092898690ab66c3451a07682f/diff:/var/lib/docker/overlay2/a4c4b7168a5306c4141de8a837d179671fd7dbb33e21cb7e1da9d322caeba5e5/diff:/var/lib/docker/overlay2/205a3b928fcf957d944f4c72d9d9b2bb15bc4723655b9376547b30b45db02575/diff:/var/lib/docker/overlay2/19ad75b84bca09d1105d2ddbacaeaae817f9b4617ebd49f2ec7da26799d0fef2/diff:/var/lib/docker/overlay2/b7f802c9aae6ffc1d5e6e7813b88b57a5f5097c0ea80cffd1c3ea2a513b5916a/diff:/var/lib/docker/overlay2/175038b463670224548b6da8e06595523480d3660dd53acf57334ae57a0077a0/diff:/var/lib/docker/overlay2/b146a204ff3a157f6273f99cd0ef9703dd8c73998e86f9ec3dab3c0ce445b7b5/diff:/var/lib/docker/overlay2/95a6073a2ee0623785b98f3c1cc20e3907e546e3195c528533f544a4ccceac9c/diff:/var/lib/docker/overlay2/ae7a8d43f473d402b35a55ae57c0bd7be35a811c11f31b6374367115c2617471/diff:/var/lib/docker/overlay2/4d642a507ed09a36c6a1a2add37d6138fc74e6593143b676c5988a5bf905140f/diff:/var/lib/docker/overlay2/9e3ac3f3d39aa13a4d1baf769a6819206b6b667b313ed29dabed61598167b750/diff:/var/lib/docker/overlay2/ebe9427f5546d8208d588a8ecd674aea338b5b40e88d4304f4b5f48bb4817733/diff:/var/lib/docker/overlay2/e40f6ca9bc1e17547503194818c60ae6280656faa253ccd1ff16d24e3c528c72/diff:/var/lib/docker/overlay2/f5028a2d0c876b6ae8908f8fa5b6f5ec96aadf1346a89a2f974dae9b5567b573/diff:/var/lib/docker/overlay2/2efd0dc1b011a6fab4094a446872fb200835f6083b939e7d7afdffd8b8dabcd9/diff:/var/lib/docker/overlay2/7bde34727a534931754f5b8d98e7f1ffde4621681a62f97dce18dd0539b54246/diff:/var/lib/docker/overlay2/cfb7fcb8524cabf359d97ba9690fc557edd420204d299b9c4b297e97746e49e1/diff:/var/lib/docker/overlay2/e5f34428eb3ae862f8d274877cd227196ec5990840c774f4936177e44afd9f48/diff:/var/lib/docker/overlay2/fc8762cccdca4b70e4bd3d7c2d1340ec46efaeb1b2b4cff6667e29ce58f23aa4/diff:/var/lib/docker/overlay2/76dd482b52cdaaf6b11a6b28aa8422ae62b2ddcc0abe16bf1586ee93f2c7f4eb/diff:/var/lib/docker/overlay2/ed76d74c37895e88532df350fc96dd16bc8136d7a4f324efb3c0fce892634a6b/diff:/var/lib/docker/overlay2/3f330655a672559c5b0bcd80bdd339ec7014c8bdc1994c18bb8a10b57db5568c/diff:/var/lib/docker/overlay2/24455530c66b3561e30793fe594dc5521685da3da7d40ebb116cd666e3613c09/diff:/var/lib/docker/overlay2/ef54943c082bdb71f58903dd773802ef7b732f78304c28db2e2689a68636e6af/diff:/var/lib/docker/overlay2/eec09c8d423a17fffa283bc45559a68292e0f537849b2a75db551ab3aba7b88c/diff:/var/lib/docker/overlay2/5f1630463727574518d638036df0bbee787b5f4e5a2459edec24e3e93e534065/diff:/var/lib/docker/overlay2/9eb7e3138b4060ab3c94f73a14adf69e6fe66d8f9a0c559c845a14348b43a985/diff:/var/lib/docker/overlay2/2678722ee3fa71efa858d2901572c508544031fff832e7f8c1a391