feat: update aphrodite to 0.20.0 (#1628)

* feat: update aphrodite to 0.20.0

* fix: stale package reference

* fix: flash attn

* fix: aphrodite_flash_attn path

* fix: rejection sampler

* fix: gitignore triton_kernels and fix flash_attn src copy again

* chore: some logging cleanup

* chore: a couple more logger fixes

* chore: remove redundant progress bar for dynamo

* chore: some centering relative to logo

* fix: silence deprecation warnings in transformers processor

* Revert "fix: silence deprecation warnings in transformers processor"

This reverts commit 7886110.

* fix: demote log to debug for compilation time messages

* fix: coerce dry strings

* fix: dry not being applied

* feat: kobold api

* fix: copyright headers

* fix: a bunch of sampler issues

* refactor: remove sampler priority

* feat: temperature_last

* chore: docker stuff

* fix: some related tooling around docker

* fix: remove stale aphrodite_kernels

* fix: no precompile step in docker

* fix: remove wheel size check from Dockerfile and export_wheels.sh

* fix: update export_wheels.sh script to remove caching comments and clarify usage

* feat: exllamav3 dense-only support

* feat: exl3 MoE support

* chore: move SiluAndMul activation to SharedExperts outside of afmoe

* feat: Qwen3.5 support for exl3

* fix: don't compile CUTLASS W8A8 in the _C extension

* refactor: remove unused permute_cols function and its registration

* feat: add v1/tokenize and v1/detokenize endpoints

* feat: add /api/latest alias for Kobold API

* chore: make pre-commit pass

* fix: more pre-commit stuff

Signed-off-by: AlpinDale <alpindale@gmail.com>

* refactor: improve shellcheck script structure and readability

Signed-off-by: AlpinDale <alpindale@gmail.com>

* refactor: remove update-dockerfile-graph hook from pre-commit configuration

Signed-off-by: AlpinDale <alpindale@gmail.com>

---------

Signed-off-by: AlpinDale <alpindale@gmail.com>

Author: AlpinDale

.buildkite/check-wheel-size.py

@@ -0,0 +1,50 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the Aphrodite project
+
+import os
+import sys
+import zipfile
+
+# Read the APHRODITE_MAX_SIZE_MB environment variable, defaulting to 500 MiB
+# Note that we have 800 MiB quota, please use it wisely.
+# See https://github.com/pypi/support/issues/6326 .
+# Please also sync the value with the one in Dockerfile.
+APHRODITE_MAX_SIZE_MB = int(os.environ.get("APHRODITE_MAX_SIZE_MB", 500))
+
+
+def print_top_10_largest_files(zip_file):
+    """Print the top 10 largest files in the given zip file."""
+    with zipfile.ZipFile(zip_file, "r") as z:
+        file_sizes = [(f, z.getinfo(f).file_size) for f in z.namelist()]
+        file_sizes.sort(key=lambda x: x[1], reverse=True)
+        for f, size in file_sizes[:10]:
+            print(f"{f}: {size / (1024 * 1024):.2f} MBs uncompressed.")
+
+
+def check_wheel_size(directory):
+    """Check the size of .whl files in the given directory."""
+    for root, _, files in os.walk(directory):
+        for file_name in files:
+            if file_name.endswith(".whl"):
+                wheel_path = os.path.join(root, file_name)
+                wheel_size_mb = os.path.getsize(wheel_path) / (1024 * 1024)
+                if wheel_size_mb > APHRODITE_MAX_SIZE_MB:
+                    print(
+                        f"Not allowed: Wheel {wheel_path} is larger "
+                        f"({wheel_size_mb:.2f} MB) than the limit "
+                        f"({APHRODITE_MAX_SIZE_MB} MB)."
+                    )
+                    print_top_10_largest_files(wheel_path)
+                    return 1
+                else:
+                    print(f"Wheel {wheel_path} is within the allowed size ({wheel_size_mb:.2f} MB).")
+    return 0
+
+
+if __name__ == "__main__":
+    if len(sys.argv) < 2:
+        print("Usage: python check-wheel-size.py <directory>")
+        sys.exit(1)
+
+    directory = sys.argv[1]
+    sys.exit(check_wheel_size(directory))

.gitignore

@@ -215,4 +215,11 @@ shellcheck-stable/
 ep_kernels_workspace/
 CMakeUserPresets.json
 cmake-build*/
-aphrodite/aphrodite_flash_attn/*
+aphrodite/vllm_flash_attn/*
+!aphrodite/vllm_flash_attn/__init__.py
+!aphrodite/vllm_flash_attn/flash_attn_interface.py
+
+aphrodite/third_party/triton_kernels/*
+aphrodite/third_party/deep_gemm/
+
+.triton

.markdownlint.yaml

@@ -0,0 +1,17 @@
+MD007:
+  indent: 4
+MD013: false
+MD024:
+  siblings_only: true
+MD031:
+  list_items: false
+MD033: false
+MD040: false
+MD046: false
+MD052: false
+MD059: false
+MD001: false
+MD025: false
+MD041: false
+MD045: false
+MD060: false

