minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[grpc] Auto-generate protobuf files during wheel build (#16409)

Browse Source
This commit is contained in:
Chang Su
2026-01-08 09:09:54 -08:00
committed by GitHub
parent cda35611d4
commit 16880235d1
6 changed files with 119 additions and 943 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.gitignore
+5
View File
@@ -261,3 +261,8 @@ outputs/
# setuptools-scm generated version file # setuptools-scm generated version file
python/sglang/_version.py python/sglang/_version.py
# Generated protobuf files (regenerate during wheel build or with compile_proto.py)
python/sglang/srt/grpc/*_pb2.py
python/sglang/srt/grpc/*_pb2_grpc.py
python/sglang/srt/grpc/*_pb2.pyi
python/pyproject.toml
+1 -1
View File
@@ -1,5 +1,5 @@
[build-system] [build-system]
requires = ["setuptools>=61.0", "setuptools-scm>=8.0", "wheel"] requires = ["setuptools>=61.0", "setuptools-scm>=8.0", "wheel", "grpcio-tools==1.75.1"]
build-backend = "setuptools.build_meta" build-backend = "setuptools.build_meta"
[project] [project]
python/setup.py
+113
View File
@@ -0,0 +1,113 @@
"""
Custom setup.py for SGLang that compiles protobuf files during build.
This file works alongside pyproject.toml. It hooks into the build process
to automatically generate gRPC/protobuf Python files from .proto sources
when building the wheel or doing editable installs.
"""
import subprocess
import sys
from pathlib import Path
from setuptools import setup
from setuptools.command.build_py import build_py
from setuptools.command.develop import develop
from setuptools.command.egg_info import egg_info
from setuptools.errors import SetupError
PROTO_SOURCE = "sglang/srt/grpc/sglang_scheduler.proto"
def compile_proto():
"""Compile the protobuf file to Python using grpc_tools.protoc."""
proto_path = Path(__file__).parent / PROTO_SOURCE
if not proto_path.exists():
print(f"Warning: Proto file not found at {proto_path}, skipping generation")
return
print(f"Generating gRPC files from {PROTO_SOURCE}")
output_dir = proto_path.parent
proto_dir = proto_path.parent
# Build the protoc command
cmd = [
sys.executable,
"-m",
"grpc_tools.protoc",
f"-I{proto_dir}",
f"--python_out={output_dir}",
f"--grpc_python_out={output_dir}",
f"--pyi_out={output_dir}",
proto_path.name,
]
print(f"Running: {' '.join(cmd)}")
try:
subprocess.run(
cmd,
capture_output=True,
text=True,
cwd=proto_dir,
check=True,
)
except subprocess.CalledProcessError as e:
error_msg = e.stderr or e.stdout or "Unknown error"
raise SetupError(f"protoc failed with exit code {e.returncode}: {error_msg}")
# Fix imports in generated grpc file (change absolute to relative imports)
_fix_imports(output_dir, proto_path.stem)
print(f"Successfully generated gRPC files in {output_dir}")
def _fix_imports(output_dir: Path, proto_stem: str):
"""Fix imports in generated files to use relative imports."""
grpc_file = output_dir / f"{proto_stem}_pb2_grpc.py"
if grpc_file.exists():
content = grpc_file.read_text()
# Change absolute import to relative import
old_import = f"import {proto_stem}_pb2"
new_import = f"from . import {proto_stem}_pb2"
if old_import in content:
content = content.replace(old_import, new_import)
grpc_file.write_text(content)
print("Fixed imports in generated gRPC file")
class BuildPyWithProto(build_py):
"""Build Python modules, generating gRPC files from .proto sources first."""
def run(self):
compile_proto()
super().run()
class DevelopWithProto(develop):
"""Editable install with gRPC file generation."""
def run(self):
compile_proto()
super().run()
class EggInfoWithProto(egg_info):
"""Egg info generation with gRPC file generation."""
def run(self):
compile_proto()
super().run()
setup(
cmdclass={
"build_py": BuildPyWithProto,
"develop": DevelopWithProto,
"egg_info": EggInfoWithProto,
},
)
python/sglang/srt/grpc/sglang_scheduler_pb2.py
-119
View File
File diff suppressed because one or more lines are too long
python/sglang/srt/grpc/sglang_scheduler_pb2.pyi
-496
View File
@@ -1,496 +0,0 @@
import datetime
from google.protobuf import timestamp_pb2 as _timestamp_pb2
from google.protobuf import struct_pb2 as _struct_pb2
from google.protobuf.internal import containers as _containers
from google.protobuf import descriptor as _descriptor
from google.protobuf import message as _message
from collections.abc import Iterable as _Iterable, Mapping as _Mapping
from typing import ClassVar as _ClassVar, Optional as _Optional, Union as _Union
DESCRIPTOR: _descriptor.FileDescriptor
class SamplingParams(_message.Message):
__slots__ = ("temperature", "top_p", "top_k", "min_p", "frequency_penalty", "presence_penalty", "repetition_penalty", "max_new_tokens", "stop", "stop_token_ids", "skip_special_tokens", "spaces_between_special_tokens", "regex", "json_schema", "ebnf_grammar", "structural_tag", "n", "min_new_tokens", "ignore_eos", "no_stop_trim", "stream_interval", "logit_bias", "custom_params")
class LogitBiasEntry(_message.Message):
__slots__ = ("key", "value")
KEY_FIELD_NUMBER: _ClassVar[int]
VALUE_FIELD_NUMBER: _ClassVar[int]
key: str
value: float
def __init__(self, key: _Optional[str] = ..., value: _Optional[float] = ...) -> None: ...
