File size: 8,738 Bytes
fcaa164 |
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 |
# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
from typing import Any, Callable, List, Literal, Type, Union
from pydantic import BaseModel
from .base import BaseConverter
class OutlinesConverter(BaseConverter):
r"""OutlinesConverter is a class that converts a string or a function
into a BaseModel schema.
Args:
model_type (str, optional): The model type to be used.
platform (str, optional): The platform to be used.
1. transformers
2. mamba
3. vllm
4. llamacpp
5. mlx
(default: "transformers")
**kwargs: The keyword arguments to be used. See the outlines
documentation for more details. See
https://dottxt-ai.github.io/outlines/latest/reference/models/models/
"""
def __init__(
self,
model_type: str,
platform: Literal[
"vllm", "transformers", "mamba", "llamacpp", "mlx"
] = "transformers",
**kwargs: Any,
):
self.model_type = model_type
from outlines import models
match platform:
case "vllm":
self._outlines_model = models.vllm(model_type, **kwargs)
case "transformers":
self._outlines_model = models.transformers(
model_type, **kwargs
)
case "mamba":
self._outlines_model = models.mamba(model_type, **kwargs)
case "llamacpp":
self._outlines_model = models.llamacpp(model_type, **kwargs)
case "mlx":
self._outlines_model = models.mlxlm(model_type, **kwargs)
case _:
raise ValueError(f"Unsupported platform: {platform}")
def convert_regex(self, content: str, regex_pattern: str) -> str:
r"""Convert the content to the specified regex pattern.
Args:
content (str): The content to be converted.
regex_pattern (str): The regex pattern to be used.
Returns:
str: The converted content.
"""
import outlines
regex_generator = outlines.generate.regex(
self._outlines_model, regex_pattern
)
return regex_generator(content)
def convert_json(
self,
content: str,
output_schema: Union[str, Callable],
) -> dict:
r"""Convert the content to the specified JSON schema given by
output_schema.
Args:
content (str): The content to be converted.
output_schema (Union[str, Callable]): The expected format of the
response.
Returns:
dict: The converted content in JSON format.
"""
import outlines
json_generator = outlines.generate.json(
self._outlines_model, output_schema
)
return json_generator(content)
def convert_pydantic(
self,
content: str,
output_schema: Type[BaseModel],
) -> BaseModel:
r"""Convert the content to the specified Pydantic schema.
Args:
content (str): The content to be converted.
output_schema (Type[BaseModel]): The expected format of the
response.
Returns:
BaseModel: The converted content in pydantic model format.
"""
import outlines
json_generator = outlines.generate.json(
self._outlines_model, output_schema
)
return json_generator(content)
def convert_type(self, content: str, type_name: type) -> str:
r"""Convert the content to the specified type.
The following types are currently available:
1. int
2. float
3. bool
4. datetime.date
5. datetime.time
6. datetime.datetime
7. custom types (https://dottxt-ai.github.io/outlines/latest/reference/generation/types/)
Args:
content (str): The content to be converted.
type_name (type): The type to be used.
Returns:
str: The converted content.
"""
import outlines
type_generator = outlines.generate.format(
self._outlines_model, type_name
)
return type_generator(content)
def convert_choice(self, content: str, choices: List[str]) -> str:
r"""Convert the content to the specified choice.
Args:
content (str): The content to be converted.
choices (List[str]): The choices to be used.
Returns:
str: The converted content.
"""
import outlines
choices_generator = outlines.generate.choice(
self._outlines_model, choices
)
return choices_generator(content)
def convert_grammar(self, content: str, grammar: str) -> str:
r"""Convert the content to the specified grammar.
Args:
content (str): The content to be converted.
grammar (str): The grammar to be used.
Returns:
str: The converted content.
"""
import outlines
grammar_generator = outlines.generate.cfg(
self._outlines_model, grammar
)
return grammar_generator(content)
def convert( # type: ignore[override]
self,
content: str,
type: Literal["regex", "json", "type", "choice", "grammar"],
**kwargs,
) -> Any:
r"""Formats the input content into the expected BaseModel.
Args:
type (Literal["regex", "json", "type", "choice", "grammar"]):
The type of conversion to perform. Options are:
- "regex": Match the content against a regex pattern.
- "pydantic": Convert the content into a pydantic model.
- "json": Convert the content into a JSON based on a
schema.
- "type": Convert the content into a specified type.
- "choice": Match the content against a list of valid
choices.
- "grammar": Convert the content using a specified grammar.
content (str): The content to be formatted.
**kwargs: Additional keyword arguments specific to the conversion
type.
- For "regex":
regex_pattern (str): The regex pattern to use for matching.
- For "pydantic":
output_schema (Type[BaseModel]): The schema to validate and
format the pydantic model.
- For "json":
output_schema (Union[str, Callable]): The schema to validate
and format the JSON object.
- For "type":
type_name (str): The target type name for the conversion.
- For "choice":
choices (List[str]): A list of valid choices to match against.
- For "grammar":
grammar (str): The grammar definition to use for content
conversion.
"""
match type:
case "regex":
return self.convert_regex(content, kwargs.get("regex_pattern")) # type: ignore[arg-type]
case "pydantic":
return self.convert_pydantic(
content, kwargs.get("output_schema")
) # type: ignore[arg-type]
case "json":
return self.convert_json(content, kwargs.get("output_schema")) # type: ignore[arg-type]
case "type":
return self.convert_type(content, kwargs.get("type_name")) # type: ignore[arg-type]
case "choice":
return self.convert_choice(content, kwargs.get("choices")) # type: ignore[arg-type]
case "grammar":
return self.convert_grammar(content, kwargs.get("grammar")) # type: ignore[arg-type]
case _:
raise ValueError("Unsupported output schema type")
|