mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-02-12 14:03:48 +08:00
Merge branch 'main' into fix_spec_gate
Signed-off-by: Zheyu Fu <zheyuf@nvidia.com>
This commit is contained in:
Zheyu Fu
GitHub
commit
9dcea00d91
@ -916,13 +916,13 @@ typing = ["types-PyYAML", "types-simplejson", "types-toml", "types-tqdm", "types
|
||||
|
||||
[[package]]
|
||||
name = "humanize"
|
||||
version = "4.14.0"
|
||||
version = "4.15.0"
|
||||
description = "Python humanize utilities"
|
||||
optional = false
|
||||
python-versions = ">=3.10"
|
||||
files = [
|
||||
{file = "humanize-4.14.0-py3-none-any.whl", hash = "sha256:d57701248d040ad456092820e6fde56c930f17749956ac47f4f655c0c547bfff"},
|
||||
{file = "humanize-4.14.0.tar.gz", hash = "sha256:2fa092705ea640d605c435b1ca82b2866a1b601cdf96f076d70b79a855eba90d"},
|
||||
{file = "humanize-4.15.0-py3-none-any.whl", hash = "sha256:b1186eb9f5a9749cd9cb8565aee77919dd7c8d076161cf44d70e59e3301e1769"},
|
||||
{file = "humanize-4.15.0.tar.gz", hash = "sha256:1dd098483eb1c7ee8e32eb2e99ad1910baefa4b75c3aff3a82f4d78688993b10"},
|
||||
]
|
||||
|
||||
[package.extras]
|
||||
|
||||
@ -419,13 +419,13 @@ files = [
|
||||
|
||||
[[package]]
|
||||
name = "click"
|
||||
version = "8.2.1"
|
||||
version = "8.3.1"
|
||||
description = "Composable command line interface toolkit"
|
||||
optional = false
|
||||
python-versions = ">=3.10"
|
||||
files = [
|
||||
{file = "click-8.2.1-py3-none-any.whl", hash = "sha256:61a3265b914e850b85317d0b3109c7f8cd35a670f963866005d6ef1d5175a12b"},
|
||||
{file = "click-8.2.1.tar.gz", hash = "sha256:27c491cc05d968d271d5a1db13e3b5a184636d9d930f148c50b038f0d0646202"},
|
||||
{file = "click-8.3.1-py3-none-any.whl", hash = "sha256:981153a64e25f12d547d3426c367a4857371575ee7ad18df2a6183ab0545b2a6"},
|
||||
{file = "click-8.3.1.tar.gz", hash = "sha256:12ff4785d337a1bb490bb7e9c2b1ee5da3112e94a8622f26a6c77f5d2fc6842a"},
|
||||
]
|
||||
|
||||
[package.dependencies]
|
||||
@ -1511,56 +1511,56 @@ files = [
|
||||
|
||||
[[package]]
|
||||
name = "ray"
|
||||
version = "2.52.1"
|
||||
version = "2.53.0"
|
||||
description = "Ray provides a simple, universal API for building distributed applications."
|
||||
optional = false
|
||||
python-versions = ">=3.9"
|
||||
files = [
|
||||
{file = "ray-2.52.1-cp310-cp310-macosx_12_0_arm64.whl", hash = "sha256:993194a8be70540e0f819862031bbf19a64401fbe6c31b42065fd313ba466d34"},
|
||||
{file = "ray-2.52.1-cp310-cp310-manylinux2014_aarch64.whl", hash = "sha256:65bf461fdfe4ffa667c46f9455f8740b2ad6c1fa471b461d5f5cf6b7baf177b5"},
|
||||
{file = "ray-2.52.1-cp310-cp310-manylinux2014_x86_64.whl", hash = "sha256:b3f9e61b799fb3cc8fd7077a3d2eb676ddfef7db644f6b6a2b657c5c3214cf19"},
|
||||
{file = "ray-2.52.1-cp310-cp310-win_amd64.whl", hash = "sha256:24694e60cdc7770b90f123cc578cabb9d1a231c1fe673b5da0027b118de45846"},
|
||||
{file = "ray-2.52.1-cp311-cp311-macosx_12_0_arm64.whl", hash = "sha256:f59e3b2d1a1466ac0778f2c6fac9ccb5f30107d77e3dddd1d60167248d268474"},
|
||||
{file = "ray-2.52.1-cp311-cp311-manylinux2014_aarch64.whl", hash = "sha256:2b57ef272a2a0a0dbae6d18d70aa541eab620b4fe3b44d50466d3a533c16f9d9"},
|
||||
{file = "ray-2.52.1-cp311-cp311-manylinux2014_x86_64.whl", hash = "sha256:a5a3c268d45060c50cd029979ecc5f1eaaec040b19fa88dd4fe9e927d19ff13e"},
|
||||
{file = "ray-2.52.1-cp311-cp311-win_amd64.whl", hash = "sha256:4e8478544fef69a17d865431c0bebdcfeff7c0f76a306f29b73c3bc3cbb0bdb9"},
|
||||
{file = "ray-2.52.1-cp312-cp312-macosx_12_0_arm64.whl", hash = "sha256:6831592fedf0a122016f5dab4b67d85fa3d4db3b21f588d18834b5c031396d1c"},
|
||||
{file = "ray-2.52.1-cp312-cp312-manylinux2014_aarch64.whl", hash = "sha256:08eb8f5fd55292ba6bee363a32491136a5e54af54e007f81e0603986fbea41a4"},
|
||||