TEMPERATURE_FIELD_NUMBER: _ClassVar[int]
TOP_P_FIELD_NUMBER: _ClassVar[int]
TOP_K_FIELD_NUMBER: _ClassVar[int]
MIN_P_FIELD_NUMBER: _ClassVar[int]
FREQUENCY_PENALTY_FIELD_NUMBER: _ClassVar[int]
PRESENCE_PENALTY_FIELD_NUMBER: _ClassVar[int]
REPETITION_PENALTY_FIELD_NUMBER: _ClassVar[int]
MAX_NEW_TOKENS_FIELD_NUMBER: _ClassVar[int]
STOP_FIELD_NUMBER: _ClassVar[int]
STOP_TOKEN_IDS_FIELD_NUMBER: _ClassVar[int]
SKIP_SPECIAL_TOKENS_FIELD_NUMBER: _ClassVar[int]
SPACES_BETWEEN_SPECIAL_TOKENS_FIELD_NUMBER: _ClassVar[int]
REGEX_FIELD_NUMBER: _ClassVar[int]
JSON_SCHEMA_FIELD_NUMBER: _ClassVar[int]
EBNF_GRAMMAR_FIELD_NUMBER: _ClassVar[int]
STRUCTURAL_TAG_FIELD_NUMBER: _ClassVar[int]
N_FIELD_NUMBER: _ClassVar[int]
MIN_NEW_TOKENS_FIELD_NUMBER: _ClassVar[int]
IGNORE_EOS_FIELD_NUMBER: _ClassVar[int]
NO_STOP_TRIM_FIELD_NUMBER: _ClassVar[int]
STREAM_INTERVAL_FIELD_NUMBER: _ClassVar[int]
LOGIT_BIAS_FIELD_NUMBER: _ClassVar[int]
CUSTOM_PARAMS_FIELD_NUMBER: _ClassVar[int]
temperature: float
top_p: float
top_k: int
min_p: float
frequency_penalty: float
presence_penalty: float
repetition_penalty: float
max_new_tokens: int
stop: _containers.RepeatedScalarFieldContainer[str]
stop_token_ids: _containers.RepeatedScalarFieldContainer[int]
skip_special_tokens: bool
spaces_between_special_tokens: bool
regex: str
json_schema: str
ebnf_grammar: str
structural_tag: str
n: int
min_new_tokens: int
ignore_eos: bool
no_stop_trim: bool
stream_interval: int
logit_bias: _containers.ScalarMap[str, float]
custom_params: _struct_pb2.Struct
def __init__(self, temperature: _Optional[float] = ..., top_p: _Optional[float] = ..., top_k: _Optional[int] = ..., min_p: _Optional[float] = ..., frequency_penalty: _Optional[float] = ..., presence_penalty: _Optional[float] = ..., repetition_penalty: _Optional[float] = ..., max_new_tokens: _Optional[int] = ..., stop: _Optional[_Iterable[str]] = ..., stop_token_ids: _Optional[_Iterable[int]] = ..., skip_special_tokens: bool = ..., spaces_between_special_tokens: bool = ..., regex: _Optional[str] = ..., json_schema: _Optional[str] = ..., ebnf_grammar: _Optional[str] = ..., structural_tag: _Optional[str] = ..., n: _Optional[int] = ..., min_new_tokens: _Optional[int] = ..., ignore_eos: bool = ..., no_stop_trim: bool = ..., stream_interval: _Optional[int] = ..., logit_bias: _Optional[_Mapping[str, float]] = ..., custom_params: _Optional[_Union[_struct_pb2.Struct, _Mapping]] = ...) -> None: ...
class DisaggregatedParams(_message.Message):
__slots__ = ("bootstrap_host", "bootstrap_port", "bootstrap_room")
BOOTSTRAP_HOST_FIELD_NUMBER: _ClassVar[int]
BOOTSTRAP_PORT_FIELD_NUMBER: _ClassVar[int]
BOOTSTRAP_ROOM_FIELD_NUMBER: _ClassVar[int]
bootstrap_host: str
bootstrap_port: int
bootstrap_room: int
def __init__(self, bootstrap_host: _Optional[str] = ..., bootstrap_port: _Optional[int] = ..., bootstrap_room: _Optional[int] = ...) -> None: ...
class GenerateRequest(_message.Message):
__slots__ = ("request_id", "tokenized", "mm_inputs", "sampling_params", "return_logprob", "logprob_start_len", "top_logprobs_num", "token_ids_logprob", "return_hidden_states", "disaggregated_params", "custom_logit_processor", "timestamp", "log_metrics", "input_embeds", "lora_id", "data_parallel_rank", "stream")
REQUEST_ID_FIELD_NUMBER: _ClassVar[int]
TOKENIZED_FIELD_NUMBER: _ClassVar[int]
MM_INPUTS_FIELD_NUMBER: _ClassVar[int]
SAMPLING_PARAMS_FIELD_NUMBER: _ClassVar[int]
RETURN_LOGPROB_FIELD_NUMBER: _ClassVar[int]
LOGPROB_START_LEN_FIELD_NUMBER: _ClassVar[int]
TOP_LOGPROBS_NUM_FIELD_NUMBER: _ClassVar[int]
TOKEN_IDS_LOGPROB_FIELD_NUMBER: _ClassVar[int]
RETURN_HIDDEN_STATES_FIELD_NUMBER: _ClassVar[int]
DISAGGREGATED_PARAMS_FIELD_NUMBER: _ClassVar[int]
CUSTOM_LOGIT_PROCESSOR_FIELD_NUMBER: _ClassVar[int]
TIMESTAMP_FIELD_NUMBER: _ClassVar[int]
LOG_METRICS_FIELD_NUMBER: _ClassVar[int]
INPUT_EMBEDS_FIELD_NUMBER: _ClassVar[int]
LORA_ID_FIELD_NUMBER: _ClassVar[int]
DATA_PARALLEL_RANK_FIELD_NUMBER: _ClassVar[int]
STREAM_FIELD_NUMBER: _ClassVar[int]
request_id: str
tokenized: TokenizedInput
mm_inputs: MultimodalInputs
sampling_params: SamplingParams
return_logprob: bool
logprob_start_len: int
top_logprobs_num: int
token_ids_logprob: _containers.RepeatedScalarFieldContainer[int]
return_hidden_states: bool
disaggregated_params: DisaggregatedParams
custom_logit_processor: str
timestamp: _timestamp_pb2.Timestamp
log_metrics: bool
input_embeds: _containers.RepeatedScalarFieldContainer[float]
lora_id: str
data_parallel_rank: int
stream: bool
def __init__(self, request_id: _Optional[str] = ..., tokenized: _Optional[_Union[TokenizedInput, _Mapping]] = ..., mm_inputs: _Optional[_Union[MultimodalInputs, _Mapping]] = ..., sampling_params: _Optional[_Union[SamplingParams, _Mapping]] = ..., return_logprob: bool = ..., logprob_start_len: _Optional[int] = ..., top_logprobs_num: _Optional[int] = ..., token_ids_logprob: _Optional[_Iterable[int]] = ..., return_hidden_states: bool = ..., disaggregated_params: _Optional[_Union[DisaggregatedParams, _Mapping]] = ..., custom_logit_processor: _Optional[str] = ..., timestamp: _Optional[_Union[datetime.datetime, _timestamp_pb2.Timestamp, _Mapping]] = ..., log_metrics: bool = ..., input_embeds: _Optional[_Iterable[float]] = ..., lora_id: _Optional[str] = ..., data_parallel_rank: _Optional[int] = ..., stream: bool = ...) -> None: ...
