Kye
1 year ago
parent
563827fc1a
commit
1a79475c11
File diff suppressed because one or more lines are too long
File diff suppressed because one or more lines are too long
@ -0,0 +1,9 @@
|
||||
from swarms.models.openai_models import OpenAIChat
|
||||
|
||||
openai = OpenAIChat(
|
||||
openai_api_key="",
|
||||
verbose=False
|
||||
)
|
||||
|
||||
chat = openai("Are quantum fields everywhere?")
|
||||
print(chat)
|
||||
@ -1,9 +1,4 @@
|
||||
# prompts
|
||||
from swarms.models.anthropic import Anthropic
|
||||
|
||||
# from swarms.models.palm import GooglePalm
|
||||
from swarms.models.petals import Petals
|
||||
|
||||
# from swarms.models.chat_openai import OpenAIChat
|
||||
from swarms.prompts.debate import *
|
||||
from swarms.models.mistral import Mistral
|
||||
from swarms.models.openai_llm import OpenAIModel
|
||||
|
||||
@ -0,0 +1,598 @@
|
||||
import importlib.util
|
||||
import json
|
||||
import os
|
||||
from dataclasses import dataclass
|
||||
from typing import Dict
|
||||
|
||||
from huggingface_hub import hf_hub_download, list_spaces
|
||||
from transformers.tools.base import (
|
||||
TASK_MAPPING,
|
||||
TOOL_CONFIG_FILE,
|
||||
Tool,
|
||||
load_tool,
|
||||
supports_remote,
|
||||
)
|
||||
from transformers.tools.prompts import CHAT_MESSAGE_PROMPT, download_prompt
|
||||
from transformers.tools.python_interpreter import evaluate
|
||||
from transformers.utils import is_offline_mode, is_openai_available, logging
|
||||
|
||||
# utils
|
||||
logger = logging.get_logger(__name__)
|
||||
|
||||
|
||||
if is_openai_available():
|
||||
import openai
|
||||
|
||||
else:
|
||||
StoppingCriteria = object
|
||||
|
||||
_tools_are_initialized = False
|
||||
|
||||
|
||||
BASE_PYTHON_TOOLS = {
|
||||
"print": print,
|
||||
"range": range,
|
||||
"float": float,
|
||||
"int": int,
|
||||
"bool": bool,
|
||||
"str": str,
|
||||
}
|
||||
|
||||
|
||||
@dataclass
|
||||
class PreTool:
|
||||
task: str
|
||||
description: str
|
||||
repo_id: str
|
||||
|
||||
|
||||
HUGGINGFACE_DEFAULT_TOOLS = {}
|
||||
|
||||
|
||||
HUGGINGFACE_DEFAULT_TOOLS_FROM_HUB = [
|
||||
"image-transformation",
|
||||
"text-download",
|
||||
"text-to-image",
|
||||
"text-to-video",
|
||||
]
|
||||
|
||||
|
||||
def get_remote_tools(organization="huggingface-tools"):
|
||||
if is_offline_mode():
|
||||
logger.info("You are in offline mode, so remote tools are not available.")
|
||||
return {}
|
||||
|
||||
spaces = list_spaces(author=organization)
|
||||
tools = {}
|
||||
for space_info in spaces:
|
||||
repo_id = space_info.id
|
||||
resolved_config_file = hf_hub_download(
|
||||
repo_id, TOOL_CONFIG_FILE, repo_type="space"
|
||||
)
|
||||
with open(resolved_config_file, encoding="utf-8") as reader:
|
||||
config = json.load(reader)
|
||||
|
||||
task = repo_id.split("/")[-1]
|
||||
tools[config["name"]] = PreTool(
|
||||
task=task, description=config["description"], repo_id=repo_id
|
||||
)
|
||||
|
||||
return tools
|
||||
|
||||
|
||||
def _setup_default_tools():
|
||||
global HUGGINGFACE_DEFAULT_TOOLS
|
||||
global _tools_are_initialized
|
||||
|
||||
if _tools_are_initialized:
|
||||
return
|
||||
|
||||
main_module = importlib.import_module("transformers")
|
||||
tools_module = main_module.tools
|
||||
|
||||
remote_tools = get_remote_tools()
|
||||
for task_name, tool_class_name in TASK_MAPPING.items():
|
||||
tool_class = getattr(tools_module, tool_class_name)
|
||||
description = tool_class.description
|
||||
HUGGINGFACE_DEFAULT_TOOLS[tool_class.name] = PreTool(
|
||||
task=task_name, description=description, repo_id=None
|
||||
)
|
||||
|
||||
if not is_offline_mode():
|
||||
for task_name in HUGGINGFACE_DEFAULT_TOOLS_FROM_HUB:
|
||||
found = False
|
||||
for tool_name, tool in remote_tools.items():
|
||||
if tool.task == task_name:
|
||||
HUGGINGFACE_DEFAULT_TOOLS[tool_name] = tool
|
||||
found = True
|
||||
break
|
||||
|
||||
if not found:
|
||||
raise ValueError(f"{task_name} is not implemented on the Hub.")
|
||||
|
||||
_tools_are_initialized = True
|
||||
|
||||
|
||||
def resolve_tools(code, toolbox, remote=False, cached_tools=None):
|
||||
if cached_tools is None:
|
||||
resolved_tools = BASE_PYTHON_TOOLS.copy()
|
||||
else:
|
||||
resolved_tools = cached_tools
|
||||
for name, tool in toolbox.items():
|
||||
if name not in code or name in resolved_tools:
|
||||
continue
|
||||
|
||||
if isinstance(tool, Tool):
|
||||
resolved_tools[name] = tool
|
||||
else:
|
||||
task_or_repo_id = tool.task if tool.repo_id is None else tool.repo_id
|
||||
_remote = remote and supports_remote(task_or_repo_id)
|
||||
resolved_tools[name] = load_tool(task_or_repo_id, remote=_remote)
|
||||
|
||||
return resolved_tools
|
||||
|
||||
|
||||
def get_tool_creation_code(code, toolbox, remote=False):
|
||||
code_lines = ["from transformers import load_tool", ""]
|
||||
for name, tool in toolbox.items():
|
||||
if name not in code or isinstance(tool, Tool):
|
||||
continue
|
||||
|
||||
task_or_repo_id = tool.task if tool.repo_id is None else tool.repo_id
|
||||
line = f'{name} = load_tool("{task_or_repo_id}"'
|
||||
if remote:
|
||||
line += ", remote=True"
|
||||
line += ")"
|
||||
code_lines.append(line)
|
||||
|
||||
return "\n".join(code_lines) + "\n"
|
||||
|
||||
|
||||
def clean_code_for_chat(result):
|
||||
lines = result.split("\n")
|
||||
idx = 0
|
||||
while idx < len(lines) and not lines[idx].lstrip().startswith("```"):
|
||||
idx += 1
|
||||
explanation = "\n".join(lines[:idx]).strip()
|
||||
if idx == len(lines):
|
||||
return explanation, None
|
||||
|
||||
idx += 1
|
||||
start_idx = idx
|
||||
while not lines[idx].lstrip().startswith("```"):
|
||||
idx += 1
|
||||
code = "\n".join(lines[start_idx:idx]).strip()
|
||||
|
||||
return explanation, code
|
||||
|
||||
|
||||
def clean_code_for_run(result):
|
||||
result = f"I will use the following {result}"
|
||||
explanation, code = result.split("Answer:")
|
||||
explanation = explanation.strip()
|
||||
code = code.strip()
|
||||
|
||||
code_lines = code.split("\n")
|
||||
if code_lines[0] in ["```", "```py", "```python"]:
|
||||
code_lines = code_lines[1:]
|
||||
if code_lines[-1] == "```":
|
||||
code_lines = code_lines[:-1]
|
||||
code = "\n".join(code_lines)
|
||||
|
||||
return explanation, code
|
||||
|
||||
|
||||
class Agent:
|
||||
"""
|
||||
Base class for all agents which contains the main API methods.