{file = "ray-2.52.1-cp312-cp312-manylinux2014_x86_64.whl", hash = "sha256:843c0108ad72bb7fc6c23a22e29e6099546a5eaad3ad675c78a146d9080f6ec6"},
|
||||
{file = "ray-2.52.1-cp312-cp312-win_amd64.whl", hash = "sha256:8045172ad3fcff62b9dab9a4cd2e0991ad0e27fc814fe625a8d3a120306651d6"},
|
||||
{file = "ray-2.52.1-cp313-cp313-macosx_12_0_arm64.whl", hash = "sha256:b5bc29548abb0a0a7ae9e6ff3b0ccca2824edaf011a4336e15a32793d574fbfd"},
|
||||
{file = "ray-2.52.1-cp313-cp313-manylinux2014_aarch64.whl", hash = "sha256:e3826aeb4e4399de0c6885bd8be7ce2f629fa0010f0013f1183e0726b3d25e40"},
|
||||
{file = "ray-2.52.1-cp313-cp313-manylinux2014_x86_64.whl", hash = "sha256:bbe492c780a39a64bd3d0766cad10d54cf12222df88d287ec2d8f2d52de37c79"},
|
||||
{file = "ray-2.53.0-cp310-cp310-macosx_12_0_arm64.whl", hash = "sha256:4db914a0a6dd608fa49c066929a1282745a2dbd73caee67d7b80fe684ca65bdd"},
|
||||
{file = "ray-2.53.0-cp310-cp310-manylinux2014_aarch64.whl", hash = "sha256:4108280d8a1cb90d7d68e5c954c35e63b8bb9a4ba15f88c5e7da0e2025647712"},
|
||||
{file = "ray-2.53.0-cp310-cp310-manylinux2014_x86_64.whl", hash = "sha256:4dbb5fce1364763f29741055f50abe33cf726397141f9cc0e845dd3cc963e455"},
|
||||
{file = "ray-2.53.0-cp310-cp310-win_amd64.whl", hash = "sha256:90faf630d20b6abf3135997fb3edb5842134aff92e04ee709865db04816d97ef"},
|
||||
{file = "ray-2.53.0-cp311-cp311-macosx_12_0_arm64.whl", hash = "sha256:bd3ec4c342776ddac23ae2b108c64f5939f417ccc4875900d586c7c978463269"},
|
||||
{file = "ray-2.53.0-cp311-cp311-manylinux2014_aarch64.whl", hash = "sha256:a0bbb98b0b0f25a3ee075ca10171e1260e70b6bc690cd509ecd7ce1228af854d"},
|
||||
{file = "ray-2.53.0-cp311-cp311-manylinux2014_x86_64.whl", hash = "sha256:eb000c17f7301071fdd15c44c4cd3ac0f7953bb4c7c227e61719fe7048195bcd"},
|
||||
{file = "ray-2.53.0-cp311-cp311-win_amd64.whl", hash = "sha256:4a1bb3fe09ab4cd0d16ddc96b9f60c9ed83b3f93b87aa8506e0d3b746fd4e825"},
|
||||
{file = "ray-2.53.0-cp312-cp312-macosx_12_0_arm64.whl", hash = "sha256:d8b95d047d947493803fb8417aea31225dcacdab15afdc75b8a238901949d457"},
|
||||
{file = "ray-2.53.0-cp312-cp312-manylinux2014_aarch64.whl", hash = "sha256:65e2ce58d3dc6baa3cf45824d889c1968ebde565ee54dfd80a98af8f31af8e4a"},
|
||||
{file = "ray-2.53.0-cp312-cp312-manylinux2014_x86_64.whl", hash = "sha256:14f46363e9b4cf0c1c8b4d8623ec337c5bd408377831b5e5b50067930137bbca"},
|
||||
{file = "ray-2.53.0-cp312-cp312-win_amd64.whl", hash = "sha256:b828c147f9ff2f277b1d254e4fe9a746fdfaee7e313a93a97c7edf4dae9b81a4"},
|
||||
{file = "ray-2.53.0-cp313-cp313-macosx_12_0_arm64.whl", hash = "sha256:85b472ab6fb8f1189f8cef81913fd91b24dd69b3fa7dcca7e144827bd924f6c0"},
|
||||
{file = "ray-2.53.0-cp313-cp313-manylinux2014_aarch64.whl", hash = "sha256:7196e5358dfcc8211be864f45e6dfe4827202df294af3c7a76ff8fbc080e0522"},
|
||||
{file = "ray-2.53.0-cp313-cp313-manylinux2014_x86_64.whl", hash = "sha256:73dbbaa7962a7f5e38aa8cf9483e0e9817205e989aa3dc859c738c2af1ae01df"},
|
||||
]
|
||||
|
||||
[package.dependencies]
|
||||
click = ">=7.0,<8.3.dev0 || >=8.4.dev0"
|
||||
click = ">=7.0"
|
||||
filelock = "*"
|
||||
jsonschema = "*"
|
||||
msgpack = ">=1.0.0,<2.0.0"
|
||||
packaging = "*"
|
||||
packaging = ">=24.2"
|
||||
protobuf = ">=3.20.3"
|
||||
pyyaml = "*"
|
||||
requests = "*"
|
||||
|
||||
[package.extras]
|
||||
adag = ["cupy-cuda12x"]
|
||||
air = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pyyaml", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all-cpp = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pyyaml", "ray-cpp (==2.52.1)", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
air = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pyyaml", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all-cpp = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pyyaml", "ray-cpp (==2.53.0)", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
