gpustack-runner 0.1.24.post3__tar.gz → 0.1.24.post4__tar.gz

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: gpustack-runner
-Version: 0.1.24.post3
+Version: 0.1.24.post4
 Summary: GPUStack Runner is library for registering runnable accelerated backends and services in GPUStack.
 Project-URL: Homepage, https://github.com/gpustack/runner
 Project-URL: Bug Tracker, https://github.com/gpustack/gpustack/issues
@@ -128,10 +128,10 @@ The following table lists the supported accelerated backends and their correspon
 > - ROCm 6.4 SGLang supports `gfx942` only.
 > - ROCm 7.0 SGLang supports `gfx950` only.
 
-| ROCm Version <br/> (Variant) | vLLM                                            | SGLang                                     |
-|------------------------------|-------------------------------------------------|--------------------------------------------|
-| 7.0                          | **`0.13.0`**, `0.12.0`, <br/>`0.11.2`           | `0.5.7`, `0.5.6.post2`                     |
-| 6.4                          | **`0.13.0`**, `0.12.0`, <br/>`0.11.2`, `0.10.2` | `0.5.7`, `0.5.6.post2`, <br/>`0.5.5.post3` |
+| ROCm Version <br/> (Variant) | vLLM                                                      | SGLang                                              |
+|------------------------------|-----------------------------------------------------------|-----------------------------------------------------|
+| 7.0                          | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`           | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`                |
+| 6.4                          | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, `0.10.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`, `0.5.5.post3` |
 
 ## Directory Structure
 

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/README.md

@@ -108,10 +108,10 @@ The following table lists the supported accelerated backends and their correspon
 > - ROCm 6.4 SGLang supports `gfx942` only.
 > - ROCm 7.0 SGLang supports `gfx950` only.
 
-| ROCm Version <br/> (Variant) | vLLM                                            | SGLang                                     |
-|------------------------------|-------------------------------------------------|--------------------------------------------|
-| 7.0                          | **`0.13.0`**, `0.12.0`, <br/>`0.11.2`           | `0.5.7`, `0.5.6.post2`                     |
-| 6.4                          | **`0.13.0`**, `0.12.0`, <br/>`0.11.2`, `0.10.2` | `0.5.7`, `0.5.6.post2`, <br/>`0.5.5.post3` |
+| ROCm Version <br/> (Variant) | vLLM                                                      | SGLang                                              |
+|------------------------------|-----------------------------------------------------------|-----------------------------------------------------|
+| 7.0                          | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`           | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`                |
+| 6.4                          | `0.14.1`, **`0.13.0`**, <br/>`0.12.0`, `0.11.2`, `0.10.2` | `0.5.8`, `0.5.7`, <br/>`0.5.6.post2`, `0.5.5.post3` |
 
 ## Directory Structure
 

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/gpustack_runner/_version.py

@@ -27,8 +27,8 @@ version_tuple: VERSION_TUPLE
 __commit_id__: COMMIT_ID
 commit_id: COMMIT_ID
 
-__version__ = version = '0.1.24.post3'
-__version_tuple__ = version_tuple = (0, 1, 24, 'post3')
+__version__ = version = '0.1.24.post4'
+__version_tuple__ = version_tuple = (0, 1, 24, 'post4')
 try:
     from ._version_appendix import git_commit
     __commit_id__ = commit_id = git_commit

gpustack_runner-0.1.24.post4/gpustack_runner/_version_appendix.py

@@ -0,0 +1 @@
+git_commit = "907c39c"

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/gpustack_runner/runner.py.json