|
||||
|
||||
Args:
|
||||
chat_prompt_template (`str`, *optional*):
|
||||
Pass along your own prompt if you want to override the default template for the `chat` method. Can be the
|
||||
actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named
|
||||
`chat_prompt_template.txt` in this repo in this case.
|
||||
run_prompt_template (`str`, *optional*):
|
||||
Pass along your own prompt if you want to override the default template for the `run` method. Can be the
|
||||
actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named
|
||||
`run_prompt_template.txt` in this repo in this case.
|
||||
additional_tools ([`Tool`], list of tools or dictionary with tool values, *optional*):
|
||||
Any additional tools to include on top of the default ones. If you pass along a tool with the same name as
|
||||
one of the default tools, that default tool will be overridden.
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self, chat_prompt_template=None, run_prompt_template=None, additional_tools=None
|
||||
):
|
||||
_setup_default_tools()
|
||||
|
||||
agent_name = self.__class__.__name__
|
||||
self.chat_prompt_template = download_prompt(
|
||||
chat_prompt_template, agent_name, mode="chat"
|
||||
)
|
||||
self.run_prompt_template = download_prompt(
|
||||
run_prompt_template, agent_name, mode="run"
|
||||
)
|
||||
self._toolbox = HUGGINGFACE_DEFAULT_TOOLS.copy()
|
||||
self.log = print
|
||||
if additional_tools is not None:
|
||||
if isinstance(additional_tools, (list, tuple)):
|
||||
additional_tools = {t.name: t for t in additional_tools}
|
||||
elif not isinstance(additional_tools, dict):
|
||||
additional_tools = {additional_tools.name: additional_tools}
|
||||
|
||||
replacements = {
|
||||
name: tool
|
||||
for name, tool in additional_tools.items()
|
||||
if name in HUGGINGFACE_DEFAULT_TOOLS
|
||||
}
|
||||
self._toolbox.update(additional_tools)
|
||||
if len(replacements) > 1:
|
||||
names = "\n".join([f"- {n}: {t}" for n, t in replacements.items()])
|
||||
logger.warning(
|
||||
f"The following tools have been replaced by the ones provided in `additional_tools`:\n{names}."
|
||||
)
|
||||
elif len(replacements) == 1:
|
||||
name = list(replacements.keys())[0]
|
||||
logger.warning(
|
||||
f"{name} has been replaced by {replacements[name]} as provided in `additional_tools`."
|
||||
)
|
||||
|
||||
self.prepare_for_new_chat()
|
||||
|
||||
@property
|
||||
def toolbox(self) -> Dict[str, Tool]:
|
||||
"""Get all tool currently available to the agent"""
|
||||
return self._toolbox
|
||||
|
||||
def format_prompt(self, task, chat_mode=False):
|
||||
description = "\n".join(
|
||||
[f"- {name}: {tool.description}" for name, tool in self.toolbox.items()]
|
||||
)
|
||||
if chat_mode:
|
||||
if self.chat_history is None:
|
||||
prompt = self.chat_prompt_template.replace("<<all_tools>>", description)
|
||||
else:
|
||||
prompt = self.chat_history
|
||||
prompt += CHAT_MESSAGE_PROMPT.replace("<<task>>", task)
|
||||
else:
|
||||
prompt = self.run_prompt_template.replace("<<all_tools>>", description)
|
||||
prompt = prompt.replace("<<prompt>>", task)
|
||||
return prompt
|
||||
|
||||
def set_stream(self, streamer):
|
||||
"""
|
||||
Set the function use to stream results (which is `print` by default).
|
||||
|
||||
Args:
|
||||
streamer (`callable`): The function to call when streaming results from the LLM.
|
||||
"""
|
||||
self.log = streamer
|
||||
|
||||
def chat(self, task, *, return_code=False, remote=False, **kwargs):
|
||||
"""
|
||||
Sends a new request to the agent in a chat. Will use the previous ones in its history.
|
||||
|
||||
Args:
|
||||
task (`str`): The task to perform
|
||||
return_code (`bool`, *optional*, defaults to `False`):
|
||||
Whether to just return code and not evaluate it.
|
||||
remote (`bool`, *optional*, defaults to `False`):
|
||||
Whether or not to use remote tools (inference endpoints) instead of local ones.
|
||||
kwargs (additional keyword arguments, *optional*):
|
||||
Any keyword argument to send to the agent when evaluating the code.
|
||||
|
||||
Example:
|
||||
|
||||
```py
|
||||
from transformers import HfAgent
|
||||
|
||||
agent = HfAgent("https://api-inference.huggingface.co/models/bigcode/starcoder")
|
||||
agent.chat("Draw me a picture of rivers and lakes")
|
||||
|
||||
agent.chat("Transform the picture so that there is a rock in there")
|
||||
```
|
||||
"""
|
||||
prompt = self.format_prompt(task, chat_mode=True)
|
||||
result = self.generate_one(prompt, stop=["Human:", "====="])
|
||||
self.chat_history = prompt + result.strip() + "\n"
|
||||
explanation, code = clean_code_for_chat(result)
|
||||
|
||||
self.log(f"==Explanation from the agent==\n{explanation}")
|
||||
|
||||
if code is not None:
|
||||
self.log(f"\n\n==Code generated by the agent==\n{code}")
|
||||
if not return_code:
|
||||
self.log("\n\n==Result==")
|
||||
self.cached_tools = resolve_tools(
|
||||
code, self.toolbox, remote=remote, cached_tools=self.cached_tools
|
||||
)
|
||||
self.chat_state.update(kwargs)
|
||||
return evaluate(
|
||||
code, self.cached_tools, self.chat_state, chat_mode=True
|
||||
)
|
||||
else:
|
||||
tool_code = get_tool_creation_code(code, self.toolbox, remote=remote)
|
||||
return f"{tool_code}\n{code}"
|
||||
|
||||
def prepare_for_new_chat(self):
|
||||
"""
|
||||
Clears the history of prior calls to [`~Agent.chat`].