class TokenizedInput(_message.Message):
__slots__ = ("original_text", "input_ids")
ORIGINAL_TEXT_FIELD_NUMBER: _ClassVar[int]
INPUT_IDS_FIELD_NUMBER: _ClassVar[int]
original_text: str
input_ids: _containers.RepeatedScalarFieldContainer[int]
def __init__(self, original_text: _Optional[str] = ..., input_ids: _Optional[_Iterable[int]] = ...) -> None: ...
class MultimodalInputs(_message.Message):
__slots__ = ("image_urls", "video_urls", "audio_urls", "processed_features", "image_data", "video_data", "audio_data", "modalities")
IMAGE_URLS_FIELD_NUMBER: _ClassVar[int]
VIDEO_URLS_FIELD_NUMBER: _ClassVar[int]
AUDIO_URLS_FIELD_NUMBER: _ClassVar[int]
PROCESSED_FEATURES_FIELD_NUMBER: _ClassVar[int]
IMAGE_DATA_FIELD_NUMBER: _ClassVar[int]
VIDEO_DATA_FIELD_NUMBER: _ClassVar[int]
AUDIO_DATA_FIELD_NUMBER: _ClassVar[int]
MODALITIES_FIELD_NUMBER: _ClassVar[int]
image_urls: _containers.RepeatedScalarFieldContainer[str]
video_urls: _containers.RepeatedScalarFieldContainer[str]
audio_urls: _containers.RepeatedScalarFieldContainer[str]
processed_features: _struct_pb2.Struct
image_data: _containers.RepeatedScalarFieldContainer[bytes]
video_data: _containers.RepeatedScalarFieldContainer[bytes]
audio_data: _containers.RepeatedScalarFieldContainer[bytes]
modalities: _containers.RepeatedScalarFieldContainer[str]
def __init__(self, image_urls: _Optional[_Iterable[str]] = ..., video_urls: _Optional[_Iterable[str]] = ..., audio_urls: _Optional[_Iterable[str]] = ..., processed_features: _Optional[_Union[_struct_pb2.Struct, _Mapping]] = ..., image_data: _Optional[_Iterable[bytes]] = ..., video_data: _Optional[_Iterable[bytes]] = ..., audio_data: _Optional[_Iterable[bytes]] = ..., modalities: _Optional[_Iterable[str]] = ...) -> None: ...
class GenerateResponse(_message.Message):
__slots__ = ("request_id", "chunk", "complete", "error")
REQUEST_ID_FIELD_NUMBER: _ClassVar[int]
CHUNK_FIELD_NUMBER: _ClassVar[int]
COMPLETE_FIELD_NUMBER: _ClassVar[int]
ERROR_FIELD_NUMBER: _ClassVar[int]
request_id: str
chunk: GenerateStreamChunk
complete: GenerateComplete
error: GenerateError
def __init__(self, request_id: _Optional[str] = ..., chunk: _Optional[_Union[GenerateStreamChunk, _Mapping]] = ..., complete: _Optional[_Union[GenerateComplete, _Mapping]] = ..., error: _Optional[_Union[GenerateError, _Mapping]] = ...) -> None: ...
class GenerateStreamChunk(_message.Message):
__slots__ = ("token_ids", "prompt_tokens", "completion_tokens", "cached_tokens", "output_logprobs", "hidden_states", "input_logprobs", "index")
TOKEN_IDS_FIELD_NUMBER: _ClassVar[int]
PROMPT_TOKENS_FIELD_NUMBER: _ClassVar[int]
COMPLETION_TOKENS_FIELD_NUMBER: _ClassVar[int]
CACHED_TOKENS_FIELD_NUMBER: _ClassVar[int]
OUTPUT_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
HIDDEN_STATES_FIELD_NUMBER: _ClassVar[int]
INPUT_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
INDEX_FIELD_NUMBER: _ClassVar[int]
token_ids: _containers.RepeatedScalarFieldContainer[int]
prompt_tokens: int
completion_tokens: int
cached_tokens: int
output_logprobs: OutputLogProbs
hidden_states: _containers.RepeatedScalarFieldContainer[float]
input_logprobs: InputLogProbs
index: int
def __init__(self, token_ids: _Optional[_Iterable[int]] = ..., prompt_tokens: _Optional[int] = ..., completion_tokens: _Optional[int] = ..., cached_tokens: _Optional[int] = ..., output_logprobs: _Optional[_Union[OutputLogProbs, _Mapping]] = ..., hidden_states: _Optional[_Iterable[float]] = ..., input_logprobs: _Optional[_Union[InputLogProbs, _Mapping]] = ..., index: _Optional[int] = ...) -> None: ...
