Spaces:
Sleeping
Sleeping
File size: 11,211 Bytes
469eae6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 |
"""
Translates from OpenAI's `/v1/chat/completions` endpoint to Triton's `/generate` endpoint.
"""
import json
from typing import Any, AsyncIterator, Dict, Iterator, List, Literal, Optional, Union
from httpx import Headers, Response
from litellm.constants import DEFAULT_MAX_TOKENS_FOR_TRITON
from litellm.litellm_core_utils.prompt_templates.factory import prompt_factory
from litellm.llms.base_llm.base_model_iterator import BaseModelResponseIterator
from litellm.llms.base_llm.chat.transformation import (
BaseConfig,
BaseLLMException,
LiteLLMLoggingObj,
)
from litellm.types.llms.openai import AllMessageValues
from litellm.types.utils import (
ChatCompletionToolCallChunk,
ChatCompletionUsageBlock,
Choices,
GenericStreamingChunk,
Message,
ModelResponse,
)
from ..common_utils import TritonError
class TritonConfig(BaseConfig):
"""
Base class for Triton configurations.
Handles routing between /infer and /generate triton completion llms
"""
def get_error_class(
self, error_message: str, status_code: int, headers: Union[Dict, Headers]
) -> BaseLLMException:
return TritonError(
status_code=status_code, message=error_message, headers=headers
)
def validate_environment(
self,
headers: Dict,
model: str,
messages: List[AllMessageValues],
optional_params: Dict,
litellm_params: dict,
api_key: Optional[str] = None,
api_base: Optional[str] = None,
) -> Dict:
return {"Content-Type": "application/json"}
def get_supported_openai_params(self, model: str) -> List:
return ["max_tokens", "max_completion_tokens"]
def map_openai_params(
self,
non_default_params: Dict,
optional_params: Dict,
model: str,
drop_params: bool,
) -> Dict:
for param, value in non_default_params.items():
if param == "max_tokens" or param == "max_completion_tokens":
optional_params[param] = value
return optional_params
def get_complete_url(
self,
api_base: Optional[str],
api_key: Optional[str],
model: str,
optional_params: dict,
litellm_params: dict,
stream: Optional[bool] = None,
) -> str:
if api_base is None:
raise ValueError("api_base is required")
llm_type = self._get_triton_llm_type(api_base)
if llm_type == "generate" and stream:
return api_base + "_stream"
return api_base
def transform_response(
self,
model: str,
raw_response: Response,
model_response: ModelResponse,
logging_obj: LiteLLMLoggingObj,
request_data: Dict,
messages: List[AllMessageValues],
optional_params: Dict,
litellm_params: Dict,
encoding: Any,
api_key: Optional[str] = None,
json_mode: Optional[bool] = None,
) -> ModelResponse:
api_base = litellm_params.get("api_base", "")
llm_type = self._get_triton_llm_type(api_base)
if llm_type == "generate":
return TritonGenerateConfig().transform_response(
model=model,
raw_response=raw_response,
model_response=model_response,
logging_obj=logging_obj,
request_data=request_data,
messages=messages,
optional_params=optional_params,
litellm_params=litellm_params,
encoding=encoding,
api_key=api_key,
json_mode=json_mode,
)
elif llm_type == "infer":
return TritonInferConfig().transform_response(
model=model,
raw_response=raw_response,
model_response=model_response,
logging_obj=logging_obj,
request_data=request_data,
messages=messages,
optional_params=optional_params,
litellm_params=litellm_params,
encoding=encoding,
api_key=api_key,
json_mode=json_mode,
)
return model_response
def transform_request(
self,
model: str,
messages: List[AllMessageValues],
optional_params: dict,
litellm_params: dict,
headers: dict,
) -> dict:
api_base = litellm_params.get("api_base", "")
llm_type = self._get_triton_llm_type(api_base)
if llm_type == "generate":
return TritonGenerateConfig().transform_request(
model=model,
messages=messages,
optional_params=optional_params,
litellm_params=litellm_params,
headers=headers,
)
elif llm_type == "infer":
return TritonInferConfig().transform_request(
model=model,
messages=messages,
optional_params=optional_params,
litellm_params=litellm_params,
headers=headers,
)
return {}
def _get_triton_llm_type(self, api_base: str) -> Literal["generate", "infer"]:
if api_base.endswith("/generate"):
return "generate"
elif api_base.endswith("/infer"):
