Spaces:
Runtime error
Runtime error
File size: 7,891 Bytes
ed4d993 |
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 |
"""Wrapper around Together AI's Completion API."""
import logging
import warnings
from typing import Any, Dict, List, Optional
import requests
from aiohttp import ClientSession
from langchain_core.callbacks import (
AsyncCallbackManagerForLLMRun,
CallbackManagerForLLMRun,
)
from langchain_core.language_models.llms import LLM
from langchain_core.pydantic_v1 import Extra, SecretStr, root_validator
from langchain_core.utils import convert_to_secret_str, get_from_dict_or_env
logger = logging.getLogger(__name__)
class Together(LLM):
"""LLM models from `Together`.
To use, you'll need an API key which you can find here:
https://api.together.ai/settings/api-keys. This can be passed in as init param
``together_api_key`` or set as environment variable ``TOGETHER_API_KEY``.
Together AI API reference: https://docs.together.ai/reference/completions
Example:
.. code-block:: python
from langchain_together import Together
model = Together(model_name="mistralai/Mixtral-8x7B-Instruct-v0.1")
"""
base_url: str = "https://api.together.ai/v1/completions"
"""Base completions API URL."""
together_api_key: SecretStr
"""Together AI API key. Get it here: https://api.together.ai/settings/api-keys"""
model: str
"""Model name. Available models listed here:
Base Models: https://docs.together.ai/docs/inference-models#language-models
Chat Models: https://docs.together.ai/docs/inference-models#chat-models
"""
temperature: Optional[float] = None
"""Model temperature."""
top_p: Optional[float] = None
"""Used to dynamically adjust the number of choices for each predicted token based
on the cumulative probabilities. A value of 1 will always yield the same
output. A temperature less than 1 favors more correctness and is appropriate
for question answering or summarization. A value greater than 1 introduces more
randomness in the output.
"""
top_k: Optional[int] = None
"""Used to limit the number of choices for the next predicted word or token. It
specifies the maximum number of tokens to consider at each step, based on their
probability of occurrence. This technique helps to speed up the generation
process and can improve the quality of the generated text by focusing on the
most likely options.
"""
max_tokens: Optional[int] = None
"""The maximum number of tokens to generate."""
repetition_penalty: Optional[float] = None
"""A number that controls the diversity of generated text by reducing the
likelihood of repeated sequences. Higher values decrease repetition.
"""
logprobs: Optional[int] = None
"""An integer that specifies how many top token log probabilities are included in
the response for each token generation step.
"""
class Config:
"""Configuration for this pydantic object."""
extra = Extra.forbid
@root_validator(pre=True)
def validate_environment(cls, values: Dict) -> Dict:
"""Validate that api key exists in environment."""
values["together_api_key"] = convert_to_secret_str(
get_from_dict_or_env(values, "together_api_key", "TOGETHER_API_KEY")
)
return values
@root_validator()
def validate_max_tokens(cls, values: Dict) -> Dict:
"""
The v1 completions endpoint, has max_tokens as required parameter.
Set a default value and warn if the parameter is missing.
"""
if values.get("max_tokens") is None:
warnings.warn(
"The completions endpoint, has 'max_tokens' as required argument. "
"The default value is being set to 200 "
"Consider setting this value, when initializing LLM"
)
values["max_tokens"] = 200 # Default Value
return values
@property
def _llm_type(self) -> str:
"""Return type of model."""
return "together"
def _format_output(self, output: dict) -> str:
return output["choices"][0]["text"]
@property
def default_params(self) -> Dict[str, Any]:
return {
"model": self.model,
"temperature": self.temperature,
"top_p": self.top_p,
"top_k": self.top_k,
"max_tokens": self.max_tokens,
"repetition_penalty": self.repetition_penalty,
}
def _call(
self,
prompt: str,
stop: Optional[List[str]] = None,
run_manager: Optional[CallbackManagerForLLMRun] = None,
**kwargs: Any,
) -> str:
"""Call out to Together's text generation endpoint.
Args:
prompt: The prompt to pass into the model.
Returns:
The string generated by the model..
"""
headers = {
"Authorization": f"Bearer {self.together_api_key.get_secret_value()}",
"Content-Type": "application/json",
}
stop_to_use = stop[0] if stop and len(stop) == 1 else stop
payload: Dict[str, Any] = {
**self.default_params,
"prompt": prompt,
"stop": stop_to_use,
**kwargs,
}
# filter None values to not pass them to the http payload
payload = {k: v for k, v in payload.items() if v is not None}
response = requests.post(url=self.base_url, json=payload, headers=headers)
if response.status_code >= 500:
raise Exception(f"Together Server: Error {response.status_code}")
elif response.status_code >= 400:
raise ValueError(f"Together received an invalid payload: {response.text}")
elif response.status_code != 200:
raise Exception(
f"Together returned an unexpected response with status "
f"{response.status_code}: {response.text}"
)
data = response.json()
output = self._format_output(data)
return output
async def _acall(
self,
prompt: str,
stop: Optional[List[str]] = None,
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
**kwargs: Any,
) -> str:
"""Call Together model to get predictions based on the prompt.
Args:
prompt: The prompt to pass into the model.
Returns:
The string generated by the model.
"""
headers = {
"Authorization": f"Bearer {self.together_api_key.get_secret_value()}",
"Content-Type": "application/json",
}
stop_to_use = stop[0] if stop and len(stop) == 1 else stop
payload: Dict[str, Any] = {
**self.default_params,
"prompt": prompt,
"stop": stop_to_use,
**kwargs,
}
# filter None values to not pass them to the http payload
payload = {k: v for k, v in payload.items() if v is not None}
async with ClientSession() as session:
async with session.post(
self.base_url, json=payload, headers=headers
) as response:
if response.status >= 500:
raise Exception(f"Together Server: Error {response.status}")
elif response.status >= 400:
raise ValueError(
f"Together received an invalid payload: {response.text}"
)
elif response.status != 200:
raise Exception(
f"Together returned an unexpected response with status "
f"{response.status}: {response.text}"
)
response_json = await response.json()
output = self._format_output(response_json)
return output
|