class GenerateComplete(_message.Message):
__slots__ = ("output_ids", "finish_reason", "prompt_tokens", "completion_tokens", "cached_tokens", "output_logprobs", "all_hidden_states", "matched_token_id", "matched_stop_str", "input_logprobs", "index")
OUTPUT_IDS_FIELD_NUMBER: _ClassVar[int]
FINISH_REASON_FIELD_NUMBER: _ClassVar[int]
PROMPT_TOKENS_FIELD_NUMBER: _ClassVar[int]
COMPLETION_TOKENS_FIELD_NUMBER: _ClassVar[int]
CACHED_TOKENS_FIELD_NUMBER: _ClassVar[int]
OUTPUT_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
ALL_HIDDEN_STATES_FIELD_NUMBER: _ClassVar[int]
MATCHED_TOKEN_ID_FIELD_NUMBER: _ClassVar[int]
MATCHED_STOP_STR_FIELD_NUMBER: _ClassVar[int]
INPUT_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
INDEX_FIELD_NUMBER: _ClassVar[int]
output_ids: _containers.RepeatedScalarFieldContainer[int]
finish_reason: str
prompt_tokens: int
completion_tokens: int
cached_tokens: int
output_logprobs: OutputLogProbs
all_hidden_states: _containers.RepeatedCompositeFieldContainer[HiddenStates]
matched_token_id: int
matched_stop_str: str
input_logprobs: InputLogProbs
index: int
def __init__(self, output_ids: _Optional[_Iterable[int]] = ..., finish_reason: _Optional[str] = ..., prompt_tokens: _Optional[int] = ..., completion_tokens: _Optional[int] = ..., cached_tokens: _Optional[int] = ..., output_logprobs: _Optional[_Union[OutputLogProbs, _Mapping]] = ..., all_hidden_states: _Optional[_Iterable[_Union[HiddenStates, _Mapping]]] = ..., matched_token_id: _Optional[int] = ..., matched_stop_str: _Optional[str] = ..., input_logprobs: _Optional[_Union[InputLogProbs, _Mapping]] = ..., index: _Optional[int] = ...) -> None: ...
class GenerateError(_message.Message):
__slots__ = ("message", "http_status_code", "details")
MESSAGE_FIELD_NUMBER: _ClassVar[int]
HTTP_STATUS_CODE_FIELD_NUMBER: _ClassVar[int]
DETAILS_FIELD_NUMBER: _ClassVar[int]
message: str
http_status_code: str
details: str
def __init__(self, message: _Optional[str] = ..., http_status_code: _Optional[str] = ..., details: _Optional[str] = ...) -> None: ...
class OutputLogProbs(_message.Message):
__slots__ = ("token_logprobs", "token_ids", "top_logprobs")
TOKEN_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
TOKEN_IDS_FIELD_NUMBER: _ClassVar[int]
TOP_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
token_logprobs: _containers.RepeatedScalarFieldContainer[float]
token_ids: _containers.RepeatedScalarFieldContainer[int]
top_logprobs: _containers.RepeatedCompositeFieldContainer[TopLogProbs]
def __init__(self, token_logprobs: _Optional[_Iterable[float]] = ..., token_ids: _Optional[_Iterable[int]] = ..., top_logprobs: _Optional[_Iterable[_Union[TopLogProbs, _Mapping]]] = ...) -> None: ...
class InputLogProbs(_message.Message):
__slots__ = ("token_logprobs", "token_ids", "top_logprobs")
TOKEN_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
TOKEN_IDS_FIELD_NUMBER: _ClassVar[int]
TOP_LOGPROBS_FIELD_NUMBER: _ClassVar[int]
token_logprobs: _containers.RepeatedCompositeFieldContainer[InputTokenLogProb]
token_ids: _containers.RepeatedScalarFieldContainer[int]
top_logprobs: _containers.RepeatedCompositeFieldContainer[TopLogProbs]
def __init__(self, token_logprobs: _Optional[_Iterable[_Union[InputTokenLogProb, _Mapping]]] = ..., token_ids: _Optional[_Iterable[int]] = ..., top_logprobs: _Optional[_Iterable[_Union[TopLogProbs, _Mapping]]] = ...) -> None: ...
class InputTokenLogProb(_message.Message):
__slots__ = ("value",)
VALUE_FIELD_NUMBER: _ClassVar[int]
value: float
def __init__(self, value: _Optional[float] = ...) -> None: ...
class TopLogProbs(_message.Message):
__slots__ = ("values", "token_ids")
VALUES_FIELD_NUMBER: _ClassVar[int]
TOKEN_IDS_FIELD_NUMBER: _ClassVar[int]
values: _containers.RepeatedScalarFieldContainer[float]
token_ids: _containers.RepeatedScalarFieldContainer[int]
def __init__(self, values: _Optional[_Iterable[float]] = ..., token_ids: _Optional[_Iterable[int]] = ...) -> None: ...
class HiddenStates(_message.Message):
__slots__ = ("values", "layer", "position")
VALUES_FIELD_NUMBER: _ClassVar[int]
LAYER_FIELD_NUMBER: _ClassVar[int]
POSITION_FIELD_NUMBER: _ClassVar[int]
values: _containers.RepeatedScalarFieldContainer[float]
layer: int
position: int
def __init__(self, values: _Optional[_Iterable[float]] = ..., layer: _Optional[int] = ..., position: _Optional[int] = ...) -> None: ...