return "infer"
else:
raise ValueError(f"Invalid Triton API base: {api_base}")
def get_model_response_iterator(
self,
streaming_response: Union[Iterator[str], AsyncIterator[str], ModelResponse],
sync_stream: bool,
json_mode: Optional[bool] = False,
) -> Any:
return TritonResponseIterator(
streaming_response=streaming_response,
sync_stream=sync_stream,
json_mode=json_mode,
)
class TritonGenerateConfig(TritonConfig):
"""
Transformations for triton /generate endpoint (This is a trtllm model)
"""
def transform_request(
self,
model: str,
messages: List[AllMessageValues],
optional_params: dict,
litellm_params: dict,
headers: dict,
) -> dict:
inference_params = optional_params.copy()
stream = inference_params.pop("stream", False)
data_for_triton: Dict[str, Any] = {
"text_input": prompt_factory(model=model, messages=messages),
"parameters": {
"max_tokens": int(
optional_params.get("max_tokens", DEFAULT_MAX_TOKENS_FOR_TRITON)
),
},
"stream": bool(stream),
}
data_for_triton["parameters"].update(inference_params)
return data_for_triton
def transform_response(
self,
model: str,
raw_response: Response,
model_response: ModelResponse,
logging_obj: LiteLLMLoggingObj,
request_data: Dict,
messages: List[AllMessageValues],
optional_params: Dict,
litellm_params: Dict,
encoding: Any,
api_key: Optional[str] = None,
json_mode: Optional[bool] = None,
) -> ModelResponse:
try:
raw_response_json = raw_response.json()
except Exception:
raise TritonError(
message=raw_response.text, status_code=raw_response.status_code
)
model_response.choices = [
Choices(index=0, message=Message(content=raw_response_json["text_output"]))
]
return model_response
class TritonInferConfig(TritonConfig):
"""
Transformations for triton /infer endpoint (his is an infer model with a custom model on triton)
"""
def transform_request(
self,
model: str,
messages: List[AllMessageValues],
optional_params: dict,
litellm_params: dict,
headers: dict,
) -> dict:
text_input = messages[0].get("content", "")
data_for_triton = {
"inputs": [
{
"name": "text_input",
"shape": [1],
"datatype": "BYTES",
"data": [text_input],
}
]
}
for k, v in optional_params.items():
if not (k == "stream" or k == "max_retries"):
datatype = "INT32" if isinstance(v, int) else "BYTES"
datatype = "FP32" if isinstance(v, float) else datatype
data_for_triton["inputs"].append(
{"name": k, "shape": [1], "datatype": datatype, "data": [v]}
)
if "max_tokens" not in optional_params:
data_for_triton["inputs"].append(
{
"name": "max_tokens",
"shape": [1],
"datatype": "INT32",
"data": [20],
}
)
return data_for_triton
def transform_response(
self,
model: str,
raw_response: Response,
model_response: ModelResponse,
logging_obj: LiteLLMLoggingObj,
request_data: Dict,
messages: List[AllMessageValues],
optional_params: Dict,
litellm_params: Dict,
encoding: Any,
api_key: Optional[str] = None,
json_mode: Optional[bool] = None,
) -> ModelResponse:
try:
raw_response_json = raw_response.json()
except Exception:
raise TritonError(
message=raw_response.text, status_code=raw_response.status_code
)
_triton_response_data = raw_response_json["outputs"][0]["data"]
triton_response_data: Optional[str] = None
if isinstance(_triton_response_data, list):
triton_response_data = "".join(_triton_response_data)
else:
triton_response_data = _triton_response_data
model_response.choices = [
Choices(
index=0,
message=Message(content=triton_response_data),
)
]
return model_response
class TritonResponseIterator(BaseModelResponseIterator):
def chunk_parser(self, chunk: dict) -> GenericStreamingChunk:
try:
text = ""
tool_use: Optional[ChatCompletionToolCallChunk] = None
is_finished = False
finish_reason = ""
usage: Optional[ChatCompletionUsageBlock] = None
provider_specific_fields = None
index = int(chunk.get("index", 0))
# set values
text = chunk.get("text_output", "")
finish_reason = chunk.get("stop_reason", "")
is_finished = chunk.get("is_finished", False)
return GenericStreamingChunk(
text=text,
tool_use=tool_use,
is_finished=is_finished,
finish_reason=finish_reason,
usage=usage,
index=index,
provider_specific_fields=provider_specific_fields,
)
except json.JSONDecodeError:
raise ValueError(f"Failed to decode JSON from chunk: {chunk}")
|