ae3fbe2200/diff:/var/lib/docker/overlay2/ef2956881c2637255e992412ce34ace4b50d201f82a201ccba4a17644f7e1514/diff:/var/lib/docker/overlay2/d9452bcedd9787c5abad7eed58ecbc0af15477092548e31e73745b6d1b733c38/diff:/var/lib/docker/overlay2/735c0a799106215c26efddf0dce51e95e5fe0c1ff2471ed29c564a501ce234fc/diff:/var/lib/docker/overlay2/4892c6c642803de028725f60e9b0a5e11b64c62b0a31bf8b71754705efe1a16e/diff:/var/lib/docker/overlay2/01edf92adcbb11d8510c8421cda41b9625068e6f0c97c41a9ccfafe3f2f20d4e/diff:/var/lib/docker/overlay2/8db573f05a092eee10a89f76f00ee24fb6b269f64f8d739958c181a088a9f87b/diff:/var/lib/docker/overlay2/f214d99772916fe536a335fc87813d1264b90bb76ddfb24c2cc263cda55f1d29/diff:/var/lib/docker/overlay2/a6604a0830d63d91395a6547348678be56d7cd724457508ebf6c0e917739b99c/diff:/var/lib/docker/overlay2/cd2d70ea538970b6bed5a52be261953572bf12d380bf0d0f822a134bd90cd7ab/diff:/var/lib/docker/overlay2/f403a003ed32b114e9a336d2a234e1dd82289c68f1dc1817818c86d22c1b02b3/diff:/var/lib/docker/overlay2/f9cbccd3e0bb67e1ed94df62e2f269c8793ba5a69dde0656232de002011fbbcf/diff:/var/lib/docker/overlay2/45c6b4ef2d7c124fac26ce5b34ca653254552eceeab15976f94c4a107252be83/diff:/var/lib/docker/overlay2/dc9c1f8ec4624038305a9a2241dbb22214209be57428b7ad9cbb4145ede33891/diff:/var/lib/docker/overlay2/7e5aae577356a950f66f38590dcff58d5a74122f8b70655ecbb3cc0fde7ece8e/diff:/var/lib/docker/overlay2/dd580fefc51a22e0fbcf2b432a0adebffb19263a9aadf7406b6445b66811218b/diff:/var/lib/docker/overlay2/59388de726bc92bd4ded8f651cf67cbcdf5935043d4c674397a2a66b49ec143a/diff:/var/lib/docker/overlay2/85adf578c12f838fbfc3a9c29ed9253b35d32c1990e1ae438e24450e9d1cb5b6/diff",
            "MergedDir": "/var/lib/docker/overlay2/6832d65d1bc79dbb9d72f0eede88cc5c3299821bf790e619fc8d4e39e1aeaaa1/merged",
            "UpperDir": "/var/lib/docker/overlay2/6832d65d1bc79dbb9d72f0eede88cc5c3299821bf790e619fc8d4e39e1aeaaa1/diff",
            "WorkDir": "/var/lib/docker/overlay2/6832d65d1bc79dbb9d72f0eede88cc5c3299821bf790e619fc8d4e39e1aeaaa1/work"
        },
        "Name": "overlay2"
    },
    "RootFS": {
        "Type": "layers",
        "Layers": [
            "sha256:0d1a800648d7469f0278fe8acc8f0affd3e7199368ec9d0e4bfee32f62775852",
            "sha256:544242a509ad1e870fbf30caadd264c7b3612a1facd587d0b3e353ac6cc269d4",
            "sha256:e4e9bbb9e44e974e738e4deb5f25ffd582afca0b0e2c3102b25b2b1d7a8b5a5d",
            "sha256:baa1952ab39e611f83823cf10e34036d742a6c214c536c50eb65d1cc5e2ebb52",
            "sha256:b7ec8fd7c485bbedbe64cbc5a7a810ceabb9156e95c83ba60596ddb9b4955d48",