|
||||
"""
|
||||
self.chat_history = None
|
||||
self.chat_state = {}
|
||||
self.cached_tools = None
|
||||
|
||||
def run(self, task, *, return_code=False, remote=False, **kwargs):
|
||||
"""
|
||||
Sends a request to the agent.
|
||||
|
||||
Args:
|
||||
task (`str`): The task to perform
|
||||
return_code (`bool`, *optional*, defaults to `False`):
|
||||
Whether to just return code and not evaluate it.
|
||||
remote (`bool`, *optional*, defaults to `False`):
|
||||
Whether or not to use remote tools (inference endpoints) instead of local ones.
|
||||
kwargs (additional keyword arguments, *optional*):
|
||||
Any keyword argument to send to the agent when evaluating the code.
|
||||
|
||||
Example:
|
||||
|
||||
```py
|
||||
from transformers import HfAgent
|
||||
|
||||
agent = HfAgent("https://api-inference.huggingface.co/models/bigcode/starcoder")
|
||||
agent.run("Draw me a picture of rivers and lakes")
|
||||
```
|
||||
"""
|
||||
prompt = self.format_prompt(task)
|
||||
result = self.generate_one(prompt, stop=["Task:"])
|
||||
explanation, code = clean_code_for_run(result)
|
||||
|
||||
self.log(f"==Explanation from the agent==\n{explanation}")
|
||||
|
||||
self.log(f"\n\n==Code generated by the agent==\n{code}")
|
||||
if not return_code:
|
||||
self.log("\n\n==Result==")
|
||||
self.cached_tools = resolve_tools(
|
||||
code, self.toolbox, remote=remote, cached_tools=self.cached_tools
|
||||
)
|
||||
return evaluate(code, self.cached_tools, state=kwargs.copy())
|
||||
else:
|
||||
tool_code = get_tool_creation_code(code, self.toolbox, remote=remote)
|
||||
return f"{tool_code}\n{code}"
|
||||
|
||||
def generate_one(self, prompt, stop):
|
||||
# This is the method to implement in your custom agent.
|
||||
raise NotImplementedError
|
||||
|
||||
def generate_many(self, prompts, stop):
|
||||
# Override if you have a way to do batch generation faster than one by one
|
||||
return [self.generate_one(prompt, stop) for prompt in prompts]
|
||||
|
||||
|
||||
class OpenAIModel(Agent):
|
||||
"""
|
||||
Agent that uses the openai API to generate code.
|
||||
|
||||
<Tip warning={true}>
|
||||
|
||||
The openAI models are used in generation mode, so even for the `chat()` API, it's better to use models like
|
||||
`"text-davinci-003"` over the chat-GPT variant. Proper support for chat-GPT models will come in a next version.
|
||||
|
||||
</Tip>
|
||||
|
||||
Args:
|
||||
model (`str`, *optional*, defaults to `"text-davinci-003"`):
|
||||
The name of the OpenAI model to use.
|
||||
api_key (`str`, *optional*):
|
||||
The API key to use. If unset, will look for the environment variable `"OPENAI_API_KEY"`.
|
||||
chat_prompt_template (`str`, *optional*):
|
||||
Pass along your own prompt if you want to override the default template for the `chat` method. Can be the
|
||||
actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named
|
||||
`chat_prompt_template.txt` in this repo in this case.
|
||||
run_prompt_template (`str`, *optional*):
|
||||
Pass along your own prompt if you want to override the default template for the `run` method. Can be the
|
||||
actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named
|
||||
`run_prompt_template.txt` in this repo in this case.
|
||||
additional_tools ([`Tool`], list of tools or dictionary with tool values, *optional*):
|
||||
Any additional tools to include on top of the default ones. If you pass along a tool with the same name as
|
||||
one of the default tools, that default tool will be overridden.
|
||||
|
||||
Example:
|
||||
|
||||
```py
|
||||
from swarms.models import OpenAI
|
||||
|
||||
agent = OpenAiAgent(model="text-davinci-003", api_key=xxx)
|
||||
agent.run("Is the following `text` (in Spanish) positive or negative?", text="¡Este es un API muy agradable!")
|
||||
```
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
model="text-davinci-003",
|
||||
api_key=None,
|
||||
chat_prompt_template=None,
|
||||
run_prompt_template=None,
|
||||
additional_tools=None,
|
||||
):
|
||||
if not is_openai_available():
|
||||
raise ImportError(
|
||||
"Using `OpenAiAgent` requires `openai`: `pip install openai`."
|
||||
)
|
||||
|
||||
if api_key is None:
|
||||
api_key = os.environ.get("OPENAI_API_KEY", None)
|
||||
if api_key is None:
|
||||
raise ValueError(
|
||||
"You need an openai key to use `OpenAIAgent`. You can get one here: Get one here "
|
||||
"https://openai.com/api/`. If you have one, set it in your env with `os.environ['OPENAI_API_KEY'] = "
|
||||
"xxx."
|
||||
)
|
||||
else:
|
||||
openai.api_key = api_key
|
||||
self.model = model
|
||||
super().__init__(
|
||||
chat_prompt_template=chat_prompt_template,
|
||||
run_prompt_template=run_prompt_template,
|
||||
additional_tools=additional_tools,
|
||||
)
|
||||
|
||||
def generate_many(self, prompts, stop):
|
||||
if "gpt" in self.model:
|
||||
return [self._chat_generate(prompt, stop) for prompt in prompts]
|
||||
else:
|
||||
return self._completion_generate(prompts, stop)
|
||||
|
||||
def generate_one(self, prompt, stop):
|
||||
if "gpt" in self.model:
|
||||
return self._chat_generate(prompt, stop)
|
||||
else:
|
||||
return self._completion_generate([prompt], stop)[0]
|
||||
|
||||
def _chat_generate(self, prompt, stop):
|
||||
result = openai.ChatCompletion.create(
|
||||
model=self.model,
|
||||
messages=[{"role": "user", "content": prompt}],
|
||||
temperature=0,
|
||||
stop=stop,
|
||||
)
|
||||
return result["choices"][0]["message"]["content"]
|
||||
|
||||
def _completion_generate(self, prompts, stop):
|
||||
result = openai.Completion.create(
|
||||
model=self.model,
|
||||
prompt=prompts,
|
||||
temperature=0,
|
||||
stop=stop,
|
||||
max_tokens=200,
|
||||
)
|
||||
return [answer["text"] for answer in result["choices"]]
|
||||
|
||||
|
||||
class AzureOpenAI(Agent):
|
||||
"""
|
||||
Agent that uses Azure OpenAI to generate code. See the [official
|
||||
documentation](https://learn.microsoft.com/en-us/azure/cognitive-services/openai/) to learn how to deploy an openAI
|
||||
model on Azure
|
||||
|
||||
<Tip warning={true}>
|
||||
|
||||
The openAI models are used in generation mode, so even for the `chat()` API, it's better to use models like
|
||||
`"text-davinci-003"` over the chat-GPT variant. Proper support for chat-GPT models will come in a next version.