cgraph = ["cupy-cuda12x"]
|
||||
client = ["grpcio", "grpcio (!=1.56.0)"]
|
||||
cpp = ["ray-cpp (==2.52.1)"]
|
||||
cpp = ["ray-cpp (==2.53.0)"]
|
||||
data = ["fsspec", "numpy (>=1.20)", "pandas (>=1.3)", "pyarrow (>=9.0.0)"]
|
||||
default = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "virtualenv (>=20.0.24,!=20.21.1)"]
|
||||
llm = ["aiohttp (>=3.7)", "aiohttp_cors", "async-timeout", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "hf_transfer", "jsonref (>=1.1.0)", "jsonschema", "meson", "ninja", "nixl (>=0.6.1)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pybind11", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "typer", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "vllm[audio] (>=0.11.0)", "watchfiles"]
|
||||
default = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "virtualenv (>=20.0.24,!=20.21.1)"]
|
||||
llm = ["aiohttp (>=3.7)", "aiohttp_cors", "async-timeout", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "hf_transfer", "jsonref (>=1.1.0)", "jsonschema", "meson", "ninja", "nixl (>=0.6.1)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pybind11", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "transformers (>=4.57.3)", "typer", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "vllm[audio] (>=0.12.0)", "watchfiles"]
|
||||
observability = ["memray"]
|
||||
rllib = ["dm_tree", "fsspec", "gymnasium (==1.1.1)", "lz4", "ormsgpack (==1.7.0)", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pyyaml", "requests", "scipy", "tensorboardX (>=1.9)"]
|
||||
serve = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-async-inference = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-grpc = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
train = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
tune = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
rllib = ["dm_tree", "fsspec", "gymnasium (==1.1.1)", "lz4", "ormsgpack (==1.7.0)", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pyyaml", "requests", "scipy", "tensorboardX (>=1.9)"]
|
||||
serve = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-async-inference = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-grpc = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
train = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
tune = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
|
||||
[[package]]
|
||||
name = "referencing"
|
||||
@ -2343,4 +2343,4 @@ dev = ["pytest", "setuptools"]
|
||||
[metadata]
|
||||
lock-version = "2.0"
|
||||
python-versions = ">=3.10,<3.13"
|
||||
content-hash = "863af11dfe73e05a959f4ba4e34f744a9385580fd2ec1c220b01fd176e776056"
|
||||
content-hash = "1dd6c5ec9bea98c49f5a6401d4af6e4406a1940a1c91823b730838f9bdf18ac9"
|
||||
|
||||
@ -12,7 +12,7 @@ av = "12.3.0"
|
||||
rotary-embedding-torch = "0.5.3"
|
||||
fabric = "^3.2.2"
|
||||
contexttimer = "^0.3.3"
|
||||
ray = "^2.52.1"
|
||||
ray = "^2.53.0"
|
||||
protobuf = "^6.33.2"
|
||||
bitsandbytes = ">=0.39.0"
|
||||
rpyc = "6.0.0"
|
||||
|
||||
@ -907,13 +907,13 @@ typing = ["types-PyYAML", "types-simplejson", "types-toml", "types-tqdm", "types
|
||||
|
||||
[[package]]
|
||||
name = "humanize"
|
||||
version = "4.14.0"
|
||||
version = "4.15.0"
|
||||
description = "Python humanize utilities"
|
||||
optional = false
|
||||
python-versions = ">=3.10"
|
||||
files = [
|
||||
{file = "humanize-4.14.0-py3-none-any.whl", hash = "sha256:d57701248d040ad456092820e6fde56c930f17749956ac47f4f655c0c547bfff"},
|
||||
{file = "humanize-4.14.0.tar.gz", hash = "sha256:2fa092705ea640d605c435b1ca82b2866a1b601cdf96f076d70b79a855eba90d"},
|
||||
{file = "humanize-4.15.0-py3-none-any.whl", hash = "sha256:b1186eb9f5a9749cd9cb8565aee77919dd7c8d076161cf44d70e59e3301e1769"},
|
||||