@@ -1748,6 +1748,17 @@
     "docker_image": "gpustack/runner:musa4.1-vllm0.9.2",
     "deprecated": false
   },
+  {
+    "backend": "rocm",
+    "backend_version": "7.0",
+    "original_backend_version": "7.0.2",
+    "backend_variant": "",
+    "service": "sglang",
+    "service_version": "0.5.8",
+    "platform": "linux/amd64",
+    "docker_image": "gpustack/runner:rocm7.0-sglang0.5.8",
+    "deprecated": false
+  },
   {
     "backend": "rocm",
     "backend_version": "7.0",
@@ -1770,6 +1781,17 @@
     "docker_image": "gpustack/runner:rocm7.0-sglang0.5.6.post2",
     "deprecated": false
   },
+  {
+    "backend": "rocm",
+    "backend_version": "7.0",
+    "original_backend_version": "7.0.2",
+    "backend_variant": "",
+    "service": "vllm",
+    "service_version": "0.14.1",
+    "platform": "linux/amd64",
+    "docker_image": "gpustack/runner:rocm7.0-vllm0.14.1",
+    "deprecated": false
+  },
   {
     "backend": "rocm",
     "backend_version": "7.0",
@@ -1814,6 +1836,17 @@
     "docker_image": "gpustack/runner:rocm7.0-vllm0.11.0",
     "deprecated": true
   },
+  {
+    "backend": "rocm",
+    "backend_version": "6.4",
+    "original_backend_version": "6.4.4",
+    "backend_variant": "",
+    "service": "sglang",
+    "service_version": "0.5.8",
+    "platform": "linux/amd64",
+    "docker_image": "gpustack/runner:rocm6.4-sglang0.5.8",
+    "deprecated": false
+  },
   {
     "backend": "rocm",
     "backend_version": "6.4",
@@ -1847,6 +1880,17 @@
     "docker_image": "gpustack/runner:rocm6.4-sglang0.5.5.post3",
     "deprecated": false
   },
+  {
+    "backend": "rocm",
+    "backend_version": "6.4",
+    "original_backend_version": "6.4.4",
+    "backend_variant": "",
+    "service": "vllm",
+    "service_version": "0.14.1",
+    "platform": "linux/amd64",
+    "docker_image": "gpustack/runner:rocm6.4-vllm0.14.1",
+    "deprecated": false
+  },
   {
     "backend": "rocm",
     "backend_version": "6.4",

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/pack/cann/Dockerfile

@@ -59,7 +59,7 @@ ARG VLLM_VERSION=0.14.1
 ARG VLLM_ASCEND_VERSION=0.14.0rc1
 ARG VLLM_TORCH_VERSION=2.9.0
 ARG VLLM_MOONCAKE_VERSION=0.3.7.post2
-ARG VLLM_OMNI_COMMIT=e8aa32b
+ARG VLLM_OMNI_COMMIT=b11d436
 ARG SGLANG_BASE_IMAGE=gpustack/runner:cann${CANN_VERSION}-${CANN_ARCHS}-python${PYTHON_VERSION}
 ARG SGLANG_VERSION=0.5.8
 ARG SGLANG_TORCH_VERSION=2.8.0
@@ -865,6 +865,15 @@ RUN --mount=type=bind,from=vllm-build-omni,source=/,target=/omni,rw <<EOF
     uv pip install --no-build-isolation \
         /omni/workspace/*.whl
 
+    # Dependencies
+    uv pip uninstall onnxruntime || true
+    cat <<EOT >/tmp/requirements.txt
+onnxruntime-cann
+sox
+EOT
+    uv pip install \
+        -r /tmp/requirements.txt
+
     # Cleanup
     rm -rf /var/tmp/* \
         && rm -rf /tmp/*
@@ -956,7 +965,11 @@ RUN --mount=type=bind,target=/workspace,rw <<EOF
 
     tree -hs /workspace/patches
     pushd $(pip show vllm | grep Location: | cut -d" " -f 2) \
-        && patch -p1 < /workspace/patches/vllm_*.patch
+        && patch -p1 < /workspace/patches/vllm/*.patch
+    if pip show vllm_omni > /dev/null 2>&1; then \
+        pushd $(pip show vllm_omni | grep Location: | cut -d" " -f 2) \
+            && patch -p1 < /workspace/patches/vllm_omni/*.patch; \
+    fi
 EOF
 
 ## Entrypoint

gpustack_runner-0.1.24.post4/pack/cann/patches/vllm_omni/001_wrong_patch.patch

@@ -0,0 +1,13 @@
+diff --git a/vllm_omni/patch.py b/vllm_omni/patch.py
+index 687ff51..6b67924 100644
+--- a/vllm_omni/patch.py
++++ b/vllm_omni/patch.py
+@@ -19,6 +19,8 @@ for module_name, module in sys.modules.items():
+     # only do patch on module of vllm, pass others
+     if "vllm" not in module_name:
+         continue
++    if "--omni" not in sys.argv:
++        continue
+     if hasattr(module, "EngineCoreOutput") and module.EngineCoreOutput == _OriginalEngineCoreOutput:
+         module.EngineCoreOutput = OmniEngineCoreOutput
+     if hasattr(module, "EngineCoreOutputs") and module.EngineCoreOutputs == _OriginalEngineCoreOutputs:

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/pack/cuda/Dockerfile

@@ -116,7 +116,7 @@ ARG VLLM_FLASHINFER_VERSION=0.6.1
 ARG VLLM_FLASHATTENTION_VERSION=2.8.3
 ARG VLLM_LMCACHE_VERSION=0.3.12
 ARG VLLM_MOONCAKE_VERSION=0.3.8.post1
-ARG VLLM_OMNI_COMMIT=e8aa32b
+ARG VLLM_OMNI_COMMIT=2c294a7
 ARG SGLANG_BASE_IMAGE=vllm
 ARG SGLANG_VERSION=0.5.8
 ARG SGLANG_BUILD_BASE_IMAGE=vllm-build
@@ -1357,6 +1357,14 @@ RUN --mount=type=bind,from=vllm-build-omni,source=/,target=/omni,rw <<EOF
     uv pip install --no-build-isolation \
         /omni/workspace/*.whl
 