class EmbedRequest(_message.Message):
__slots__ = ("request_id", "tokenized", "mm_inputs", "sampling_params", "log_metrics", "token_type_ids", "data_parallel_rank", "is_cross_encoder", "texts")
REQUEST_ID_FIELD_NUMBER: _ClassVar[int]
TOKENIZED_FIELD_NUMBER: _ClassVar[int]
MM_INPUTS_FIELD_NUMBER: _ClassVar[int]
SAMPLING_PARAMS_FIELD_NUMBER: _ClassVar[int]
LOG_METRICS_FIELD_NUMBER: _ClassVar[int]
TOKEN_TYPE_IDS_FIELD_NUMBER: _ClassVar[int]
DATA_PARALLEL_RANK_FIELD_NUMBER: _ClassVar[int]
IS_CROSS_ENCODER_FIELD_NUMBER: _ClassVar[int]
TEXTS_FIELD_NUMBER: _ClassVar[int]
request_id: str
tokenized: TokenizedInput
mm_inputs: MultimodalInputs
sampling_params: SamplingParams
log_metrics: bool
token_type_ids: _containers.RepeatedScalarFieldContainer[int]
data_parallel_rank: int
is_cross_encoder: bool
texts: _containers.RepeatedScalarFieldContainer[str]
def __init__(self, request_id: _Optional[str] = ..., tokenized: _Optional[_Union[TokenizedInput, _Mapping]] = ..., mm_inputs: _Optional[_Union[MultimodalInputs, _Mapping]] = ..., sampling_params: _Optional[_Union[SamplingParams, _Mapping]] = ..., log_metrics: bool = ..., token_type_ids: _Optional[_Iterable[int]] = ..., data_parallel_rank: _Optional[int] = ..., is_cross_encoder: bool = ..., texts: _Optional[_Iterable[str]] = ...) -> None: ...
class EmbedResponse(_message.Message):
__slots__ = ("request_id", "complete", "error")
REQUEST_ID_FIELD_NUMBER: _ClassVar[int]
COMPLETE_FIELD_NUMBER: _ClassVar[int]
ERROR_FIELD_NUMBER: _ClassVar[int]
request_id: str
complete: EmbedComplete
error: EmbedError
def __init__(self, request_id: _Optional[str] = ..., complete: _Optional[_Union[EmbedComplete, _Mapping]] = ..., error: _Optional[_Union[EmbedError, _Mapping]] = ...) -> None: ...
class EmbedComplete(_message.Message):
__slots__ = ("embedding", "prompt_tokens", "cached_tokens", "embedding_dim", "batch_embeddings")
EMBEDDING_FIELD_NUMBER: _ClassVar[int]
PROMPT_TOKENS_FIELD_NUMBER: _ClassVar[int]
CACHED_TOKENS_FIELD_NUMBER: _ClassVar[int]
EMBEDDING_DIM_FIELD_NUMBER: _ClassVar[int]
BATCH_EMBEDDINGS_FIELD_NUMBER: _ClassVar[int]
embedding: _containers.RepeatedScalarFieldContainer[float]
prompt_tokens: int
cached_tokens: int
embedding_dim: int
batch_embeddings: _containers.RepeatedCompositeFieldContainer[Embedding]
def __init__(self, embedding: _Optional[_Iterable[float]] = ..., prompt_tokens: _Optional[int] = ..., cached_tokens: _Optional[int] = ..., embedding_dim: _Optional[int] = ..., batch_embeddings: _Optional[_Iterable[_Union[Embedding, _Mapping]]] = ...) -> None: ...
class Embedding(_message.Message):
__slots__ = ("values", "index")
VALUES_FIELD_NUMBER: _ClassVar[int]
INDEX_FIELD_NUMBER: _ClassVar[int]
values: _containers.RepeatedScalarFieldContainer[float]
index: int
def __init__(self, values: _Optional[_Iterable[float]] = ..., index: _Optional[int] = ...) -> None: ...
class EmbedError(_message.Message):
__slots__ = ("message", "code", "details")
MESSAGE_FIELD_NUMBER: _ClassVar[int]
CODE_FIELD_NUMBER: _ClassVar[int]
DETAILS_FIELD_NUMBER: _ClassVar[int]
message: str
code: str
details: str
def __init__(self, message: _Optional[str] = ..., code: _Optional[str] = ..., details: _Optional[str] = ...) -> None: ...
class HealthCheckRequest(_message.Message):
__slots__ = ()
def __init__(self) -> None: ...
class HealthCheckResponse(_message.Message):
__slots__ = ("healthy", "message")
HEALTHY_FIELD_NUMBER: _ClassVar[int]
MESSAGE_FIELD_NUMBER: _ClassVar[int]
healthy: bool
message: str
def __init__(self, healthy: bool = ..., message: _Optional[str] = ...) -> None: ...
class AbortRequest(_message.Message):
__slots__ = ("request_id", "reason")
REQUEST_ID_FIELD_NUMBER: _ClassVar[int]
REASON_FIELD_NUMBER: _ClassVar[int]
request_id: str
reason: str
def __init__(self, request_id: _Optional[str] = ..., reason: _Optional[str] = ...) -> None: ...
class AbortResponse(_message.Message):
__slots__ = ("success", "message")
SUCCESS_FIELD_NUMBER: _ClassVar[int]
MESSAGE_FIELD_NUMBER: _ClassVar[int]
success: bool
message: str
def __init__(self, success: bool = ..., message: _Optional[str] = ...) -> None: ...
class LoadLoRARequest(_message.Message):
__slots__ = ("adapter_id", "adapter_path", "rank")
ADAPTER_ID_FIELD_NUMBER: _ClassVar[int]
ADAPTER_PATH_FIELD_NUMBER: _ClassVar[int]
RANK_FIELD_NUMBER: _ClassVar[int]
adapter_id: str
adapter_path: str
rank: int
def __init__(self, adapter_id: _Optional[str] = ..., adapter_path: _Optional[str] = ..., rank: _Optional[int] = ...) -> None: ...
class LoadLoRAResponse(_message.Message):
__slots__ = ("success", "adapter_id", "message")
SUCCESS_FIELD_NUMBER: _ClassVar[int]
ADAPTER_ID_FIELD_NUMBER: _ClassVar[int]
MESSAGE_FIELD_NUMBER: _ClassVar[int]
success: bool
adapter_id: str
message: str
def __init__(self, success: bool = ..., adapter_id: _Optional[str] = ..., message: _Optional[str] = ...) -> None: ...
class UnloadLoRARequest(_message.Message):
__slots__ = ("adapter_id",)
ADAPTER_ID_FIELD_NUMBER: _ClassVar[int]
adapter_id: str
def __init__(self, adapter_id: _Optional[str] = ...) -> None: ...