.pre-commit-config.yaml

@@ -4,7 +4,7 @@ default_install_hook_types:
 default_stages:
   - pre-commit # Run locally
   - manual # Run in CI
-exclude: 'aphrodite/third_party/.*'
+exclude: '^(reference/|shellcheck-stable/|aphrodite/third_party/.*)'
 repos:
 - repo: https://github.com/astral-sh/ruff-pre-commit
   rev: v0.14.0
@@ -13,7 +13,7 @@ repos:
     args: [--output-format, github, --fix]
   - id: ruff-format
 - repo: https://github.com/crate-ci/typos
-  rev: v1.38.1
+  rev: v1.43.5
   hooks:
   - id: typos
     exclude: 'kernels/xqa/cubin/.*'
@@ -25,18 +25,68 @@ repos:
     exclude: 'kernels/(moe/topk_softmax_kernels.cu|quantization/gguf/(ggml-common.h|dequantize.cuh|vecdotq.cuh|mmq.cuh|mmvq.cuh)|flash_mla|hadamard|backup|vmm|xqa)|aphrodite/third_party/.*'
     types_or: [c++, cuda]
     args: [--style=file, --verbose]
+- repo: https://github.com/DavidAnson/markdownlint-cli2
+  rev: v0.21.0
+  hooks:
+  - id: markdownlint-cli2
+    language_version: lts
+    args: [--fix]
+    exclude: ^CLAUDE\.md$
 - repo: https://github.com/rhysd/actionlint
   rev: v1.7.7
   hooks:
   - id: actionlint
 - repo: https://github.com/astral-sh/uv-pre-commit
-  rev: 0.9.1
+  rev: 0.11.1
   hooks:
     - id: pip-compile
-      args: [requirements/test.in, -o, requirements/test.txt, --index-strategy, unsafe-best-match, --torch-backend, cu129, --python-platform, x86_64-manylinux_2_28]
-      files: ^requirements/test\.(in|txt)$
+      args: [
+        requirements/test.in,
+        -c, requirements/cuda.txt,
+        -o, requirements/test.txt,
+        --index-strategy, unsafe-best-match,
+        --torch-backend, cu129,
+        --python-platform, x86_64-manylinux_2_28,
+        --python-version, "3.12",
+      ]
+      files: ^requirements/(common|cuda|test)\.(in|txt)$
 - repo: local
   hooks:
+  - id: mypy-local
+    name: Run mypy locally for lowest supported Python version
+    entry: python tools/pre_commit/mypy.py 0 "3.10"
+    stages: [pre-commit] # Don't run in CI
+    <<: &mypy_common
+      language: python
+      types_or: [python, pyi]
+      require_serial: true
+      additional_dependencies: ["mypy[faster-cache]==1.19.1", regex, types-cachetools, types-setuptools, types-PyYAML, types-requests, types-torch, pydantic]
+  - id: mypy-3.10
+    name: Run mypy for Python 3.10
+    entry: python tools/pre_commit/mypy.py 1 "3.10"
+    <<: *mypy_common
+    stages: [manual]
+  - id: mypy-3.11
+    name: Run mypy for Python 3.11
+    entry: python tools/pre_commit/mypy.py 1 "3.11"
+    <<: *mypy_common
+    stages: [manual]
+  - id: mypy-3.12
+    name: Run mypy for Python 3.12
+    entry: python tools/pre_commit/mypy.py 1 "3.12"
+    <<: *mypy_common
+    stages: [manual]
+  - id: mypy-3.13
+    name: Run mypy for Python 3.13
+    entry: python tools/pre_commit/mypy.py 1 "3.13"
+    <<: *mypy_common
+    stages: [manual]
+  - id: shellcheck
+    name: Lint shell scripts
+    entry: tools/pre_commit/shellcheck.sh
+    language: script
+    types: [shell]
+    files: ^tools/pre_commit/.*\.sh$
   - id: png-lint
     name: Lint PNG exports from excalidraw
     entry: tools/pre_commit/png-lint.sh
@@ -54,6 +104,12 @@ repos:
     language: system
     verbose: true
     stages: [commit-msg]
+  - id: check-spdx-header
+    name: Check SPDX headers
+    entry: python tools/pre_commit/check_spdx_header.py
+    language: python
+    types: [python]
+    exclude: ^(reference/|aphrodite/vllm_flash_attn/|aphrodite/third_party/|tests/|examples/|benchmarks/)
   - id: check-root-lazy-imports
     name: Check root lazy imports
     entry: python tools/pre_commit/check_init_lazy_imports.py
@@ -75,7 +131,6 @@ repos:
     types: [python]
     pass_filenames: false
     additional_dependencies: [regex]
-  # forbid directly import triton
   - id: forbid-direct-triton-import
     name: "Forbid direct 'import triton'"
     entry: python tools/pre_commit/check_triton_import.py
@@ -90,11 +145,29 @@ repos:
     types: [python]
     additional_dependencies: [regex]
     exclude: ^benchmarks/
+  - id: check-torch-cuda-call
+    name: "Prevent new 'torch.cuda' APIs call"
+    entry: python tools/pre_commit/check_torch_cuda.py
+    language: python
+    types: [python]
+    additional_dependencies: [regex]
   - id: validate-config
     name: Validate configuration has default values and that each field has a docstring
     entry: python tools/pre_commit/validate_config.py
     language: python
     additional_dependencies: [regex]
+  - id: validate-docker-versions
+    name: Validate docker/versions.json matches Dockerfile
+    entry: python tools/generate_versions_json.py --check
+    language: python
+    files: ^docker/(Dockerfile|versions\.json)$
+    pass_filenames: false
+    additional_dependencies: [dockerfile-parse]
+  - id: check-boolean-context-manager
+    name: Check for boolean ops in with-statements
+    entry: python tools/pre_commit/check_boolean_context_manager.py
+    language: python
+    types: [python]
   # Keep `suggestion` last
   - id: suggestion
     name: Suggestion