            "sha256:9b05faa4e8ddfbfe12730a58e502774173ff8be6c0c8b9b70627e727845fe054",
            "sha256:416baa63d3cabc0fd90edbe6b19007714712f1633daed14abf9b031ba929b736",
            "sha256:450ddd2a21b8c10d27ef81b2c24a572443e483fa23a76d373df004a94d8791c0",
            "sha256:c37e1f2f62dc2bf3e7d2e01b8767b3580a0e4cd04e2b63b1881b51da09d8d653",
            "sha256:5f3ec810ba3955c2607e7abb0f893ad09020ebe13b51bbcbf7ed372ba5ad3dc1",
            "sha256:d2861f46df62c6f20495abe4c0ce54953fa16fc2b55354aa89161a6812355a43",
            "sha256:4c157e838d0672da68db75bd153543d45f701be183b3ba3459fda6004cca1d57",
            "sha256:5b23fa06da10dc1834cd50d0ad3ac024af3c4ae5f6e31cc807403c746fb9cf0e",
            "sha256:0fe3342d43d627ba14a48ec20770298dcd1b86753ab699ae3e3f4d9069640dd6",
            "sha256:47020a0e5b12a2fe5e8802340ade6336dd524c3d09cea889ba6d2a9917ed63a2",
            "sha256:71f8622c456b2c1c79b049b4daf8c51de9287215dbec1858fe07d248d1bf17ff",
            "sha256:cbb8dea81ecc1f5afd1a9704c94b6adb44ddcf56f72b6420f6e2e34f30af8ac0",
            "sha256:ab1517524bba98780775cae4e097973b226666436f0d0b78c26556350e3b892a",
            "sha256:80089dd2113a0a37afe1313111650940af1cebb6fd4184b741b2306d13feead0",
            "sha256:85ff9f7d7eca7aabe87b81fc134f29a390cfa81e696d02078f6370b6565ff430",
            "sha256:647b997058987bfca0cb1803d78dfd9744d779a2e0e9d4a7cc10aa19937346ee",
            "sha256:972bd89abe9aa350d91a1943f62c9ea643d72c010dd87b2d5d23cb6606cbb6c4",
            "sha256:48b72973ea083f103070d7ffcccee20d759965038735323a17f6798455244e4e",
            "sha256:d618be6b7c159265f993e4161ad688c56fb90d951a400dc21eb7873f3ead6af9",
            "sha256:ff0f8b3a9d70da570a6085918f36b38e230f1e23644d9a17c6944202db339a36",
            "sha256:242a08fd767088c77557055dcf0063a1a3abe8ca4a9f42e6a6c9f397741b01d2",
            "sha256:b450fb62703ec72ddf7fd9430be56db9b2ae01f0a8afe535041b9323e3843ae7",
            "sha256:8442af9ea58298a6afa76bf34dc61b72e9cba4673afa9105503caf06f147b018",
            "sha256:15c0c005a740f9f5fa5cea47b1829ecb94e70a019ed3732e33216261e4169b97",
            "sha256:5f70bf18a086007016e948b04aed3b82103a36bea41755b6cddfaf10ace3c6ef",
            "sha256:35c14a70d3177eabf7a12149e9e2d14b35b8da770eb68938a801e8e7d2e4d70e",
            "sha256:1bdb9193635a8551f0591351e71f03c26f972769b57b2b9caf37feb3cc01ebef",
            "sha256:ba6b5357be557e4fffa6958a02afe5eaf072d7d1e1c2c4ac44fc543ecfb3f226",
            "sha256:849754a460bafd7471da102f2b9ef993258f3e8cd981dafd1a5cd1470a33a253",
            "sha256:71382f2fc469fac7e1d94d15224122384ce96a7bd1d15dc2fef592676c71f081",