class UnloadLoRAResponse(_message.Message):
__slots__ = ("success", "message")
SUCCESS_FIELD_NUMBER: _ClassVar[int]
MESSAGE_FIELD_NUMBER: _ClassVar[int]
success: bool
message: str
def __init__(self, success: bool = ..., message: _Optional[str] = ...) -> None: ...
class UpdateWeightsRequest(_message.Message):
__slots__ = ("disk_path", "tensor_data", "remote_url", "weight_name")
DISK_PATH_FIELD_NUMBER: _ClassVar[int]
TENSOR_DATA_FIELD_NUMBER: _ClassVar[int]
REMOTE_URL_FIELD_NUMBER: _ClassVar[int]
WEIGHT_NAME_FIELD_NUMBER: _ClassVar[int]
disk_path: str
tensor_data: bytes
remote_url: str
weight_name: str
def __init__(self, disk_path: _Optional[str] = ..., tensor_data: _Optional[bytes] = ..., remote_url: _Optional[str] = ..., weight_name: _Optional[str] = ...) -> None: ...
class UpdateWeightsResponse(_message.Message):
__slots__ = ("success", "message")
SUCCESS_FIELD_NUMBER: _ClassVar[int]
MESSAGE_FIELD_NUMBER: _ClassVar[int]
success: bool
message: str
def __init__(self, success: bool = ..., message: _Optional[str] = ...) -> None: ...
class GetInternalStateRequest(_message.Message):
__slots__ = ("state_keys",)
STATE_KEYS_FIELD_NUMBER: _ClassVar[int]
state_keys: _containers.RepeatedScalarFieldContainer[str]
def __init__(self, state_keys: _Optional[_Iterable[str]] = ...) -> None: ...
class GetInternalStateResponse(_message.Message):
__slots__ = ("state",)
STATE_FIELD_NUMBER: _ClassVar[int]
state: _struct_pb2.Struct
def __init__(self, state: _Optional[_Union[_struct_pb2.Struct, _Mapping]] = ...) -> None: ...
class SetInternalStateRequest(_message.Message):
__slots__ = ("state",)
STATE_FIELD_NUMBER: _ClassVar[int]
state: _struct_pb2.Struct
def __init__(self, state: _Optional[_Union[_struct_pb2.Struct, _Mapping]] = ...) -> None: ...
class SetInternalStateResponse(_message.Message):
__slots__ = ("success", "message")
SUCCESS_FIELD_NUMBER: _ClassVar[int]
MESSAGE_FIELD_NUMBER: _ClassVar[int]
success: bool
message: str
def __init__(self, success: bool = ..., message: _Optional[str] = ...) -> None: ...
class GetModelInfoRequest(_message.Message):
__slots__ = ()
def __init__(self) -> None: ...
class GetModelInfoResponse(_message.Message):
__slots__ = ("model_path", "tokenizer_path", "is_generation", "preferred_sampling_params", "weight_version", "served_model_name", "max_context_length", "vocab_size", "supports_vision", "model_type", "eos_token_ids", "pad_token_id", "bos_token_id", "max_req_input_len", "architectures", "id2label_json", "num_labels")
MODEL_PATH_FIELD_NUMBER: _ClassVar[int]
TOKENIZER_PATH_FIELD_NUMBER: _ClassVar[int]
IS_GENERATION_FIELD_NUMBER: _ClassVar[int]
PREFERRED_SAMPLING_PARAMS_FIELD_NUMBER: _ClassVar[int]
WEIGHT_VERSION_FIELD_NUMBER: _ClassVar[int]
SERVED_MODEL_NAME_FIELD_NUMBER: _ClassVar[int]
MAX_CONTEXT_LENGTH_FIELD_NUMBER: _ClassVar[int]
VOCAB_SIZE_FIELD_NUMBER: _ClassVar[int]
SUPPORTS_VISION_FIELD_NUMBER: _ClassVar[int]
MODEL_TYPE_FIELD_NUMBER: _ClassVar[int]
EOS_TOKEN_IDS_FIELD_NUMBER: _ClassVar[int]
PAD_TOKEN_ID_FIELD_NUMBER: _ClassVar[int]
BOS_TOKEN_ID_FIELD_NUMBER: _ClassVar[int]
MAX_REQ_INPUT_LEN_FIELD_NUMBER: _ClassVar[int]
ARCHITECTURES_FIELD_NUMBER: _ClassVar[int]
ID2LABEL_JSON_FIELD_NUMBER: _ClassVar[int]
NUM_LABELS_FIELD_NUMBER: _ClassVar[int]
model_path: str
tokenizer_path: str
is_generation: bool
preferred_sampling_params: str
weight_version: str
served_model_name: str
max_context_length: int
vocab_size: int
supports_vision: bool
model_type: str
eos_token_ids: _containers.RepeatedScalarFieldContainer[int]
pad_token_id: int
bos_token_id: int
max_req_input_len: int
architectures: _containers.RepeatedScalarFieldContainer[str]
id2label_json: str
num_labels: int
def __init__(self, model_path: _Optional[str] = ..., tokenizer_path: _Optional[str] = ..., is_generation: bool = ..., preferred_sampling_params: _Optional[str] = ..., weight_version: _Optional[str] = ..., served_model_name: _Optional[str] = ..., max_context_length: _Optional[int] = ..., vocab_size: _Optional[int] = ..., supports_vision: bool = ..., model_type: _Optional[str] = ..., eos_token_ids: _Optional[_Iterable[int]] = ..., pad_token_id: _Optional[int] = ..., bos_token_id: _Optional[int] = ..., max_req_input_len: _Optional[int] = ..., architectures: _Optional[_Iterable[str]] = ..., id2label_json: _Optional[str] = ..., num_labels: _Optional[int] = ...) -> None: ...
class GetServerInfoRequest(_message.Message):
__slots__ = ()
def __init__(self) -> None: ...