|
||||
|
||||
</Tip>
|
||||
|
||||
Args:
|
||||
deployment_id (`str`):
|
||||
The name of the deployed Azure openAI model to use.
|
||||
api_key (`str`, *optional*):
|
||||
The API key to use. If unset, will look for the environment variable `"AZURE_OPENAI_API_KEY"`.
|
||||
resource_name (`str`, *optional*):
|
||||
The name of your Azure OpenAI Resource. If unset, will look for the environment variable
|
||||
`"AZURE_OPENAI_RESOURCE_NAME"`.
|
||||
api_version (`str`, *optional*, default to `"2022-12-01"`):
|
||||
The API version to use for this agent.
|
||||
is_chat_mode (`bool`, *optional*):
|
||||
Whether you are using a completion model or a chat model (see note above, chat models won't be as
|
||||
efficient). Will default to `gpt` being in the `deployment_id` or not.
|
||||
chat_prompt_template (`str`, *optional*):
|
||||
Pass along your own prompt if you want to override the default template for the `chat` method. Can be the
|
||||
actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named
|
||||
`chat_prompt_template.txt` in this repo in this case.
|
||||
run_prompt_template (`str`, *optional*):
|
||||
Pass along your own prompt if you want to override the default template for the `run` method. Can be the
|
||||
actual prompt template or a repo ID (on the Hugging Face Hub). The prompt should be in a file named
|
||||
`run_prompt_template.txt` in this repo in this case.
|
||||
additional_tools ([`Tool`], list of tools or dictionary with tool values, *optional*):
|
||||
Any additional tools to include on top of the default ones. If you pass along a tool with the same name as
|
||||
one of the default tools, that default tool will be overridden.
|
||||
|
||||
Example:
|
||||
|
||||
```py
|
||||
from transformers import AzureOpenAiAgent
|
||||
|
||||
agent = AzureAiAgent(deployment_id="Davinci-003", api_key=xxx, resource_name=yyy)
|
||||
agent.run("Is the following `text` (in Spanish) positive or negative?", text="¡Este es un API muy agradable!")
|
||||
```
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
deployment_id,
|
||||
api_key=None,
|
||||
resource_name=None,
|
||||
api_version="2022-12-01",
|
||||
is_chat_model=None,
|
||||
chat_prompt_template=None,
|
||||
run_prompt_template=None,
|
||||
additional_tools=None,
|
||||
):
|
||||
if not is_openai_available():
|
||||
raise ImportError(
|
||||
"Using `OpenAiAgent` requires `openai`: `pip install openai`."
|
||||
)
|
||||
|
||||
self.deployment_id = deployment_id
|
||||
openai.api_type = "azure"
|
||||
if api_key is None:
|
||||
api_key = os.environ.get("AZURE_OPENAI_API_KEY", None)
|
||||
if api_key is None:
|
||||
raise ValueError(
|
||||
"You need an Azure openAI key to use `AzureOpenAIAgent`. If you have one, set it in your env with "
|
||||
"`os.environ['AZURE_OPENAI_API_KEY'] = xxx."
|
||||
)
|
||||
else:
|
||||
openai.api_key = api_key
|
||||
if resource_name is None:
|
||||
resource_name = os.environ.get("AZURE_OPENAI_RESOURCE_NAME", None)
|
||||
if resource_name is None:
|
||||
raise ValueError(
|
||||
"You need a resource_name to use `AzureOpenAIAgent`. If you have one, set it in your env with "
|
||||
"`os.environ['AZURE_OPENAI_RESOURCE_NAME'] = xxx."
|
||||
)
|
||||
else:
|
||||
openai.api_base = f"https://{resource_name}.openai.azure.com"
|
||||
openai.api_version = api_version
|
||||
|
||||
if is_chat_model is None:
|
||||
is_chat_model = "gpt" in deployment_id.lower()
|
||||
self.is_chat_model = is_chat_model
|
||||
|
||||
super().__init__(
|
||||
chat_prompt_template=chat_prompt_template,
|
||||
run_prompt_template=run_prompt_template,
|
||||
additional_tools=additional_tools,
|
||||
)
|
||||
|
||||
def generate_many(self, prompts, stop):
|
||||
if self.is_chat_model:
|
||||
return [self._chat_generate(prompt, stop) for prompt in prompts]
|
||||
else:
|
||||
return self._completion_generate(prompts, stop)
|
||||
|
||||
def generate_one(self, prompt, stop):
|
||||
if self.is_chat_model:
|
||||
return self._chat_generate(prompt, stop)
|
||||
else:
|
||||
return self._completion_generate([prompt], stop)[0]
|
||||
|
||||
def _chat_generate(self, prompt, stop):
|
||||
result = openai.ChatCompletion.create(
|
||||
engine=self.deployment_id,
|
||||
messages=[{"role": "user", "content": prompt}],
|
||||
temperature=0,
|
||||
stop=stop,
|
||||
)
|
||||
return result["choices"][0]["message"]["content"]
|
||||
|
||||
def _completion_generate(self, prompts, stop):
|
||||
result = openai.Completion.create(
|
||||
engine=self.deployment_id,
|
||||
prompt=prompts,
|
||||
temperature=0,
|
||||
stop=stop,
|
||||
max_tokens=200,
|
||||
)
|
||||
return [answer["text"] for answer in result["choices"]]
|
||||
@ -0,0 +1,958 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import sys
|
||||
import warnings
|
||||
from typing import (
|
||||
AbstractSet,
|
||||
Any,
|
||||
AsyncIterator,
|
||||
Callable,
|
||||
Collection,
|
||||
Dict,
|
||||
Iterator,
|
||||
List,
|
||||
Literal,
|
||||
Mapping,
|
||||
Optional,
|
||||
Set,
|
||||
Tuple,
|
||||
Union,
|
||||
)
|
||||
|
||||
from langchain.callbacks.manager import (
|
||||
AsyncCallbackManagerForLLMRun,
|
||||
CallbackManagerForLLMRun,
|
||||
)
|
||||
from langchain.llms.base import BaseLLM, create_base_retry_decorator
|
||||
from langchain.pydantic_v1 import Field, root_validator
|
||||
from langchain.schema import Generation, LLMResult
|
||||
from langchain.schema.output import GenerationChunk
|
||||
from langchain.utils import get_from_dict_or_env, get_pydantic_field_names
|
||||
from langchain.utils.utils import build_extra_kwargs
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
def update_token_usage(
|
||||
keys: Set[str], response: Dict[str, Any], token_usage: Dict[str, Any]
|
||||
) -> None:
|
||||
"""Update token usage."""