{file = "humanize-4.15.0.tar.gz", hash = "sha256:1dd098483eb1c7ee8e32eb2e99ad1910baefa4b75c3aff3a82f4d78688993b10"},
|
||||
]
|
||||
|
||||
[package.extras]
|
||||
|
||||
@ -652,25 +652,25 @@ test = ["pytest (>=6)"]
|
||||
|
||||
[[package]]
|
||||
name = "fastapi"
|
||||
version = "0.125.0"
|
||||
version = "0.126.0"
|
||||
description = "FastAPI framework, high performance, easy to learn, fast to code, ready for production"
|
||||
optional = false
|
||||
python-versions = ">=3.9"
|
||||
files = [
|
||||
{file = "fastapi-0.125.0-py3-none-any.whl", hash = "sha256:2570ec4f3aecf5cca8f0428aed2398b774fcdfee6c2116f86e80513f2f86a7a1"},
|
||||
{file = "fastapi-0.125.0.tar.gz", hash = "sha256:16b532691a33e2c5dee1dac32feb31dc6eb41a3dd4ff29a95f9487cb21c054c0"},
|
||||
{file = "fastapi-0.126.0-py3-none-any.whl", hash = "sha256:c9330b9731e3bd2caae0a00e76353f86adbf592c5a25649a1682f3a92aeaff41"},
|
||||
{file = "fastapi-0.126.0.tar.gz", hash = "sha256:f099fceb2a6d56dd21c59c4543d00be123dedacff869e76ae31ba3c0f963e2cd"},
|
||||
]
|
||||
|
||||
[package.dependencies]
|
||||
annotated-doc = ">=0.0.2"
|
||||
pydantic = ">=1.7.4,<1.8 || >1.8,<1.8.1 || >1.8.1,<2.0.0 || >2.0.0,<2.0.1 || >2.0.1,<2.1.0 || >2.1.0,<3.0.0"
|
||||
pydantic = ">=2.7.0"
|
||||
starlette = ">=0.40.0,<0.51.0"
|
||||
typing-extensions = ">=4.8.0"
|
||||
|
||||
[package.extras]
|
||||
all = ["email-validator (>=2.0.0)", "fastapi-cli[standard] (>=0.0.8)", "httpx (>=0.23.0,<1.0.0)", "itsdangerous (>=1.1.0)", "jinja2 (>=3.1.5)", "orjson (>=3.2.1)", "pydantic-extra-types (>=2.0.0)", "pydantic-settings (>=2.0.0)", "python-multipart (>=0.0.18)", "pyyaml (>=5.3.1)", "ujson (>=4.0.1,!=4.0.2,!=4.1.0,!=4.2.0,!=4.3.0,!=5.0.0,!=5.1.0)", "uvicorn[standard] (>=0.12.0)"]
|
||||
standard = ["email-validator (>=2.0.0)", "fastapi-cli[standard] (>=0.0.8)", "httpx (>=0.23.0,<1.0.0)", "jinja2 (>=3.1.5)", "python-multipart (>=0.0.18)", "uvicorn[standard] (>=0.12.0)"]
|
||||
standard-no-fastapi-cloud-cli = ["email-validator (>=2.0.0)", "fastapi-cli[standard-no-fastapi-cloud-cli] (>=0.0.8)", "httpx (>=0.23.0,<1.0.0)", "jinja2 (>=3.1.5)", "python-multipart (>=0.0.18)", "uvicorn[standard] (>=0.12.0)"]
|
||||
standard = ["email-validator (>=2.0.0)", "fastapi-cli[standard] (>=0.0.8)", "httpx (>=0.23.0,<1.0.0)", "jinja2 (>=3.1.5)", "pydantic-extra-types (>=2.0.0)", "pydantic-settings (>=2.0.0)", "python-multipart (>=0.0.18)", "uvicorn[standard] (>=0.12.0)"]
|
||||
standard-no-fastapi-cloud-cli = ["email-validator (>=2.0.0)", "fastapi-cli[standard-no-fastapi-cloud-cli] (>=0.0.8)", "httpx (>=0.23.0,<1.0.0)", "jinja2 (>=3.1.5)", "pydantic-extra-types (>=2.0.0)", "pydantic-settings (>=2.0.0)", "python-multipart (>=0.0.18)", "uvicorn[standard] (>=0.12.0)"]
|
||||
|
||||
[[package]]
|
||||
name = "ffmpy"
|
||||
|
||||
@ -361,13 +361,13 @@ files = [
|
||||
|
||||
[[package]]
|
||||
name = "click"
|
||||
version = "8.2.1"
|
||||
version = "8.3.1"
|
||||
description = "Composable command line interface toolkit"
|
||||
optional = false
|
||||
python-versions = ">=3.10"
|
||||
files = [
|
||||
{file = "click-8.2.1-py3-none-any.whl", hash = "sha256:61a3265b914e850b85317d0b3109c7f8cd35a670f963866005d6ef1d5175a12b"},
|
||||
{file = "click-8.2.1.tar.gz", hash = "sha256:27c491cc05d968d271d5a1db13e3b5a184636d9d930f148c50b038f0d0646202"},
|
||||
{file = "click-8.3.1-py3-none-any.whl", hash = "sha256:981153a64e25f12d547d3426c367a4857371575ee7ad18df2a6183ab0545b2a6"},
|
||||
{file = "click-8.3.1.tar.gz", hash = "sha256:12ff4785d337a1bb490bb7e9c2b1ee5da3112e94a8622f26a6c77f5d2fc6842a"},
|
||||
]
|
||||
|
||||
[package.dependencies]
|
||||
@ -1598,32 +1598,32 @@ files = [
|
||||
|
||||
[[package]]
|
||||
name = "ray"
|
||||
version = "2.52.1"
|
||||
version = "2.53.0"
|
||||
description = "Ray provides a simple, universal API for building distributed applications."
