Sign inSign up

sabreshao/vllm:k3_0814

Manifest digest

sha256:7c8a137960813c4a1a23375561a491d9379442732420b07d7255587a04867b4d

OS/ARCH

linux/amd64

Compressed size

14 GB

Last pushed

about 2 months by sabreshao

Type

Image

Manifest digest

sha256:7c8a137960813c4a1a23375561a491d9379442732420b07d7255587a04867b4d

Image Layers

1ARG RELEASE0 B
2ARG LAUNCHPAD_BUILD_ARCH0 B
3LABEL org.opencontainers.image.version=22.040 B
4ADD file ... in / 29.23 MB
5CMD ["/bin/bash"]0 B
7ARG ROCM_VERSION=5.30 B
8ARG AMDGPU_VERSION=5.30 B
9ARG APT_PREF0 B
10RUN |3 ROCM_VERSION=7.2.3 AMDGPU_VERSION=30.30.3 APT_PREF=Package:240 B
11RUN |3 ROCM_VERSION=7.2.3 AMDGPU_VERSION=30.30.3 APT_PREF=Package:6.76 GB
12RUN |3 ROCM_VERSION=7.2.3 AMDGPU_VERSION=30.30.3 APT_PREF=Package:655 B
13ENV PATH=/opt/rocm/llvm/bin:/opt/rocm/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin0 B
14ENV ROCM_PATH=/opt/rocm0 B
15ENV LD_LIBRARY_PATH=/opt/rocm/lib:/usr/local/lib:0 B
16ARG PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx11510 B
17ENV PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx11510 B
18ENV AITER_ROCM_ARCH=gfx942;gfx9500 B
19ENV HSA_NO_SCRATCH_RECLAIM=10 B
20ARG PYTHON_VERSION=3.120 B
21ENV PYTHON_VERSION=3.120 B
22RUN |2 PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151 PYTHON_VERSION=3.12 /bin/sh93 B
23WORKDIR /app32 B
24ENV DEBIAN_FRONTEND=noninteractive0 B
25RUN |2 PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151 PYTHON_VERSION=3.12 /bin/sh149.76 MB
26RUN |2 PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151 PYTHON_VERSION=3.12 /bin/sh65.89 MB
27RUN |2 PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151 PYTHON_VERSION=3.12 /bin/sh4.06 MB
28ARG USE_SCCACHE0 B
29ARG SCCACHE_DOWNLOAD_URL0 B
30ARG SCCACHE_ENDPOINT0 B
31ARG SCCACHE_BUCKET_NAME0 B
32ARG SCCACHE_REGION_NAME0 B
33ARG SCCACHE_S3_NO_CREDENTIALS0 B
34RUN |8 PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151 PYTHON_VERSION=3.12 USE_SCCACHE=19.11 MB
35RUN |8 PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx1151 PYTHON_VERSION=3.12 USE_SCCACHE=1229 B
36ARG USE_SCCACHE0 B
37ENV SCCACHE_BUCKET=vllm-build-sccache0 B
38ENV SCCACHE_REGION=us-west-20 B
39ENV SCCACHE_S3_NO_CREDENTIALS=00 B
40ENV SCCACHE_IDLE_TIMEOUT=00 B
41RUN /bin/sh -c pip install2.41 GB
42ARG BASE_IMAGE0 B
43ARG TRITON_BRANCH0 B
44ARG TRITON_REPO0 B
45ARG PYTORCH_BRANCH0 B
46ARG PYTORCH_VISION_BRANCH0 B
47ARG PYTORCH_REPO0 B
48ARG PYTORCH_VISION_REPO0 B
49ARG PYTORCH_AUDIO_BRANCH0 B
50ARG PYTORCH_AUDIO_REPO0 B
51ARG FA_BRANCH0 B
52ARG FA_REPO0 B
53ARG AITER_BRANCH0 B