|
||||
_keys_to_use = keys.intersection(response["usage"])
|
||||
for _key in _keys_to_use:
|
||||
if _key not in token_usage:
|
||||
token_usage[_key] = response["usage"][_key]
|
||||
else:
|
||||
token_usage[_key] += response["usage"][_key]
|
||||
|
||||
|
||||
def _stream_response_to_generation_chunk(
|
||||
stream_response: Dict[str, Any],
|
||||
) -> GenerationChunk:
|
||||
"""Convert a stream response to a generation chunk."""
|
||||
return GenerationChunk(
|
||||
text=stream_response["choices"][0]["text"],
|
||||
generation_info=dict(
|
||||
finish_reason=stream_response["choices"][0].get("finish_reason", None),
|
||||
logprobs=stream_response["choices"][0].get("logprobs", None),
|
||||
),
|
||||
)
|
||||
|
||||
|
||||
def _update_response(response: Dict[str, Any], stream_response: Dict[str, Any]) -> None:
|
||||
"""Update response from the stream response."""
|
||||
response["choices"][0]["text"] += stream_response["choices"][0]["text"]
|
||||
response["choices"][0]["finish_reason"] = stream_response["choices"][0].get(
|
||||
"finish_reason", None
|
||||
)
|
||||
response["choices"][0]["logprobs"] = stream_response["choices"][0]["logprobs"]
|
||||
|
||||
|
||||
def _streaming_response_template() -> Dict[str, Any]:
|
||||
return {
|
||||
"choices": [
|
||||
{
|
||||
"text": "",
|
||||
"finish_reason": None,
|
||||
"logprobs": None,
|
||||
}
|
||||
]
|
||||
}
|
||||
|
||||
|
||||
def _create_retry_decorator(
|
||||
llm: Union[BaseOpenAI, OpenAIChat],
|
||||
run_manager: Optional[
|
||||
Union[AsyncCallbackManagerForLLMRun, CallbackManagerForLLMRun]
|
||||
] = None,
|
||||
) -> Callable[[Any], Any]:
|
||||
import openai
|
||||
|
||||
errors = [
|
||||
openai.error.Timeout,
|
||||
openai.error.APIError,
|
||||
openai.error.APIConnectionError,
|
||||
openai.error.RateLimitError,
|
||||
openai.error.ServiceUnavailableError,
|
||||
]
|
||||
return create_base_retry_decorator(
|
||||
error_types=errors, max_retries=llm.max_retries, run_manager=run_manager
|
||||
)
|
||||
|
||||
|
||||
def completion_with_retry(
|
||||
llm: Union[BaseOpenAI, OpenAIChat],
|
||||
run_manager: Optional[CallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> Any:
|
||||
"""Use tenacity to retry the completion call."""
|
||||
retry_decorator = _create_retry_decorator(llm, run_manager=run_manager)
|
||||
|
||||
@retry_decorator
|
||||
def _completion_with_retry(**kwargs: Any) -> Any:
|
||||
return llm.client.create(**kwargs)
|
||||
|
||||
return _completion_with_retry(**kwargs)
|
||||
|
||||
|
||||
async def acompletion_with_retry(
|
||||
llm: Union[BaseOpenAI, OpenAIChat],
|
||||
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> Any:
|
||||
"""Use tenacity to retry the async completion call."""
|
||||
retry_decorator = _create_retry_decorator(llm, run_manager=run_manager)
|
||||
|
||||
@retry_decorator
|
||||
async def _completion_with_retry(**kwargs: Any) -> Any:
|
||||
# Use OpenAI's async api https://github.com/openai/openai-python#async-api
|
||||
return await llm.client.acreate(**kwargs)
|
||||
|
||||
return await _completion_with_retry(**kwargs)
|
||||
|
||||
|
||||
class BaseOpenAI(BaseLLM):
|
||||
"""Base OpenAI large language model class."""
|
||||
|
||||
@property
|
||||
def lc_secrets(self) -> Dict[str, str]:
|
||||
return {"openai_api_key": "OPENAI_API_KEY"}
|
||||
|
||||
@property
|
||||
def lc_attributes(self) -> Dict[str, Any]:
|
||||
attributes: Dict[str, Any] = {}
|
||||
if self.openai_api_base != "":
|
||||
attributes["openai_api_base"] = self.openai_api_base
|
||||
|
||||
if self.openai_organization != "":
|
||||
attributes["openai_organization"] = self.openai_organization
|
||||
|
||||
if self.openai_proxy != "":
|
||||
attributes["openai_proxy"] = self.openai_proxy
|
||||
|
||||
return attributes
|
||||
|
||||
@classmethod
|
||||
def is_lc_serializable(cls) -> bool:
|
||||
return True
|
||||
|
||||
client: Any = None #: :meta private:
|
||||
model_name: str = Field(default="text-davinci-003", alias="model")
|
||||
"""Model name to use."""
|
||||
temperature: float = 0.7
|
||||
"""What sampling temperature to use."""
|
||||
max_tokens: int = 256
|
||||
"""The maximum number of tokens to generate in the completion.
|
||||
-1 returns as many tokens as possible given the prompt and
|
||||
the models maximal context size."""
|
||||
top_p: float = 1
|
||||
"""Total probability mass of tokens to consider at each step."""
|
||||
frequency_penalty: float = 0
|
||||
"""Penalizes repeated tokens according to frequency."""
|
||||
presence_penalty: float = 0
|
||||
"""Penalizes repeated tokens."""
|
||||
n: int = 1
|
||||
"""How many completions to generate for each prompt."""
|
||||
best_of: int = 1
|
||||
"""Generates best_of completions server-side and returns the "best"."""
|
||||
model_kwargs: Dict[str, Any] = Field(default_factory=dict)
|
||||
"""Holds any model parameters valid for `create` call not explicitly specified."""
|
||||
openai_api_key: Optional[str] = None
|
||||
openai_api_base: Optional[str] = None
|
||||
openai_organization: Optional[str] = None
|
||||
# to support explicit proxy for OpenAI
|
||||
openai_proxy: Optional[str] = None
|
||||
batch_size: int = 20
|
||||
"""Batch size to use when passing multiple documents to generate."""
|
||||
request_timeout: Optional[Union[float, Tuple[float, float]]] = None
|
||||
"""Timeout for requests to OpenAI completion API. Default is 600 seconds."""
|
||||
logit_bias: Optional[Dict[str, float]] = Field(default_factory=dict)
|
||||
"""Adjust the probability of specific tokens being generated."""
|
||||
max_retries: int = 6
|
||||
"""Maximum number of retries to make when generating."""
|
||||
streaming: bool = False
|
||||
"""Whether to stream the results or not."""
|
||||
allowed_special: Union[Literal["all"], AbstractSet[str]] = set()
|
||||
"""Set of special tokens that are allowed。"""
|
||||
disallowed_special: Union[Literal["all"], Collection[str]] = "all"
|
||||
"""Set of special tokens that are not allowed。"""
|
||||
tiktoken_model_name: Optional[str] = None
|
||||
"""The model name to pass to tiktoken when using this class.
