Source code for langchain_community.llms.modal
import logging
from typing import Any, Dict, List, Mapping, Optional
import requests
from langchain_core.callbacks import CallbackManagerForLLMRun
from langchain_core.language_models.llms import LLM
from langchain_core.pydantic_v1 import Extra, Field, root_validator
from langchain_community.llms.utils import enforce_stop_tokens
logger = logging.getLogger(__name__)
[docs]class Modal(LLM):
"""大型语言模型。
要使用,您应该已经安装了``modal-client`` python包。
可以传递任何可以传递给调用的参数,即使在此类中没有明确保存。
示例:
.. code-block:: python
from langchain_community.llms import Modal
modal = Modal(endpoint_url="")"""
endpoint_url: str = ""
"""使用的模型端点"""
model_kwargs: Dict[str, Any] = Field(default_factory=dict)
"""包含在`create`调用中有效但未明确指定的任何模型参数。"""
class Config:
"""这是用于pydantic配置的设置。"""
extra = Extra.forbid
@root_validator(pre=True)
def build_extra(cls, values: Dict[str, Any]) -> Dict[str, Any]:
"""从传入的额外参数构建额外的kwargs。"""
all_required_field_names = {field.alias for field in cls.__fields__.values()}
extra = values.get("model_kwargs", {})
for field_name in list(values):
if field_name not in all_required_field_names:
if field_name in extra:
raise ValueError(f"Found {field_name} supplied twice.")
logger.warning(
f"""{field_name} was transferred to model_kwargs.
Please confirm that {field_name} is what you intended."""
)
extra[field_name] = values.pop(field_name)
values["model_kwargs"] = extra
return values
@property
def _identifying_params(self) -> Mapping[str, Any]:
"""获取识别参数。"""
return {
**{"endpoint_url": self.endpoint_url},
**{"model_kwargs": self.model_kwargs},
}
@property
def _llm_type(self) -> str:
"""llm的返回类型。"""
return "modal"
def _call(
self,
prompt: str,
stop: Optional[List[str]] = None,
run_manager: Optional[CallbackManagerForLLMRun] = None,
**kwargs: Any,
) -> str:
"""调用模态端点。"""
params = self.model_kwargs or {}
params = {**params, **kwargs}
response = requests.post(
url=self.endpoint_url,
headers={
"Content-Type": "application/json",
},
json={"prompt": prompt, **params},
)
try:
if prompt in response.json()["prompt"]:
response_json = response.json()
except KeyError:
raise KeyError("LangChain requires 'prompt' key in response.")
text = response_json["prompt"]
if stop is not None:
# I believe this is required since the stop tokens
# are not enforced by the model parameters
text = enforce_stop_tokens(text, stop)
return text