+    # Dependencies
+    cat <<EOT >/tmp/requirements.txt
+onnxruntime
+sox
+EOT
+    uv pip install \
+        -r /tmp/requirements.txt
+
     # Cleanup
     rm -rf /var/tmp/* \
         && rm -rf /tmp/*
@@ -1551,7 +1559,11 @@ RUN --mount=type=bind,target=/workspace,rw <<EOF
 
     tree -hs /workspace/patches
     pushd $(pip show vllm | grep Location: | cut -d" " -f 2) \
-        && patch -p1 < /workspace/patches/vllm_*.patch
+        && patch -p1 < /workspace/patches/vllm/*.patch
+    if pip show vllm_omni > /dev/null 2>&1; then \
+        pushd $(pip show vllm_omni | grep Location: | cut -d" " -f 2) \
+            && patch -p1 < /workspace/patches/vllm_omni/*.patch; \
+    fi
 EOF
 
 ## Entrypoint

gpustack_runner-0.1.24.post4/pack/cuda/patches/vllm_omni/001_wrong_patch.patch

@@ -0,0 +1,13 @@
+diff --git a/vllm_omni/patch.py b/vllm_omni/patch.py
+index 687ff51..6b67924 100644
+--- a/vllm_omni/patch.py
++++ b/vllm_omni/patch.py
+@@ -19,6 +19,8 @@ for module_name, module in sys.modules.items():
+     # only do patch on module of vllm, pass others
+     if "vllm" not in module_name:
+         continue
++    if "--omni" not in sys.argv:
++        continue
+     if hasattr(module, "EngineCoreOutput") and module.EngineCoreOutput == _OriginalEngineCoreOutput:
+         module.EngineCoreOutput = OmniEngineCoreOutput
+     if hasattr(module, "EngineCoreOutputs") and module.EngineCoreOutputs == _OriginalEngineCoreOutputs:

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/pack/rocm/Dockerfile

@@ -84,7 +84,7 @@ ARG VLLM_FLASHATTENTION_VERSION=2.8.3
 ARG VLLM_AITER_VERSION=0.1.8
 ARG VLLM_LMCACHE_VERSION=0.3.12
 ARG VLLM_MOONCAKE_VERSION=0.3.8.post1
-ARG VLLM_OMNI_COMMIT=e8aa32b
+ARG VLLM_OMNI_COMMIT=2c294a7
 ARG SGLANG_BASE_IMAGE=vllm
 ARG SGLANG_VERSION=0.5.8
 ARG SGLANG_BUILD_BASE_IMAGE=vllm-build
@@ -724,6 +724,11 @@ RUN --mount=type=bind,from=vllm-build-flashattention,source=/,target=/flashatten
         --depth 1 --branch v${VLLM_VERSION} --single-branch \
         https://github.com/vllm-project/vllm.git vllm
     pushd /tmp/vllm \
+        && sed -i "s/\"torch ==.*\"/\"torch\"/g" /tmp/vllm/pyproject.toml \
+        && sed -i "s/\"torch==.*\"/\"torch\"/g" /tmp/vllm/requirements/rocm-build.txt \
+        && sed -i "s/\"torchvision==.*\"/\"torchvision\"/g" /tmp/vllm/requirements/rocm-build.txt \
+        && sed -i "s/\"torchaudio==.*\"/\"torchaudio\"/g" /tmp/vllm/requirements/rocm-build.txt \
+        && sed -i "s/\"triton==.*\"/\"triton\"/g" /tmp/vllm/requirements/rocm-build.txt \
         && uv pip install protobuf>=6.33.2 grpcio-tools>=1.76.0 \
         && VLLM_TARGET_DEVICE="rocm" python -v -m build --no-isolation --wheel \
         && tree -hs /tmp/vllm/dist \
@@ -1079,6 +1084,15 @@ RUN --mount=type=bind,from=vllm-build-omni,source=/,target=/omni,rw <<EOF
     uv pip install --no-build-isolation \
         /omni/workspace/*.whl
 