            "sha256:5f70bf18a086007016e948b04aed3b82103a36bea41755b6cddfaf10ace3c6ef",
            "sha256:b35de4581860ac41f972c7741dce21e318d99fc6bd9ea6a7aaebbdef6b155b5c",
            "sha256:cdf4e7a43c01973038223080f72e627bd716caaa59f5dcfb26bc28403bb48cf2",
            "sha256:046f066aff261736f43a0e059a3dba6d8b4c31fb8e6444849a9c7a352e05fc9b",
            "sha256:3123611be36c7efc539e2cdd8930fa9475203f72e0be5d405d04585eb97b13fc",
            "sha256:bb9dc1279ae7ca4d685a1bc2361b1465eb89fbfefadb5e18cd69c827a981165e",
            "sha256:7937995096346416c4687fa2ae988304c81fb6552207bb1d7d629f74f2680a38",
            "sha256:33e7681eda8085205d906a2ad27ee9d80ccb2d1582b33d2ca1a7704f12845190",
            "sha256:eba435cbf985273afb39837b7d1d4e047daf5c6dc4e2b7ee27521cd787507f29",
            "sha256:f92429fc5d9656003697da06828ea92d69c8531d44c76f4b091534d3094696f7",
            "sha256:6583deda3c9769cecaf0d1ad056a93cd46376ca453446a75cc1c0d478403ec23",
            "sha256:77d08c9db634fe4aced902ce3b8822ec39ec2892886008fb03bdbdab5cb3e0fe",
            "sha256:25fdcf56a8bf36200ff976acf00479ce9486c16b13d0dd5fbca77a58dc73e2a6",
            "sha256:e07f6b9b7ac388bc6b242e80722d7ecf59ab11f71cef010e96182ac84ca03382",
            "sha256:efab9851f0314b6b894eadf44fe9bd53f0d5964dd873cb10e621684ed561e16e",
            "sha256:b4a46af1af798211cbf8a3d1d68c52285221588f1cfa4a7c9dd370c7d6db418a",
            "sha256:743a6c155bb1aed8d6a3f3fd9646a2bae1bd17e69aca4abd1a49270ff4f00dc2",
            "sha256:d1dc1a102fb1558488d6684950866a953eedc1c52113ecd23133b97a437bd9b2",
            "sha256:6115647aa95ac104266f01fc91ff3adc3859270b833211f8d4cd8b4e87c86b2c",
            "sha256:245fe4e3aa84943be8443af5c323d7d8103312db4a11911cc27ed70b67c11643",
            "sha256:2aad0875155deb62a066d9272f52fe43bbc9a3c3e0d3c3f7927f750a8204e436",
            "sha256:20ac2eac762d16e7d8135de2485e42187dfe809523214de9c8141b1ccede7ae5",
            "sha256:eb4af10fc33bce36e819d67f8fb87dfd4aa53a0ceaef2eeca6c0cd7d1cc4ddd1",
            "sha256:5f70bf18a086007016e948b04aed3b82103a36bea41755b6cddfaf10ace3c6ef",
            "sha256:5f70bf18a086007016e948b04aed3b82103a36bea41755b6cddfaf10ace3c6ef",
            "sha256:037facfd1305ce271d5684ee4b3c662454dfadd76df59bf39f68eccd3e689426",
            "sha256:707b9591a4615db0ddb7df435255396fa4a60255a13487bf8d438e6f1bcdde9b"
        ]
    },
    "Metadata": {
        "LastTagTime": "2026-09-18T05:51:40.157273279+08:00"
    }
}

更多版本

ghcr.io/aeon-7/aeon-vllm-ultimate:latest

linux/arm64 ghcr.io52.30GB2026-09-18 06:20
13
检测到您正在使用广告拦截插件,本站为公益站点,依赖广告维持运转 🙏 查看如何关闭 ×