|
||||
optional = false
|
||||
python-versions = ">=3.9"
|
||||
files = [
|
||||
{file = "ray-2.52.1-cp310-cp310-macosx_12_0_arm64.whl", hash = "sha256:993194a8be70540e0f819862031bbf19a64401fbe6c31b42065fd313ba466d34"},
|
||||
{file = "ray-2.52.1-cp310-cp310-manylinux2014_aarch64.whl", hash = "sha256:65bf461fdfe4ffa667c46f9455f8740b2ad6c1fa471b461d5f5cf6b7baf177b5"},
|
||||
{file = "ray-2.52.1-cp310-cp310-manylinux2014_x86_64.whl", hash = "sha256:b3f9e61b799fb3cc8fd7077a3d2eb676ddfef7db644f6b6a2b657c5c3214cf19"},
|
||||
{file = "ray-2.52.1-cp310-cp310-win_amd64.whl", hash = "sha256:24694e60cdc7770b90f123cc578cabb9d1a231c1fe673b5da0027b118de45846"},
|
||||
{file = "ray-2.52.1-cp311-cp311-macosx_12_0_arm64.whl", hash = "sha256:f59e3b2d1a1466ac0778f2c6fac9ccb5f30107d77e3dddd1d60167248d268474"},
|
||||
{file = "ray-2.52.1-cp311-cp311-manylinux2014_aarch64.whl", hash = "sha256:2b57ef272a2a0a0dbae6d18d70aa541eab620b4fe3b44d50466d3a533c16f9d9"},
|
||||
{file = "ray-2.52.1-cp311-cp311-manylinux2014_x86_64.whl", hash = "sha256:a5a3c268d45060c50cd029979ecc5f1eaaec040b19fa88dd4fe9e927d19ff13e"},
|
||||
{file = "ray-2.52.1-cp311-cp311-win_amd64.whl", hash = "sha256:4e8478544fef69a17d865431c0bebdcfeff7c0f76a306f29b73c3bc3cbb0bdb9"},
|
||||
{file = "ray-2.52.1-cp312-cp312-macosx_12_0_arm64.whl", hash = "sha256:6831592fedf0a122016f5dab4b67d85fa3d4db3b21f588d18834b5c031396d1c"},
|
||||
{file = "ray-2.52.1-cp312-cp312-manylinux2014_aarch64.whl", hash = "sha256:08eb8f5fd55292ba6bee363a32491136a5e54af54e007f81e0603986fbea41a4"},
|
||||
{file = "ray-2.52.1-cp312-cp312-manylinux2014_x86_64.whl", hash = "sha256:843c0108ad72bb7fc6c23a22e29e6099546a5eaad3ad675c78a146d9080f6ec6"},
|
||||
{file = "ray-2.52.1-cp312-cp312-win_amd64.whl", hash = "sha256:8045172ad3fcff62b9dab9a4cd2e0991ad0e27fc814fe625a8d3a120306651d6"},
|
||||
{file = "ray-2.52.1-cp313-cp313-macosx_12_0_arm64.whl", hash = "sha256:b5bc29548abb0a0a7ae9e6ff3b0ccca2824edaf011a4336e15a32793d574fbfd"},
|
||||
{file = "ray-2.52.1-cp313-cp313-manylinux2014_aarch64.whl", hash = "sha256:e3826aeb4e4399de0c6885bd8be7ce2f629fa0010f0013f1183e0726b3d25e40"},
|
||||
{file = "ray-2.52.1-cp313-cp313-manylinux2014_x86_64.whl", hash = "sha256:bbe492c780a39a64bd3d0766cad10d54cf12222df88d287ec2d8f2d52de37c79"},
|
||||
{file = "ray-2.53.0-cp310-cp310-macosx_12_0_arm64.whl", hash = "sha256:4db914a0a6dd608fa49c066929a1282745a2dbd73caee67d7b80fe684ca65bdd"},
|
||||
{file = "ray-2.53.0-cp310-cp310-manylinux2014_aarch64.whl", hash = "sha256:4108280d8a1cb90d7d68e5c954c35e63b8bb9a4ba15f88c5e7da0e2025647712"},
|
||||
{file = "ray-2.53.0-cp310-cp310-manylinux2014_x86_64.whl", hash = "sha256:4dbb5fce1364763f29741055f50abe33cf726397141f9cc0e845dd3cc963e455"},
|
||||
{file = "ray-2.53.0-cp310-cp310-win_amd64.whl", hash = "sha256:90faf630d20b6abf3135997fb3edb5842134aff92e04ee709865db04816d97ef"},
|
||||
{file = "ray-2.53.0-cp311-cp311-macosx_12_0_arm64.whl", hash = "sha256:bd3ec4c342776ddac23ae2b108c64f5939f417ccc4875900d586c7c978463269"},
|
||||
{file = "ray-2.53.0-cp311-cp311-manylinux2014_aarch64.whl", hash = "sha256:a0bbb98b0b0f25a3ee075ca10171e1260e70b6bc690cd509ecd7ce1228af854d"},
|
||||
{file = "ray-2.53.0-cp311-cp311-manylinux2014_x86_64.whl", hash = "sha256:eb000c17f7301071fdd15c44c4cd3ac0f7953bb4c7c227e61719fe7048195bcd"},
|
||||
{file = "ray-2.53.0-cp311-cp311-win_amd64.whl", hash = "sha256:4a1bb3fe09ab4cd0d16ddc96b9f60c9ed83b3f93b87aa8506e0d3b746fd4e825"},
|
||||
{file = "ray-2.53.0-cp312-cp312-macosx_12_0_arm64.whl", hash = "sha256:d8b95d047d947493803fb8417aea31225dcacdab15afdc75b8a238901949d457"},
|
||||
{file = "ray-2.53.0-cp312-cp312-manylinux2014_aarch64.whl", hash = "sha256:65e2ce58d3dc6baa3cf45824d889c1968ebde565ee54dfd80a98af8f31af8e4a"},
|
||||
{file = "ray-2.53.0-cp312-cp312-manylinux2014_x86_64.whl", hash = "sha256:14f46363e9b4cf0c1c8b4d8623ec337c5bd408377831b5e5b50067930137bbca"},