class GetServerInfoResponse(_message.Message):
__slots__ = ("server_args", "scheduler_info", "active_requests", "is_paused", "last_receive_timestamp", "uptime_seconds", "sglang_version", "server_type", "start_time")
SERVER_ARGS_FIELD_NUMBER: _ClassVar[int]
SCHEDULER_INFO_FIELD_NUMBER: _ClassVar[int]
ACTIVE_REQUESTS_FIELD_NUMBER: _ClassVar[int]
IS_PAUSED_FIELD_NUMBER: _ClassVar[int]
LAST_RECEIVE_TIMESTAMP_FIELD_NUMBER: _ClassVar[int]
UPTIME_SECONDS_FIELD_NUMBER: _ClassVar[int]
SGLANG_VERSION_FIELD_NUMBER: _ClassVar[int]
SERVER_TYPE_FIELD_NUMBER: _ClassVar[int]
START_TIME_FIELD_NUMBER: _ClassVar[int]
server_args: _struct_pb2.Struct
scheduler_info: _struct_pb2.Struct
active_requests: int
is_paused: bool
last_receive_timestamp: float
uptime_seconds: float
sglang_version: str
server_type: str
start_time: _timestamp_pb2.Timestamp
def __init__(self, server_args: _Optional[_Union[_struct_pb2.Struct, _Mapping]] = ..., scheduler_info: _Optional[_Union[_struct_pb2.Struct, _Mapping]] = ..., active_requests: _Optional[int] = ..., is_paused: bool = ..., last_receive_timestamp: _Optional[float] = ..., uptime_seconds: _Optional[float] = ..., sglang_version: _Optional[str] = ..., server_type: _Optional[str] = ..., start_time: _Optional[_Union[datetime.datetime, _timestamp_pb2.Timestamp, _Mapping]] = ...) -> None: ...
python/sglang/srt/grpc/sglang_scheduler_pb2_grpc.py
-327
View File
@@ -1,327 +0,0 @@
# This file is auto-generated. Do not edit manually.
# Regenerate with: python compile_proto.py
# Generated by the gRPC Python protocol compiler plugin. DO NOT EDIT!
"""Client and server classes corresponding to protobuf-defined services."""
import grpc
import warnings
from . import sglang_scheduler_pb2 as sglang__scheduler__pb2
GRPC_GENERATED_VERSION = '1.75.1'
GRPC_VERSION = grpc.__version__
_version_not_supported = False
try:
from grpc._utilities import first_version_is_lower
_version_not_supported = first_version_is_lower(GRPC_VERSION, GRPC_GENERATED_VERSION)
except ImportError:
_version_not_supported = True
if _version_not_supported:
raise RuntimeError(
f'The grpc package installed is at version {GRPC_VERSION},'
+ f' but the generated code in sglang_scheduler_pb2_grpc.py depends on'
+ f' grpcio>={GRPC_GENERATED_VERSION}.'
+ f' Please upgrade your grpc module to grpcio>={GRPC_GENERATED_VERSION}'
+ f' or downgrade your generated code using grpcio-tools<={GRPC_VERSION}.'
)
class SglangSchedulerStub(object):
"""Service definition for SGLang scheduler communication
This protocol bridges the Rust router and Python scheduler
"""
def __init__(self, channel):
"""Constructor.
Args:
channel: A grpc.Channel.
"""
self.Generate = channel.unary_stream(
'/sglang.grpc.scheduler.SglangScheduler/Generate',
request_serializer=sglang__scheduler__pb2.GenerateRequest.SerializeToString,
response_deserializer=sglang__scheduler__pb2.GenerateResponse.FromString,
_registered_method=True)
self.Embed = channel.unary_unary(
'/sglang.grpc.scheduler.SglangScheduler/Embed',
request_serializer=sglang__scheduler__pb2.EmbedRequest.SerializeToString,
response_deserializer=sglang__scheduler__pb2.EmbedResponse.FromString,
_registered_method=True)
self.HealthCheck = channel.unary_unary(
'/sglang.grpc.scheduler.SglangScheduler/HealthCheck',
request_serializer=sglang__scheduler__pb2.HealthCheckRequest.SerializeToString,
response_deserializer=sglang__scheduler__pb2.HealthCheckResponse.FromString,
_registered_method=True)
self.Abort = channel.unary_unary(
'/sglang.grpc.scheduler.SglangScheduler/Abort',
request_serializer=sglang__scheduler__pb2.AbortRequest.SerializeToString,
response_deserializer=sglang__scheduler__pb2.AbortResponse.FromString,
_registered_method=True)
self.GetModelInfo = channel.unary_unary(
'/sglang.grpc.scheduler.SglangScheduler/GetModelInfo',
request_serializer=sglang__scheduler__pb2.GetModelInfoRequest.SerializeToString,
response_deserializer=sglang__scheduler__pb2.GetModelInfoResponse.FromString,
_registered_method=True)
self.GetServerInfo = channel.unary_unary(
'/sglang.grpc.scheduler.SglangScheduler/GetServerInfo',
request_serializer=sglang__scheduler__pb2.GetServerInfoRequest.SerializeToString,
response_deserializer=sglang__scheduler__pb2.GetServerInfoResponse.FromString,
_registered_method=True)
class SglangSchedulerServicer(object):
"""Service definition for SGLang scheduler communication
This protocol bridges the Rust router and Python scheduler
"""
def Generate(self, request, context):
"""Submit a generation request (supports streaming)
"""
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
context.set_details('Method not implemented!')
raise NotImplementedError('Method not implemented!')
def Embed(self, request, context):
"""Submit an embedding request
"""
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
context.set_details('Method not implemented!')
raise NotImplementedError('Method not implemented!')
def HealthCheck(self, request, context):
"""Health check and metrics
"""
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
context.set_details('Method not implemented!')
raise NotImplementedError('Method not implemented!')
def Abort(self, request, context):
"""Abort a running request
"""
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
context.set_details('Method not implemented!')
raise NotImplementedError('Method not implemented!')
def GetModelInfo(self, request, context):
"""Get model information
"""
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
context.set_details('Method not implemented!')
raise NotImplementedError('Method not implemented!')
def GetServerInfo(self, request, context):
"""Get server information
"""
context.set_code(grpc.StatusCode.UNIMPLEMENTED)
context.set_details('Method not implemented!')
