Spaces:
Runtime error
Runtime error
File size: 7,107 Bytes
4bdb245 |
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 |
from __future__ import annotations
import sys
import traceback
import time
from re import compile, Match, Pattern
from typing import Callable, Coroutine, Optional, Tuple, Union, Dict
from typing_extensions import TypedDict
from fastapi import (
Request,
Response,
HTTPException,
)
from fastapi.responses import JSONResponse
from fastapi.routing import APIRoute
from llama_cpp.server.types import (
CreateCompletionRequest,
CreateEmbeddingRequest,
CreateChatCompletionRequest,
)
class ErrorResponse(TypedDict):
"""OpenAI style error response"""
message: str
type: str
param: Optional[str]
code: Optional[str]
class ErrorResponseFormatters:
"""Collection of formatters for error responses.
Args:
request (Union[CreateCompletionRequest, CreateChatCompletionRequest]):
Request body
match (Match[str]): Match object from regex pattern
Returns:
Tuple[int, ErrorResponse]: Status code and error response
"""
@staticmethod
def context_length_exceeded(
request: Union["CreateCompletionRequest", "CreateChatCompletionRequest"],
match, # type: Match[str] # type: ignore
) -> Tuple[int, ErrorResponse]:
"""Formatter for context length exceeded error"""
context_window = int(match.group(2))
prompt_tokens = int(match.group(1))
completion_tokens = request.max_tokens
if hasattr(request, "messages"):
# Chat completion
message = (
"This model's maximum context length is {} tokens. "
"However, you requested {} tokens "
"({} in the messages, {} in the completion). "
"Please reduce the length of the messages or completion."
)
else:
# Text completion
message = (
"This model's maximum context length is {} tokens, "
"however you requested {} tokens "
"({} in your prompt; {} for the completion). "
"Please reduce your prompt; or completion length."
)
return 400, ErrorResponse(
message=message.format(
context_window,
(completion_tokens or 0) + prompt_tokens,
prompt_tokens,
completion_tokens,
), # type: ignore
type="invalid_request_error",
param="messages",
code="context_length_exceeded",
)
@staticmethod
def model_not_found(
request: Union["CreateCompletionRequest", "CreateChatCompletionRequest"],
match, # type: Match[str] # type: ignore
) -> Tuple[int, ErrorResponse]:
"""Formatter for model_not_found error"""
model_path = str(match.group(1))
message = f"The model `{model_path}` does not exist"
return 400, ErrorResponse(
message=message,
type="invalid_request_error",
param=None,
code="model_not_found",
)
class RouteErrorHandler(APIRoute):
"""Custom APIRoute that handles application errors and exceptions"""
# key: regex pattern for original error message from llama_cpp
# value: formatter function
pattern_and_formatters: Dict[
"Pattern[str]",
Callable[
[
Union["CreateCompletionRequest", "CreateChatCompletionRequest"],
"Match[str]",
],
Tuple[int, ErrorResponse],
],
] = {
compile(
r"Requested tokens \((\d+)\) exceed context window of (\d+)"
): ErrorResponseFormatters.context_length_exceeded,
compile(
r"Model path does not exist: (.+)"
): ErrorResponseFormatters.model_not_found,
}
def error_message_wrapper(
self,
error: Exception,
body: Optional[
Union[
"CreateChatCompletionRequest",
"CreateCompletionRequest",
"CreateEmbeddingRequest",
]
] = None,
) -> Tuple[int, ErrorResponse]:
"""Wraps error message in OpenAI style error response"""
print(f"Exception: {str(error)}", file=sys.stderr)
traceback.print_exc(file=sys.stderr)
if body is not None and isinstance(
body,
(
CreateCompletionRequest,
CreateChatCompletionRequest,
),
):
# When text completion or chat completion
for pattern, callback in self.pattern_and_formatters.items():
match = pattern.search(str(error))
if match is not None:
return callback(body, match)
# Wrap other errors as internal server error
return 500, ErrorResponse(
message=str(error),
type="internal_server_error",
param=None,
code=None,
)
def get_route_handler(
self,
) -> Callable[[Request], Coroutine[None, None, Response]]:
"""Defines custom route handler that catches exceptions and formats
in OpenAI style error response"""
original_route_handler = super().get_route_handler()
async def custom_route_handler(request: Request) -> Response:
try:
start_sec = time.perf_counter()
response = await original_route_handler(request)
elapsed_time_ms = int((time.perf_counter() - start_sec) * 1000)
response.headers["openai-processing-ms"] = f"{elapsed_time_ms}"
return response
except HTTPException as unauthorized:
# api key check failed
raise unauthorized
except Exception as exc:
json_body = await request.json()
try:
if "messages" in json_body:
# Chat completion
body: Optional[
Union[
CreateChatCompletionRequest,
CreateCompletionRequest,
CreateEmbeddingRequest,
]
] = CreateChatCompletionRequest(**json_body)
elif "prompt" in json_body:
# Text completion
body = CreateCompletionRequest(**json_body)
else:
# Embedding
body = CreateEmbeddingRequest(**json_body)
except Exception:
# Invalid request body
body = None
# Get proper error message from the exception
(
status_code,
error_message,
) = self.error_message_wrapper(error=exc, body=body)
return JSONResponse(
{"error": error_message},
status_code=status_code,
)
return custom_route_handler
|