|
||||
{file = "ray-2.53.0-cp312-cp312-win_amd64.whl", hash = "sha256:b828c147f9ff2f277b1d254e4fe9a746fdfaee7e313a93a97c7edf4dae9b81a4"},
|
||||
{file = "ray-2.53.0-cp313-cp313-macosx_12_0_arm64.whl", hash = "sha256:85b472ab6fb8f1189f8cef81913fd91b24dd69b3fa7dcca7e144827bd924f6c0"},
|
||||
{file = "ray-2.53.0-cp313-cp313-manylinux2014_aarch64.whl", hash = "sha256:7196e5358dfcc8211be864f45e6dfe4827202df294af3c7a76ff8fbc080e0522"},
|
||||
{file = "ray-2.53.0-cp313-cp313-manylinux2014_x86_64.whl", hash = "sha256:73dbbaa7962a7f5e38aa8cf9483e0e9817205e989aa3dc859c738c2af1ae01df"},
|
||||
]
|
||||
|
||||
[package.dependencies]
|
||||
aiohttp = {version = ">=3.7", optional = true, markers = "extra == \"default\""}
|
||||
aiohttp_cors = {version = "*", optional = true, markers = "extra == \"default\""}
|
||||
click = ">=7.0,<8.3.dev0 || >=8.4.dev0"
|
||||
click = ">=7.0"
|
||||
colorful = {version = "*", optional = true, markers = "extra == \"default\""}
|
||||
filelock = "*"
|
||||
grpcio = {version = ">=1.42.0", optional = true, markers = "python_version >= \"3.10\" and extra == \"default\""}
|
||||
@ -1633,14 +1633,14 @@ opencensus = {version = "*", optional = true, markers = "extra == \"default\""}
|
||||
opentelemetry-exporter-prometheus = {version = "*", optional = true, markers = "extra == \"default\""}
|
||||
opentelemetry-proto = {version = "*", optional = true, markers = "extra == \"default\""}
|
||||
opentelemetry-sdk = {version = ">=1.30.0", optional = true, markers = "extra == \"default\""}
|
||||
packaging = "*"
|
||||
packaging = ">=24.2"
|
||||
prometheus_client = {version = ">=0.7.1", optional = true, markers = "extra == \"default\""}
|
||||
protobuf = ">=3.20.3"
|
||||
py-spy = [
|
||||
{version = ">=0.2.0", optional = true, markers = "python_version < \"3.12\" and extra == \"default\""},
|
||||
{version = ">=0.4.0", optional = true, markers = "python_version >= \"3.12\" and extra == \"default\""},
|
||||
]
|
||||
pydantic = {version = "<2.0.dev0 || >=2.5.dev0,<3", optional = true, markers = "extra == \"default\""}
|
||||
pydantic = {version = "<2.0.dev0 || >=2.12.dev0,<3", optional = true, markers = "extra == \"default\""}
|
||||
pyyaml = "*"
|
||||
requests = "*"
|
||||
smart_open = {version = "*", optional = true, markers = "extra == \"default\""}
|
||||
@ -1648,22 +1648,22 @@ virtualenv = {version = ">=20.0.24,<20.21.1 || >20.21.1", optional = true, marke
|
||||
|
||||
[package.extras]
|
||||
adag = ["cupy-cuda12x"]
|
||||
air = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pyyaml", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all-cpp = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pyyaml", "ray-cpp (==2.52.1)", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
air = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pyyaml", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
all-cpp = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "cupy-cuda12x", "dm_tree", "fastapi", "fsspec", "grpcio", "grpcio (!=1.56.0)", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "gymnasium (==1.1.1)", "lz4", "memray", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "ormsgpack (==1.7.0)", "pandas", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pyyaml", "ray-cpp (==2.53.0)", "requests", "scipy", "smart_open", "starlette", "tensorboardX (>=1.9)", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
cgraph = ["cupy-cuda12x"]
|
||||
client = ["grpcio", "grpcio (!=1.56.0)"]
|
||||
cpp = ["ray-cpp (==2.52.1)"]
|
||||
cpp = ["ray-cpp (==2.53.0)"]
|
||||
data = ["fsspec", "numpy (>=1.20)", "pandas (>=1.3)", "pyarrow (>=9.0.0)"]
|
||||
default = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "virtualenv (>=20.0.24,!=20.21.1)"]
|
||||