|
||||
Tiktoken is used to count the number of tokens in documents to constrain
|
||||
them to be under a certain limit. By default, when set to None, this will
|
||||
be the same as the embedding model name. However, there are some cases
|
||||
where you may want to use this Embedding class with a model name not
|
||||
supported by tiktoken. This can include when using Azure embeddings or
|
||||
when using one of the many model providers that expose an OpenAI-like
|
||||
API but with different models. In those cases, in order to avoid erroring
|
||||
when tiktoken is called, you can specify a model name to use here."""
|
||||
|
||||
def __new__(cls, **data: Any) -> Union[OpenAIChat, BaseOpenAI]: # type: ignore
|
||||
"""Initialize the OpenAI object."""
|
||||
model_name = data.get("model_name", "")
|
||||
return super().__new__(cls)
|
||||
|
||||
class Config:
|
||||
"""Configuration for this pydantic object."""
|
||||
|
||||
allow_population_by_field_name = True
|
||||
|
||||
@root_validator(pre=True)
|
||||
def build_extra(cls, values: Dict[str, Any]) -> Dict[str, Any]:
|
||||
"""Build extra kwargs from additional params that were passed in."""
|
||||
all_required_field_names = get_pydantic_field_names(cls)
|
||||
extra = values.get("model_kwargs", {})
|
||||
values["model_kwargs"] = build_extra_kwargs(
|
||||
extra, values, all_required_field_names
|
||||
)
|
||||
return values
|
||||
|
||||
@root_validator()
|
||||
def validate_environment(cls, values: Dict) -> Dict:
|
||||
"""Validate that api key and python package exists in environment."""
|
||||
values["openai_api_key"] = get_from_dict_or_env(
|
||||
values, "openai_api_key", "OPENAI_API_KEY"
|
||||
)
|
||||
values["openai_api_base"] = get_from_dict_or_env(
|
||||
values,
|
||||
"openai_api_base",
|
||||
"OPENAI_API_BASE",
|
||||
default="",
|
||||
)
|
||||
values["openai_proxy"] = get_from_dict_or_env(
|
||||
values,
|
||||
"openai_proxy",
|
||||
"OPENAI_PROXY",
|
||||
default="",
|
||||
)
|
||||
values["openai_organization"] = get_from_dict_or_env(
|
||||
values,
|
||||
"openai_organization",
|
||||
"OPENAI_ORGANIZATION",
|
||||
default="",
|
||||
)
|
||||
try:
|
||||
import openai
|
||||
|
||||
values["client"] = openai.Completion
|
||||
except ImportError:
|
||||
raise ImportError(
|
||||
"Could not import openai python package. "
|
||||
"Please install it with `pip install openai`."
|
||||
)
|
||||
if values["streaming"] and values["n"] > 1:
|
||||
raise ValueError("Cannot stream results when n > 1.")
|
||||
if values["streaming"] and values["best_of"] > 1:
|
||||
raise ValueError("Cannot stream results when best_of > 1.")
|
||||
return values
|
||||
|
||||
@property
|
||||
def _default_params(self) -> Dict[str, Any]:
|
||||
"""Get the default parameters for calling OpenAI API."""
|
||||
normal_params = {
|
||||
"temperature": self.temperature,
|
||||
"max_tokens": self.max_tokens,
|
||||
"top_p": self.top_p,
|
||||
"frequency_penalty": self.frequency_penalty,
|
||||
"presence_penalty": self.presence_penalty,
|
||||
"n": self.n,
|
||||
"request_timeout": self.request_timeout,
|
||||
"logit_bias": self.logit_bias,
|
||||
}
|
||||
|
||||
# Azure gpt-35-turbo doesn't support best_of
|
||||
# don't specify best_of if it is 1
|
||||
if self.best_of > 1:
|
||||
normal_params["best_of"] = self.best_of
|
||||
|
||||
return {**normal_params, **self.model_kwargs}
|
||||
|
||||
def _stream(
|
||||
self,
|
||||
prompt: str,
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[CallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> Iterator[GenerationChunk]:
|
||||
params = {**self._invocation_params, **kwargs, "stream": True}
|
||||
self.get_sub_prompts(params, [prompt], stop) # this mutates params
|
||||
for stream_resp in completion_with_retry(
|
||||
self, prompt=prompt, run_manager=run_manager, **params
|
||||
):
|
||||
chunk = _stream_response_to_generation_chunk(stream_resp)
|
||||
yield chunk
|
||||
if run_manager:
|
||||
run_manager.on_llm_new_token(
|
||||
chunk.text,
|
||||
chunk=chunk,
|
||||
verbose=self.verbose,
|
||||
logprobs=chunk.generation_info["logprobs"]
|
||||
if chunk.generation_info
|
||||
else None,
|
||||
)
|
||||
|
||||
async def _astream(
|
||||
self,
|
||||
prompt: str,
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> AsyncIterator[GenerationChunk]:
|
||||
params = {**self._invocation_params, **kwargs, "stream": True}
|
||||
self.get_sub_prompts(params, [prompt], stop) # this mutate params
|
||||
async for stream_resp in await acompletion_with_retry(
|
||||
self, prompt=prompt, run_manager=run_manager, **params
|
||||
):
|
||||
chunk = _stream_response_to_generation_chunk(stream_resp)
|
||||
yield chunk
|
||||
if run_manager:
|
||||
await run_manager.on_llm_new_token(
|
||||
chunk.text,
|
||||
chunk=chunk,
|
||||
verbose=self.verbose,
|
||||
logprobs=chunk.generation_info["logprobs"]
|
||||
if chunk.generation_info
|
||||
else None,
|
||||
)
|
||||
|
||||
def _generate(
|
||||
self,
|
||||
prompts: List[str],
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[CallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> LLMResult:
|
||||
"""Call out to OpenAI's endpoint with k unique prompts.
|
||||
|
||||
Args:
|
||||
prompts: The prompts to pass into the model.
|
||||
stop: Optional list of stop words to use when generating.
|
||||
|
||||
Returns:
|
||||
The full LLM output.
|
||||
|
||||
Example:
|
||||
.. code-block:: python
|
||||
|
||||
response = openai.generate(["Tell me a joke."])
|
||||
"""
|
||||
# TODO: write a unit test for this
|
||||
params = self._invocation_params
|
||||
params = {**params, **kwargs}
|
||||
sub_prompts = self.get_sub_prompts(params, prompts, stop)
|
||||
choices = []
|
||||
token_usage: Dict[str, int] = {}
|
||||
# Get the token usage from the response.
|
||||
# Includes prompt, completion, and total tokens used.
|
||||
_keys = {"completion_tokens", "prompt_tokens", "total_tokens"}
|
||||
for _prompts in sub_prompts:
|
||||
if self.streaming:
|
||||
if len(_prompts) > 1:
|
||||
raise ValueError("Cannot stream results with multiple prompts.")