54ARG AITER_REPO0 B
55ARG MORI_BRANCH0 B
56ARG MORI_REPO0 B
57RUN |15 BASE_IMAGE=rocm/dev-ubuntu-22.04:7.2.3-complete TRITON_BRANCH=0f380657 TRITON_REPO=https://github.com/ROCm/triton.git398 B
58ARG ARG_PYTORCH_ROCM_ARCH0 B
59ENV PYTORCH_ROCM_ARCH=gfx90a;gfx942;gfx950;gfx1100;gfx1101;gfx1200;gfx1201;gfx1150;gfx11510 B
60RUN |1 ARG_PYTORCH_ROCM_ARCH= /bin/sh -c142.07 MB
61RUN |1 ARG_PYTORCH_ROCM_ARCH= /bin/sh -c92 B
62ARG USE_SCCACHE0 B
63RUN |2 ARG_PYTORCH_ROCM_ARCH= USE_SCCACHE=1 /bin/sh32 B
64RUN |2 ARG_PYTORCH_ROCM_ARCH= USE_SCCACHE=1 /bin/sh25.66 MB
65ENV UV_HTTP_TIMEOUT=5000 B
66ENV UV_INDEX_STRATEGY=unsafe-best-match0 B
67ENV UV_LINK_MODE=copy0 B
68ENV CCACHE_DIR=/root/.cache/ccache0 B
69ENV CCACHE_COMPILERCHECK=content0 B
70ARG max_jobs0 B
71ENV MAX_JOBS=160 B
72ARG USE_SCCACHE0 B
73ARG SCCACHE_DOWNLOAD_URL0 B
74ARG SCCACHE_ENDPOINT0 B
75ARG SCCACHE_BUCKET_NAME0 B
76ARG SCCACHE_REGION_NAME0 B
77ARG SCCACHE_S3_NO_CREDENTIALS0 B
78RUN |9 ARG_PYTORCH_ROCM_ARCH= USE_SCCACHE=1 max_jobs=1632 B
79ARG USE_SCCACHE0 B
80ENV SCCACHE_BUCKET=vllm-build-sccache0 B
81ENV SCCACHE_REGION=us-west-20 B
82ENV SCCACHE_S3_NO_CREDENTIALS=00 B
83ENV SCCACHE_IDLE_TIMEOUT=00 B
84ARG COMMON_WORKDIR0 B
85WORKDIR /app32 B
86ARG NIC_BACKEND0 B
87ARG AINIC_VERSION0 B
88ARG UBUNTU_CODENAME0 B
89RUN |3 NIC_BACKEND=all AINIC_VERSION=1.117.3-hydra UBUNTU_CODENAME=jammy15.89 MB
90RUN /bin/sh -c python3 -m1.38 KB
91RUN /bin/sh -c rm -f196 B
92ENV SCCACHE_BUCKET=0 B
93ENV SCCACHE_REGION=0 B
94ENV SCCACHE_ENDPOINT=0 B
95ENV SCCACHE_S3_NO_CREDENTIALS=0 B
96ENV SCCACHE_IDLE_TIMEOUT=0 B
97RUN /bin/sh -c case "$(which32 B
98RUN /bin/sh -c uv pip6.72 MB
99RUN /bin/sh -c cd /install707.91 MB
100RUN /bin/sh -c uv pip24.65 MB
101ARG COMMON_WORKDIR0 B
102ARG BASE_IMAGE0 B
103ARG NIC_BACKEND0 B
104ARG AINIC_VERSION0 B
105COPY /benchmarks /app/vllm/benchmarks # buildkit259.44 KB
106COPY /examples /app/vllm/examples # buildkit262.16 KB
107COPY /docker /app/vllm/docker # buildkit33.03 KB
108ENV HSA_ENABLE_IPC_MODE_LEGACY=10 B
109ENV UCX_RMA_PPLN_ENABLE=y0 B
110ENV UCX_ROCM_COPY_SIGPOOL_MAX_ELEMS=inf0 B
111ENV TOKENIZERS_PARALLELISM=false0 B
112ENV SAFETENSORS_FAST_GPU=10 B
113ENV HIP_FORCE_DEV_KERNARG=10 B
114ENV PYTORCH_NVML_BASED_CUDA_CHECK=10 B
115RUN |4 COMMON_WORKDIR=/app BASE_IMAGE=public.ecr.aws/q9t5s3a7/vllm-release-repo:3f72e75809f39571-rocm-base NIC_BACKEND=all168 B
116ENV KINETO_CONFIG=/app/libkineto.conf0 B
117RUN |4 COMMON_WORKDIR=/app BASE_IMAGE=public.ecr.aws/q9t5s3a7/vllm-release-repo:3f72e75809f39571-rocm-base NIC_BACKEND=all490 B
118CMD ["/bin/bash"]0 B