llm = ["aiohttp (>=3.7)", "aiohttp_cors", "async-timeout", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "hf_transfer", "jsonref (>=1.1.0)", "jsonschema", "meson", "ninja", "nixl (>=0.6.1)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pybind11", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "typer", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "vllm[audio] (>=0.11.0)", "watchfiles"]
|
||||
default = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "virtualenv (>=20.0.24,!=20.21.1)"]
|
||||
llm = ["aiohttp (>=3.7)", "aiohttp_cors", "async-timeout", "colorful", "fastapi", "fsspec", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "hf_transfer", "jsonref (>=1.1.0)", "jsonschema", "meson", "ninja", "nixl (>=0.6.1)", "numpy (>=1.20)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "pandas (>=1.3)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyarrow (>=9.0.0)", "pybind11", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "transformers (>=4.57.3)", "typer", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "vllm[audio] (>=0.12.0)", "watchfiles"]
|
||||
observability = ["memray"]
|
||||
rllib = ["dm_tree", "fsspec", "gymnasium (==1.1.1)", "lz4", "ormsgpack (==1.7.0)", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pyyaml", "requests", "scipy", "tensorboardX (>=1.9)"]
|
||||
serve = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-async-inference = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-grpc = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
train = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
tune = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.5.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
rllib = ["dm_tree", "fsspec", "gymnasium (==1.1.1)", "lz4", "ormsgpack (==1.7.0)", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pyyaml", "requests", "scipy", "tensorboardX (>=1.9)"]
|
||||
serve = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-async-inference = ["aiohttp (>=3.7)", "aiohttp_cors", "celery", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
serve-grpc = ["aiohttp (>=3.7)", "aiohttp_cors", "colorful", "fastapi", "grpcio (>=1.32.0)", "grpcio (>=1.42.0)", "opencensus", "opentelemetry-exporter-prometheus", "opentelemetry-proto", "opentelemetry-sdk (>=1.30.0)", "prometheus_client (>=0.7.1)", "py-spy (>=0.2.0)", "py-spy (>=0.4.0)", "pyOpenSSL", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "smart_open", "starlette", "uvicorn[standard]", "virtualenv (>=20.0.24,!=20.21.1)", "watchfiles"]
|
||||
train = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
tune = ["fsspec", "pandas", "pyarrow (>=9.0.0)", "pydantic (<2.0.dev0 || >=2.12.dev0,<3)", "requests", "tensorboardX (>=1.9)"]
|
||||
|
||||
[[package]]
|
||||
name = "referencing"
|
||||
@ -2224,4 +2224,4 @@ type = ["pytest-mypy"]
|
||||
[metadata]
|
||||
lock-version = "2.0"
|
||||
python-versions = ">=3.10,<3.13"
|
||||
content-hash = "64bb4e78ac900350b3ff15fa3d91a3c2ad3bdb948044976bb6c9f05018eaa18b"
|
||||
content-hash = "89165db124a381d5ce9e301a6e6d062091a9c415d6edf2ec6f4c61a730190708"
|
||||
|
||||
@ -7,7 +7,7 @@ readme = "README.md"
|
||||
|
||||
[tool.poetry.dependencies]
|
||||
python = ">=3.10,<3.13"
|
||||
ray = {extras = ["default"], version = "^2.52.1"}
|
||||
ray = {extras = ["default"], version = "^2.53.0"}
|
||||
|
||||
|
||||
[build-system]
|
||||
|
||||
@ -1,4 +1,4 @@
|
||||
{
|
||||
"commit_hash": "3b3069b39045fc7371cd6b533fd3c1abc881236f",
|
||||
"timestamp": "2025-12-20T02:39:22Z"
|
||||
"commit_hash": "77e37d9dd08665f9dce155f5884501791078eac2",
|
||||
"timestamp": "2025-12-21T02:39:22Z"
|
||||
}
|
||||
|
||||
@ -219,7 +219,7 @@ class TestPhi4MMFusedVisionLora(LlmapiAccuracyTestHarness):
|
||||
class TestGemma3_27BInstruct(LlmapiAccuracyTestHarness):
|
||||
MODEL_NAME = "google/gemma-3-27b-it"
|
||||
MODEL_PATH = f"{llm_models_root()}/gemma/gemma-3-27b-it/"