raise NotImplementedError('Method not implemented!')
def add_SglangSchedulerServicer_to_server(servicer, server):
rpc_method_handlers = {
'Generate': grpc.unary_stream_rpc_method_handler(
servicer.Generate,
request_deserializer=sglang__scheduler__pb2.GenerateRequest.FromString,
response_serializer=sglang__scheduler__pb2.GenerateResponse.SerializeToString,
),
'Embed': grpc.unary_unary_rpc_method_handler(
servicer.Embed,
request_deserializer=sglang__scheduler__pb2.EmbedRequest.FromString,
response_serializer=sglang__scheduler__pb2.EmbedResponse.SerializeToString,
),
'HealthCheck': grpc.unary_unary_rpc_method_handler(
servicer.HealthCheck,
request_deserializer=sglang__scheduler__pb2.HealthCheckRequest.FromString,
response_serializer=sglang__scheduler__pb2.HealthCheckResponse.SerializeToString,
),
'Abort': grpc.unary_unary_rpc_method_handler(
servicer.Abort,
request_deserializer=sglang__scheduler__pb2.AbortRequest.FromString,
response_serializer=sglang__scheduler__pb2.AbortResponse.SerializeToString,
),
'GetModelInfo': grpc.unary_unary_rpc_method_handler(
servicer.GetModelInfo,
request_deserializer=sglang__scheduler__pb2.GetModelInfoRequest.FromString,
response_serializer=sglang__scheduler__pb2.GetModelInfoResponse.SerializeToString,
),
'GetServerInfo': grpc.unary_unary_rpc_method_handler(
servicer.GetServerInfo,
request_deserializer=sglang__scheduler__pb2.GetServerInfoRequest.FromString,
response_serializer=sglang__scheduler__pb2.GetServerInfoResponse.SerializeToString,
),
}
generic_handler = grpc.method_handlers_generic_handler(
'sglang.grpc.scheduler.SglangScheduler', rpc_method_handlers)
server.add_generic_rpc_handlers((generic_handler,))
server.add_registered_method_handlers('sglang.grpc.scheduler.SglangScheduler', rpc_method_handlers)
# This class is part of an EXPERIMENTAL API.
class SglangScheduler(object):
"""Service definition for SGLang scheduler communication
This protocol bridges the Rust router and Python scheduler
"""
@staticmethod
def Generate(request,
target,
options=(),
channel_credentials=None,
call_credentials=None,
insecure=False,
compression=None,
wait_for_ready=None,
timeout=None,
metadata=None):
return grpc.experimental.unary_stream(
request,
target,
'/sglang.grpc.scheduler.SglangScheduler/Generate',
sglang__scheduler__pb2.GenerateRequest.SerializeToString,
sglang__scheduler__pb2.GenerateResponse.FromString,
options,
channel_credentials,
insecure,
call_credentials,
compression,
wait_for_ready,
timeout,
metadata,
_registered_method=True)
@staticmethod
def Embed(request,
target,
options=(),
channel_credentials=None,
call_credentials=None,
insecure=False,
compression=None,
wait_for_ready=None,
timeout=None,
metadata=None):
return grpc.experimental.unary_unary(
request,
target,
'/sglang.grpc.scheduler.SglangScheduler/Embed',
sglang__scheduler__pb2.EmbedRequest.SerializeToString,
sglang__scheduler__pb2.EmbedResponse.FromString,
options,
channel_credentials,
insecure,
call_credentials,
compression,
wait_for_ready,
timeout,
metadata,
_registered_method=True)
@staticmethod
def HealthCheck(request,
target,
options=(),
channel_credentials=None,
call_credentials=None,
insecure=False,
compression=None,
wait_for_ready=None,
timeout=None,
metadata=None):
return grpc.experimental.unary_unary(
request,
target,
'/sglang.grpc.scheduler.SglangScheduler/HealthCheck',
sglang__scheduler__pb2.HealthCheckRequest.SerializeToString,
sglang__scheduler__pb2.HealthCheckResponse.FromString,
options,
channel_credentials,
insecure,
call_credentials,
compression,
wait_for_ready,
timeout,
metadata,
_registered_method=True)
@staticmethod
def Abort(request,
target,
options=(),
channel_credentials=None,
call_credentials=None,
insecure=False,
compression=None,
wait_for_ready=None,
timeout=None,
metadata=None):
return grpc.experimental.unary_unary(
request,
target,
'/sglang.grpc.scheduler.SglangScheduler/Abort',
sglang__scheduler__pb2.AbortRequest.SerializeToString,
sglang__scheduler__pb2.AbortResponse.FromString,
options,
channel_credentials,
insecure,
call_credentials,
compression,
wait_for_ready,
timeout,
metadata,
_registered_method=True)
@staticmethod
def GetModelInfo(request,
target,
options=(),
channel_credentials=None,
call_credentials=None,
insecure=False,
compression=None,
wait_for_ready=None,
timeout=None,
metadata=None):
return grpc.experimental.unary_unary(
request,
target,
'/sglang.grpc.scheduler.SglangScheduler/GetModelInfo',
sglang__scheduler__pb2.GetModelInfoRequest.SerializeToString,
sglang__scheduler__pb2.GetModelInfoResponse.FromString,
options,
channel_credentials,
insecure,
call_credentials,
compression,
wait_for_ready,
timeout,
metadata,
_registered_method=True)
@staticmethod
def GetServerInfo(request,
target,
options=(),
channel_credentials=None,
call_credentials=None,
insecure=False,
compression=None,
wait_for_ready=None,
timeout=None,
metadata=None):
return grpc.experimental.unary_unary(
request,
target,
'/sglang.grpc.scheduler.SglangScheduler/GetServerInfo',
sglang__scheduler__pb2.GetServerInfoRequest.SerializeToString,
sglang__scheduler__pb2.GetServerInfoResponse.FromString,
options,
channel_credentials,
insecure,
call_credentials,
compression,
wait_for_ready,
timeout,
metadata,
_registered_method=True)