119ENTRYPOINT ["vllm" "serve"]0 B
1202.83 GB
121392.96 MB
122ARG VLLM_SOURCE=worktrees/0814-customer-vllm0 B
123ENV GPU_ARCHS=gfx950 PYTORCH_ROCM_ARCH=gfx950 AITER_USE_SYSTEM_TRITON=1 TRITON_KERNELS_SRC_DIR=/opt/triton-src/python/triton_kernels/triton_kernels0 B
124COPY worktrees/0813-customer-aiter /opt/aiter-main # buildkit29.18 MB
125COPY ck-src /opt/aiter-main/3rdparty/composable_kernel # buildkit86.81 MB
126COPY wheels /opt/wheels # buildkit68.81 MB
127COPY triton-src /opt/triton-src # buildkit6.33 MB
128RUN |1 VLLM_SOURCE=worktrees/0814-customer-vllm /bin/sh -c73.56 MB
129COPY worktrees/0814-customer-vllm /opt/vllm-main # buildkit38.1 MB
130RUN |1 VLLM_SOURCE=worktrees/0814-customer-vllm /bin/sh -c173.46 MB
131RUN |1 VLLM_SOURCE=worktrees/0814-customer-vllm /bin/sh -c6.21 KB
132ENV AITER_CONFIG_FMOE=/opt/aiter-configs/kimik3-a8w4-tuned-fmoe.csv AITER_CONFIG_GEMM_BF16=/opt/aiter-configs/kimik3-bf16-tuned-gemm.csv0 B
133ARG VLLM_COMMIT=8e6d8e4f6a0c84db0d79129ab648492edf640fe20 B
134ARG AITER_COMMIT=b5c5912cb2bbab5d8c0cc12b08dec391a27bf8df0 B
135ARG VLLM_DIFF_SHA256=6fcf4ceadeda7c720fdf52f9511816dcfb901bb0a4fc495dfbf115626da7b9590 B
136ARG AITER_DIFF_SHA256=b2efecf9fe85831faefc38b0f6e1d0f5e24a617cd0b06e089b8d2c6e30522c7c0 B
137ARG VLLM_OVERLAYS=merged-main-includes:51860,51843,51772,51862,51653;open:51171,51647,52000,52047,522050 B
138ARG AITER_OVERLAYS=pr4710,pr47130 B
139ARG BUILD_DATE=2026-08-14T14:07:09+00:000 B
140ARG VALIDATION_STATUS=runtime-validation-pending0 B
141ARG IMAGE_NAME=kimi-k3-mi355x-customer-0814-stable-v30 B
142ARG RECIPE_VERSION=customer-0814-stable-v30 B
143ARG MODEL_SNAPSHOT=9f62e4e9fffbd0a83ddd60e1c209d828994b35690 B
144ARG DSPARK_SNAPSHOT=cf6b8244620e7ea4b0651d214f28e89eac75bed60 B
145LABEL ai.image.name=kimi-k3-mi355x-customer-0814-stable-v3 ai.base.image.digest=sha256:5aa7e626ff73672f5ca7aae46754570488c23d33ca1ac90756a1d2d1a3fe099b ai.triton_kernels.commit=0f380657dbf3ee86eb57558ff71df24f03b5d4e7 ai.vllm.commit=8e6d8e4f6a0c84db0d79129ab648492edf640fe20 B
146RUN |13 VLLM_SOURCE=worktrees/0814-customer-vllm VLLM_COMMIT=8e6d8e4f6a0c84db0d79129ab648492edf640fe2 AITER_COMMIT=b5c5912cb2bbab5d8c0cc12b08dec391a27bf8df140.42 KB
147RUN |13 VLLM_SOURCE=worktrees/0814-customer-vllm VLLM_COMMIT=8e6d8e4f6a0c84db0d79129ab648492edf640fe2 AITER_COMMIT=b5c5912cb2bbab5d8c0cc12b08dec391a27bf8df390.12 KB
148ARG BASE_IMAGE=kimi-k3-mi355x:customer-0814-stable-v30 B
149ARG RUNNER_PATCH_SHA256=7da2e83f7edd768a434cbed3ad2ec8ab2c6ad45d0687aff3f18e660defd5e1160 B