|
||||
MAX_NUM_TOKENS = 25600
|
||||
MAX_NUM_TOKENS = 12800
|
||||
|
||||
sampling_params = SamplingParams(
|
||||
max_tokens=MAX_NUM_TOKENS, truncate_prompt_tokens=MMMU.MAX_INPUT_LEN, stop="<end_of_turn>"
|
||||
@ -229,7 +229,7 @@ class TestGemma3_27BInstruct(LlmapiAccuracyTestHarness):
|
||||
kv_cache_config = KvCacheConfig(
|
||||
enable_block_reuse=False,
|
||||
enable_partial_reuse=False,
|
||||
free_gpu_memory_fraction=0.6,
|
||||
free_gpu_memory_fraction=0.4,
|
||||
)
|
||||
|
||||
def test_auto_dtype(self):
|
||||
|
||||
@ -47,7 +47,6 @@ l0_h100:
|
||||
- accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype_vswa_reuse_low_memory_available_no_partial_reuse
|
||||
- accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype_vswa_reuse_low_memory_available_partial_reuse
|
||||
- accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype_vswa_without_reuse_disable_overlap_scheduler
|
||||
- accuracy/test_llm_api_pytorch_multimodal.py::TestGemma3_27BInstruct::test_auto_dtype
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16[attn_backend=TRTLLM-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16[attn_backend=TRTLLM-torch_compile=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_chunked_prefill[attn_backend=TRTLLM] TIMEOUT (90)
|
||||
@ -261,6 +260,7 @@ l0_h100:
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_fp8_block_scales[mtp=vanilla-fp8kv=False-attention_dp=False-cuda_graph=False-overlap_scheduler=False-torch_compile=False]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_no_kv_cache_reuse[quant_dtype=none-mtp_nextn=2-fp8kv=False-attention_dp=True-cuda_graph=True-overlap_scheduler=True]
|
||||
- accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_guided_decoding[xgrammar-mtp_nextn=0]
|
||||
- accuracy/test_llm_api_pytorch_multimodal.py::TestGemma3_27BInstruct::test_auto_dtype
|
||||
- accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_fp8_prequantized
|
||||
- accuracy/test_llm_api_pytorch.py::TestGemma3_27BInstruct::test_fp8_prequantized
|
||||
- accuracy/test_llm_api_pytorch.py::TestGemma3_27BInstruct::test_auto_dtype TIMEOUT (90)
|
||||
|
||||
@ -447,7 +447,6 @@ examples/test_phi.py::test_phi_fp8_with_bf16_lora[phi-2] SKIP (https://nvbugs/57
|
||||
examples/test_phi.py::test_llm_phi_1node_2gpus_summary[Phi-3.5-MoE-instruct-nb:1] SKIP (https://nvbugs/5744293)
|
||||
examples/test_phi.py::test_llm_phi_quantization_1gpu[phi-2-fp8-bfloat16] SKIP (https://nvbugs/5744293)
|
||||
disaggregated/test_disaggregated.py::test_disaggregated_trtllm_sampler[TinyLlama-1.1B-Chat-v1.0] SKIP (https://nvbugs/5741884)
|
||||
accuracy/test_llm_api_pytorch_multimodal.py::TestGemma3_27BInstruct::test_auto_dtype SKIP (https://nvbugs/5744427)
|
||||
test_e2e.py::test_trtllm_bench_llmapi_launch[pytorch_backend-llama-v3-llama3-8b] SKIP (https://nvbugs/5744432)
|
||||
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[tp2pp2-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True] SKIP (https://nvbugs/5740087)
|
||||
accuracy/test_llm_api_pytorch.py::TestDeepSeekV3Lite::test_bfloat16_4gpus[tp4-mtp_nextn=2-attention_dp=True-cuda_graph=True-overlap_scheduler=True-torch_compile=True] SKIP (https://nvbugs/5740075)
|
||||
@ -491,3 +490,4 @@ examples/test_whisper.py::test_llm_whisper_general[large-v3-disable_gemm_plugin-
|
||||
accuracy/test_llm_api_pytorch_multimodal.py::TestQwen3VL_MOE::test_auto_dtype SKIP (https://nvbugs/5588376)
|
||||
accuracy/test_disaggregated_serving.py::TestLlama3_1_8BInstruct::test_tp_pp_symmetric[MMLU-tp2pp2] SKIP (https://nvbugs/5756008)
|
||||
unittest/_torch/speculative/test_dynamic_spec_decode.py::test_dynamic_spec_decode SKIP (https://nvbugs/5758449)
|
||||
unittest/executor/test_base_worker.py::TestWorkerBase SKIP (https://nvbugs/5759698)
|
||||
|
||||
Loading…
Reference in New Issue
Block a user