|
||||
|
||||
generation: Optional[GenerationChunk] = None
|
||||
for chunk in self._stream(_prompts[0], stop, run_manager, **kwargs):
|
||||
if generation is None:
|
||||
generation = chunk
|
||||
else:
|
||||
generation += chunk
|
||||
assert generation is not None
|
||||
choices.append(
|
||||
{
|
||||
"text": generation.text,
|
||||
"finish_reason": generation.generation_info.get("finish_reason")
|
||||
if generation.generation_info
|
||||
else None,
|
||||
"logprobs": generation.generation_info.get("logprobs")
|
||||
if generation.generation_info
|
||||
else None,
|
||||
}
|
||||
)
|
||||
else:
|
||||
response = completion_with_retry(
|
||||
self, prompt=_prompts, run_manager=run_manager, **params
|
||||
)
|
||||
choices.extend(response["choices"])
|
||||
update_token_usage(_keys, response, token_usage)
|
||||
return self.create_llm_result(choices, prompts, token_usage)
|
||||
|
||||
async def _agenerate(
|
||||
self,
|
||||
prompts: List[str],
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> LLMResult:
|
||||
"""Call out to OpenAI's endpoint async with k unique prompts."""
|
||||
params = self._invocation_params
|
||||
params = {**params, **kwargs}
|
||||
sub_prompts = self.get_sub_prompts(params, prompts, stop)
|
||||
choices = []
|
||||
token_usage: Dict[str, int] = {}
|
||||
# Get the token usage from the response.
|
||||
# Includes prompt, completion, and total tokens used.
|
||||
_keys = {"completion_tokens", "prompt_tokens", "total_tokens"}
|
||||
for _prompts in sub_prompts:
|
||||
if self.streaming:
|
||||
if len(_prompts) > 1:
|
||||
raise ValueError("Cannot stream results with multiple prompts.")
|
||||
|
||||
generation: Optional[GenerationChunk] = None
|
||||
async for chunk in self._astream(
|
||||
_prompts[0], stop, run_manager, **kwargs
|
||||
):
|
||||
if generation is None:
|
||||
generation = chunk
|
||||
else:
|
||||
generation += chunk
|
||||
assert generation is not None
|
||||
choices.append(
|
||||
{
|
||||
"text": generation.text,
|
||||
"finish_reason": generation.generation_info.get("finish_reason")
|
||||
if generation.generation_info
|
||||
else None,
|
||||
"logprobs": generation.generation_info.get("logprobs")
|
||||
if generation.generation_info
|
||||
else None,
|
||||
}
|
||||
)
|
||||
else:
|
||||
response = await acompletion_with_retry(
|
||||
self, prompt=_prompts, run_manager=run_manager, **params
|
||||
)
|
||||
choices.extend(response["choices"])
|
||||
update_token_usage(_keys, response, token_usage)
|
||||
return self.create_llm_result(choices, prompts, token_usage)
|
||||
|
||||
def get_sub_prompts(
|
||||
self,
|
||||
params: Dict[str, Any],
|
||||
prompts: List[str],
|
||||
stop: Optional[List[str]] = None,
|
||||
) -> List[List[str]]:
|
||||
"""Get the sub prompts for llm call."""
|
||||
if stop is not None:
|
||||
if "stop" in params:
|
||||
raise ValueError("`stop` found in both the input and default params.")
|
||||
params["stop"] = stop
|
||||
if params["max_tokens"] == -1:
|
||||
if len(prompts) != 1:
|
||||
raise ValueError(
|
||||
"max_tokens set to -1 not supported for multiple inputs."
|
||||
)
|
||||
params["max_tokens"] = self.max_tokens_for_prompt(prompts[0])
|
||||
sub_prompts = [
|
||||
prompts[i : i + self.batch_size]
|
||||
for i in range(0, len(prompts), self.batch_size)
|
||||
]
|
||||
return sub_prompts
|
||||
|
||||
def create_llm_result(
|
||||
self, choices: Any, prompts: List[str], token_usage: Dict[str, int]
|
||||
) -> LLMResult:
|
||||
"""Create the LLMResult from the choices and prompts."""
|
||||
generations = []
|
||||
for i, _ in enumerate(prompts):
|
||||
sub_choices = choices[i * self.n : (i + 1) * self.n]
|
||||
generations.append(
|
||||
[
|
||||
Generation(
|
||||
text=choice["text"],
|
||||
generation_info=dict(
|
||||
finish_reason=choice.get("finish_reason"),
|
||||
logprobs=choice.get("logprobs"),
|
||||
),
|
||||
)
|
||||
for choice in sub_choices
|
||||
]
|
||||
)
|
||||
llm_output = {"token_usage": token_usage, "model_name": self.model_name}
|
||||
return LLMResult(generations=generations, llm_output=llm_output)
|
||||
|
||||
@property
|
||||
def _invocation_params(self) -> Dict[str, Any]:
|
||||
"""Get the parameters used to invoke the model."""
|
||||
openai_creds: Dict[str, Any] = {
|
||||
"api_key": self.openai_api_key,
|
||||
"api_base": self.openai_api_base,
|
||||
"organization": self.openai_organization,
|
||||
}
|
||||
if self.openai_proxy:
|
||||
import openai
|
||||
|
||||
openai.proxy = {"http": self.openai_proxy, "https": self.openai_proxy} # type: ignore[assignment] # noqa: E501
|
||||
return {**openai_creds, **self._default_params}
|
||||
|
||||
@property
|
||||
def _identifying_params(self) -> Mapping[str, Any]:
|
||||
"""Get the identifying parameters."""
|
||||
return {**{"model_name": self.model_name}, **self._default_params}
|
||||
|
||||
@property
|
||||
def _llm_type(self) -> str:
|
||||
"""Return type of llm."""
|
||||
return "openai"
|
||||
|
||||
def get_token_ids(self, text: str) -> List[int]:
|
||||
"""Get the token IDs using the tiktoken package."""
|
||||
# tiktoken NOT supported for Python < 3.8
|
||||
if sys.version_info[1] < 8:
|
||||
return super().get_num_tokens(text)
|
||||
try:
|
||||
import tiktoken
|
||||
except ImportError:
|
||||
raise ImportError(
|
||||
"Could not import tiktoken python package. "
|
||||
"This is needed in order to calculate get_num_tokens. "
|
||||
"Please install it with `pip install tiktoken`."
|
||||
)
|
||||
|
||||
model_name = self.tiktoken_model_name or self.model_name
|
||||
try:
|
||||
enc = tiktoken.encoding_for_model(model_name)
|
||||
except KeyError:
|
||||
logger.warning("Warning: model not found. Using cl100k_base encoding.")
|
||||
model = "cl100k_base"
|
||||
enc = tiktoken.get_encoding(model)
|
||||
|
||||
return enc.encode(
|
||||
text,
|
||||
allowed_special=self.allowed_special,
|
||||
disallowed_special=self.disallowed_special,
|
||||
)
|
||||
|
||||
@staticmethod
|
||||
def modelname_to_contextsize(modelname: str) -> int:
|
||||
"""Calculate the maximum number of tokens possible to generate for a model.
|
||||
|
||||
Args:
|
||||
modelname: The modelname we want to know the context size for.