150ARG DIRECT_RETURN_SHA256=9a91ea1624242e38a1acdb18b81d55b6d265a9606d68a8655052daf1c53b6fda0 B
151ARG BUILD_DATE=2026-08-15T08:54:42+00:000 B
152ARG VALIDATION_STATUS=runtime-gates-passed0 B
153COPY worktrees/0814-customer-vllm/vllm/config/vllm.py /opt/vllm-main/vllm/config/vllm.py # buildkit26.05 KB
154COPY worktrees/0814-customer-vllm/vllm/models/kimi_k3/amd/linear.py /opt/vllm-main/vllm/models/kimi_k3/amd/linear.py # buildkit7.98 KB
155COPY worktrees/0814-customer-vllm/vllm/models/kimi_k3/amd/kda.py /opt/vllm-main/vllm/models/kimi_k3/amd/kda.py # buildkit5.93 KB
156LABEL ai.image.name=kimi-k3-mi355x-customer-0814-stable-v3-18ms ai.parent.image=kimi-k3-mi355x:customer-0814-stable-v3 ai.patch.name=kimi-k3-rocm-legacy-runner-default+attention-direct-return ai.patch.runner.sha256=7da2e83f7edd768a434cbed3ad2ec8ab2c6ad45d0687aff3f18e660defd5e1160 B
157RUN |5 BASE_IMAGE=kimi-k3-mi355x:customer-0814-stable-v3 RUNNER_PATCH_SHA256=7da2e83f7edd768a434cbed3ad2ec8ab2c6ad45d0687aff3f18e660defd5e116 DIRECT_RETURN_SHA256=9a91ea1624242e38a1acdb18b81d55b6d265a9606d68a8655052daf1c53b6fda66.37 KB
158ARG BASE_IMAGE=sabreshao/vllm:k3_08140 B
159ARG RUNNER_PATCH_SHA256=7da2e83f7edd768a434cbed3ad2ec8ab2c6ad45d0687aff3f18e660defd5e1160 B
160ARG DIRECT_RETURN_SHA256=9a91ea1624242e38a1acdb18b81d55b6d265a9606d68a8655052daf1c53b6fda0 B
161ARG H24_PADDING_SHA256=cbf541bc6419fe52d6288231fdfc1d2d6371e2c3d43819f0d826381c1132a9300 B
162ARG BUILD_DATE=2026-08-17T06:20:13+00:000 B
163ARG VALIDATION_STATUS=h24-pilot-passed0 B
164COPY worktrees/0814-customer-vllm/vllm/config/vllm.py /opt/vllm-main/vllm/config/vllm.py # buildkit26.05 KB
165COPY worktrees/0814-customer-vllm/vllm/models/kimi_k3/amd/linear.py /opt/vllm-main/vllm/models/kimi_k3/amd/linear.py # buildkit7.97 KB
166COPY worktrees/0814-customer-vllm/vllm/models/kimi_k3/amd/kda.py /opt/vllm-main/vllm/models/kimi_k3/amd/kda.py # buildkit5.93 KB
167COPY worktrees/0814-customer-vllm/vllm/v1/attention/backends/mla/rocm_aiter_mla.py /opt/vllm-main/vllm/v1/attention/backends/mla/rocm_aiter_mla.py # buildkit15.88 KB
168LABEL ai.image.name=kimi-k3-mi355x-customer-0814-stable-v3-18ms ai.parent.image=sabreshao/vllm:k3_0814 ai.patch.name=kimi-k3-rocm-legacy-runner-default+attention-direct-return+h24-metadata-padding ai.patch.runner.sha256=7da2e83f7edd768a434cbed3ad2ec8ab2c6ad45d0687aff3f18e660defd5e1160 B
169RUN |6 BASE_IMAGE=sabreshao/vllm:k3_0814 RUNNER_PATCH_SHA256=7da2e83f7edd768a434cbed3ad2ec8ab2c6ad45d0687aff3f18e660defd5e116 DIRECT_RETURN_SHA256=9a91ea1624242e38a1acdb18b81d55b6d265a9606d68a8655052daf1c53b6fda22.17 KB

Command

 ARG RELEASE