Spaces:
Runtime error
Runtime error
File size: 7,114 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 |
from __future__ import annotations
from typing import Any, AsyncIterator, Dict, Iterator, List, Optional
from langchain_core.callbacks import (
AsyncCallbackManagerForLLMRun,
CallbackManagerForLLMRun,
)
from langchain_core.language_models.chat_models import (
BaseChatModel,
agenerate_from_stream,
generate_from_stream,
)
from langchain_core.messages import (
AIMessage,
AIMessageChunk,
BaseMessage,
ChatMessage,
HumanMessage,
SystemMessage,
)
from langchain_core.outputs import ChatGeneration, ChatGenerationChunk, ChatResult
from langchain_community.llms.friendli import BaseFriendli
def get_role(message: BaseMessage) -> str:
"""Get role of the message.
Args:
message (BaseMessage): The message object.
Raises:
ValueError: Raised when the message is of an unknown type.
Returns:
str: The role of the message.
"""
if isinstance(message, ChatMessage) or isinstance(message, HumanMessage):
return "user"
if isinstance(message, AIMessage):
return "assistant"
if isinstance(message, SystemMessage):
return "system"
raise ValueError(f"Got unknown type {message}")
def get_chat_request(messages: List[BaseMessage]) -> Dict[str, Any]:
"""Get a request of the Friendli chat API.
Args:
messages (List[BaseMessage]): Messages comprising the conversation so far.
Returns:
Dict[str, Any]: The request for the Friendli chat API.
"""
return {
"messages": [
{"role": get_role(message), "content": message.content}
for message in messages
]
}
class ChatFriendli(BaseChatModel, BaseFriendli):
"""Friendli LLM for chat.
``friendli-client`` package should be installed with `pip install friendli-client`.
You must set ``FRIENDLI_TOKEN`` environment variable or provide the value of your
personal access token for the ``friendli_token`` argument.
Example:
.. code-block:: python
from langchain_community.chat_models import FriendliChat
chat = Friendli(
model="llama-2-13b-chat", friendli_token="YOUR FRIENDLI TOKEN"
)
chat.invoke("What is generative AI?")
"""
model: str = "llama-2-13b-chat"
@property
def lc_secrets(self) -> Dict[str, str]:
return {"friendli_token": "FRIENDLI_TOKEN"}
@property
def _default_params(self) -> Dict[str, Any]:
"""Get the default parameters for calling Friendli completions API."""
return {
"frequency_penalty": self.frequency_penalty,
"presence_penalty": self.presence_penalty,
"max_tokens": self.max_tokens,
"stop": self.stop,
"temperature": self.temperature,
"top_p": self.top_p,
}
@property
def _identifying_params(self) -> Dict[str, Any]:
"""Get the identifying parameters."""
return {"model": self.model, **self._default_params}
@property
def _llm_type(self) -> str:
return "friendli-chat"
def _get_invocation_params(
self, stop: Optional[List[str]] = None, **kwargs: Any
) -> Dict[str, Any]:
"""Get the parameters used to invoke the model."""
params = self._default_params
if self.stop is not None and stop is not None:
raise ValueError("`stop` found in both the input and default params.")
elif self.stop is not None:
params["stop"] = self.stop
else:
params["stop"] = stop
return {**params, **kwargs}
def _stream(
self,
messages: List[BaseMessage],
stop: Optional[List[str]] = None,
run_manager: Optional[CallbackManagerForLLMRun] = None,
**kwargs: Any,
) -> Iterator[ChatGenerationChunk]:
params = self._get_invocation_params(stop=stop, **kwargs)
stream = self.client.chat.completions.create(
**get_chat_request(messages), stream=True, model=self.model, **params
)
for chunk in stream:
delta = chunk.choices[0].delta.content
if delta:
yield ChatGenerationChunk(message=AIMessageChunk(content=delta))
if run_manager:
run_manager.on_llm_new_token(delta)
async def _astream(
self,
messages: List[BaseMessage],
stop: Optional[List[str]] = None,
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
**kwargs: Any,
) -> AsyncIterator[ChatGenerationChunk]:
params = self._get_invocation_params(stop=stop, **kwargs)
stream = await self.async_client.chat.completions.create(
**get_chat_request(messages), stream=True, model=self.model, **params
)
async for chunk in stream:
delta = chunk.choices[0].delta.content
if delta:
yield ChatGenerationChunk(message=AIMessageChunk(content=delta))
if run_manager:
await run_manager.on_llm_new_token(delta)
def _generate(
self,
messages: List[BaseMessage],
stop: Optional[List[str]] = None,
run_manager: Optional[CallbackManagerForLLMRun] = None,
**kwargs: Any,
) -> ChatResult:
if self.streaming:
stream_iter = self._stream(
messages, stop=stop, run_manager=run_manager, **kwargs
)
return generate_from_stream(stream_iter)
params = self._get_invocation_params(stop=stop, **kwargs)
response = self.client.chat.completions.create(
messages=[
{
"role": get_role(message),
"content": message.content,
}
for message in messages
],
stream=False,
model=self.model,
**params,
)
message = AIMessage(content=response.choices[0].message.content)
return ChatResult(generations=[ChatGeneration(message=message)])
async def _agenerate(
self,
messages: List[BaseMessage],
stop: Optional[List[str]] = None,
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
**kwargs: Any,
) -> ChatResult:
if self.streaming:
stream_iter = self._astream(
messages, stop=stop, run_manager=run_manager, **kwargs
)
return await agenerate_from_stream(stream_iter)
params = self._get_invocation_params(stop=stop, **kwargs)
response = await self.async_client.chat.completions.create(
messages=[
{
"role": get_role(message),
"content": message.content,
}
for message in messages
],
stream=False,
model=self.model,
**params,
)
message = AIMessage(content=response.choices[0].message.content)
return ChatResult(generations=[ChatGeneration(message=message)])
|