+    # Dependencies
+    uv pip uninstall onnxruntime || true
+    cat <<EOT >/tmp/requirements.txt
+onnxruntime-rocm
+sox
+EOT
+    uv pip install \
+        -r /tmp/requirements.txt
+
     # Cleanup
     rm -rf /var/tmp/* \
         && rm -rf /tmp/*
@@ -1236,7 +1250,11 @@ RUN --mount=type=bind,target=/workspace,rw <<EOF
 
     tree -hs /workspace/patches
     pushd $(pip show vllm | grep Location: | cut -d" " -f 2) \
-        && patch -p1 < /workspace/patches/vllm_*.patch
+        && patch -p1 < /workspace/patches/vllm/*.patch
+    if pip show vllm_omni > /dev/null 2>&1; then \
+        pushd $(pip show vllm_omni | grep Location: | cut -d" " -f 2) \
+            && patch -p1 < /workspace/patches/vllm_omni/*.patch; \
+    fi
 EOF
 
 ## Entrypoint
@@ -1537,7 +1555,7 @@ RUN --mount=type=bind,target=/workspace,rw <<EOF
 
     tree -hs /workspace/patches
     pushd $(pip show sglang | grep Location: | cut -d" " -f 2) \
-        && patch -p1 < /workspace/patches/sglang_*.patch
+        && patch -p1 < /workspace/patches/sglang/*.patch
 EOF
 
 ## Entrypoint

gpustack_runner-0.1.24.post4/pack/rocm/patches/vllm_omni/001_wrong_patch.patch

@@ -0,0 +1,13 @@
+diff --git a/vllm_omni/patch.py b/vllm_omni/patch.py
+index 687ff51..6b67924 100644
+--- a/vllm_omni/patch.py
++++ b/vllm_omni/patch.py
+@@ -19,6 +19,8 @@ for module_name, module in sys.modules.items():
+     # only do patch on module of vllm, pass others
+     if "vllm" not in module_name:
+         continue
++    if "--omni" not in sys.argv:
++        continue
+     if hasattr(module, "EngineCoreOutput") and module.EngineCoreOutput == _OriginalEngineCoreOutput:
+         module.EngineCoreOutput = OmniEngineCoreOutput
+     if hasattr(module, "EngineCoreOutputs") and module.EngineCoreOutputs == _OriginalEngineCoreOutputs:

{gpustack_runner-0.1.24.post3 → gpustack_runner-0.1.24.post4}/tests/gpustack_runner/fixtures/test_list_runners_by_backend.json

@@ -1810,6 +1810,17 @@
       "backend": "rocm"
     },
     [
+      {
+        "backend": "rocm",
+        "backend_version": "7.0",
+        "original_backend_version": "7.0.2",
+        "backend_variant": "",
+        "service": "sglang",
+        "service_version": "0.5.8",
+        "platform": "linux/amd64",
+        "docker_image": "gpustack/runner:rocm7.0-sglang0.5.8",
+        "deprecated": false
+      },
       {
         "backend": "rocm",
         "backend_version": "7.0",
@@ -1832,6 +1843,17 @@
         "docker_image": "gpustack/runner:rocm7.0-sglang0.5.6.post2",
         "deprecated": false
       },
+      {
+        "backend": "rocm",
+        "backend_version": "7.0",
+        "original_backend_version": "7.0.2",
+        "backend_variant": "",
+        "service": "vllm",
+        "service_version": "0.14.1",
+        "platform": "linux/amd64",
+        "docker_image": "gpustack/runner:rocm7.0-vllm0.14.1",
+        "deprecated": false
+      },
       {
         "backend": "rocm",
         "backend_version": "7.0",
@@ -1876,6 +1898,17 @@
         "docker_image": "gpustack/runner:rocm7.0-vllm0.11.0",
         "deprecated": true
       },
+      {
+        "backend": "rocm",
+        "backend_version": "6.4",
+        "original_backend_version": "6.4.4",
+        "backend_variant": "",
+        "service": "sglang",
+        "service_version": "0.5.8",
+        "platform": "linux/amd64",
+        "docker_image": "gpustack/runner:rocm6.4-sglang0.5.8",
+        "deprecated": false
+      },
       {
         "backend": "rocm",
         "backend_version": "6.4",
@@ -1909,6 +1942,17 @@
         "docker_image": "gpustack/runner:rocm6.4-sglang0.5.5.post3",
         "deprecated": false
       },
+      {
+        "backend": "rocm",
+        "backend_version": "6.4",
+        "original_backend_version": "6.4.4",
+        "backend_variant": "",
+        "service": "vllm",
+        "service_version": "0.14.1",
+        "platform": "linux/amd64",
+        "docker_image": "gpustack/runner:rocm6.4-vllm0.14.1",
+        "deprecated": false
+      },
       {
         "backend": "rocm",
         "backend_version": "6.4",

gpustack_runner-0.1.24.post3/gpustack_runner/_version_appendix.py

@@ -1 +0,0 @@
-git_commit = "dc41ed2"