|
||||
|
||||
Returns:
|
||||
The maximum context size
|
||||
|
||||
Example:
|
||||
.. code-block:: python
|
||||
|
||||
max_tokens = openai.modelname_to_contextsize("text-davinci-003")
|
||||
"""
|
||||
model_token_mapping = {
|
||||
"gpt-4": 8192,
|
||||
"gpt-4-0314": 8192,
|
||||
"gpt-4-0613": 8192,
|
||||
"gpt-4-32k": 32768,
|
||||
"gpt-4-32k-0314": 32768,
|
||||
"gpt-4-32k-0613": 32768,
|
||||
"gpt-3.5-turbo": 4096,
|
||||
"gpt-3.5-turbo-0301": 4096,
|
||||
"gpt-3.5-turbo-0613": 4096,
|
||||
"gpt-3.5-turbo-16k": 16385,
|
||||
"gpt-3.5-turbo-16k-0613": 16385,
|
||||
"gpt-3.5-turbo-instruct": 4096,
|
||||
"text-ada-001": 2049,
|
||||
"ada": 2049,
|
||||
"text-babbage-001": 2040,
|
||||
"babbage": 2049,
|
||||
"text-curie-001": 2049,
|
||||
"curie": 2049,
|
||||
"davinci": 2049,
|
||||
"text-davinci-003": 4097,
|
||||
"text-davinci-002": 4097,
|
||||
"code-davinci-002": 8001,
|
||||
"code-davinci-001": 8001,
|
||||
"code-cushman-002": 2048,
|
||||
"code-cushman-001": 2048,
|
||||
}
|
||||
|
||||
# handling finetuned models
|
||||
if "ft-" in modelname:
|
||||
modelname = modelname.split(":")[0]
|
||||
|
||||
context_size = model_token_mapping.get(modelname, None)
|
||||
|
||||
if context_size is None:
|
||||
raise ValueError(
|
||||
f"Unknown model: {modelname}. Please provide a valid OpenAI model name."
|
||||
"Known models are: " + ", ".join(model_token_mapping.keys())
|
||||
)
|
||||
|
||||
return context_size
|
||||
|
||||
@property
|
||||
def max_context_size(self) -> int:
|
||||
"""Get max context size for this model."""
|
||||
return self.modelname_to_contextsize(self.model_name)
|
||||
|
||||
def max_tokens_for_prompt(self, prompt: str) -> int:
|
||||
"""Calculate the maximum number of tokens possible to generate for a prompt.
|
||||
|
||||
Args:
|
||||
prompt: The prompt to pass into the model.
|
||||
|
||||
Returns:
|
||||
The maximum number of tokens to generate for a prompt.
|
||||
|
||||
Example:
|
||||
.. code-block:: python
|
||||
|
||||
max_tokens = openai.max_token_for_prompt("Tell me a joke.")
|
||||
"""
|
||||
num_tokens = self.get_num_tokens(prompt)
|
||||
return self.max_context_size - num_tokens
|
||||
|
||||
|
||||
class OpenAI(BaseOpenAI):
|
||||
"""OpenAI large language models.
|
||||
|
||||
To use, you should have the ``openai`` python package installed, and the
|
||||
environment variable ``OPENAI_API_KEY`` set with your API key.
|
||||
|
||||
Any parameters that are valid to be passed to the openai.create call can be passed
|
||||
in, even if not explicitly saved on this class.
|
||||
|
||||
Example:
|
||||
.. code-block:: python
|
||||
|
||||
from langchain.llms import OpenAI
|
||||
openai = OpenAI(model_name="text-davinci-003")
|
||||
"""
|
||||
|
||||
@property
|
||||
def _invocation_params(self) -> Dict[str, Any]:
|
||||
return {**{"model": self.model_name}, **super()._invocation_params}
|
||||
|
||||
|
||||
class AzureOpenAI(BaseOpenAI):
|
||||
"""Azure-specific OpenAI large language models.
|
||||
|
||||
To use, you should have the ``openai`` python package installed, and the
|
||||
environment variable ``OPENAI_API_KEY`` set with your API key.
|
||||
|
||||
Any parameters that are valid to be passed to the openai.create call can be passed
|
||||
in, even if not explicitly saved on this class.
|
||||
|
||||
Example:
|
||||
.. code-block:: python
|
||||
|
||||
from langchain.llms import AzureOpenAI
|
||||
openai = AzureOpenAI(model_name="text-davinci-003")
|
||||
"""
|
||||
|
||||
deployment_name: str = ""
|
||||
"""Deployment name to use."""
|
||||
openai_api_type: str = ""
|
||||
openai_api_version: str = ""
|
||||
|
||||
@root_validator()
|
||||
def validate_azure_settings(cls, values: Dict) -> Dict:
|
||||
values["openai_api_version"] = get_from_dict_or_env(
|
||||
values,
|
||||
"openai_api_version",
|
||||
"OPENAI_API_VERSION",
|
||||
)
|
||||
values["openai_api_type"] = get_from_dict_or_env(
|
||||
values, "openai_api_type", "OPENAI_API_TYPE", "azure"
|
||||
)
|
||||
return values
|
||||
|
||||
@property
|
||||
def _identifying_params(self) -> Mapping[str, Any]:
|
||||
return {
|
||||
**{"deployment_name": self.deployment_name},
|
||||
**super()._identifying_params,
|
||||
}
|
||||
|
||||
@property
|
||||
def _invocation_params(self) -> Dict[str, Any]:
|
||||
openai_params = {
|
||||
"engine": self.deployment_name,
|
||||
"api_type": self.openai_api_type,
|
||||
"api_version": self.openai_api_version,
|
||||
}
|
||||
return {**openai_params, **super()._invocation_params}
|
||||
|
||||
@property
|
||||
def _llm_type(self) -> str:
|
||||
"""Return type of llm."""
|
||||
return "azure"
|
||||
|
||||
@property
|
||||
def lc_attributes(self) -> Dict[str, Any]:
|
||||
return {
|
||||
"openai_api_type": self.openai_api_type,
|
||||
"openai_api_version": self.openai_api_version,
|
||||
}
|
||||
|
||||
|
||||
class OpenAIChat(BaseLLM):
|
||||
"""OpenAI Chat large language models.
|
||||
|
||||
To use, you should have the ``openai`` python package installed, and the
|
||||
environment variable ``OPENAI_API_KEY`` set with your API key.
|
||||
|
||||
Any parameters that are valid to be passed to the openai.create call can be passed
|
||||
in, even if not explicitly saved on this class.
|
||||
|
||||
Example:
|
||||
.. code-block:: python
|
||||
|
||||
from langchain.llms import OpenAIChat
|
||||
openaichat = OpenAIChat(model_name="gpt-3.5-turbo")
|
||||
"""
|
||||
|
||||
client: Any #: :meta private:
|
||||
model_name: str = "gpt-3.5-turbo"
|
||||
"""Model name to use."""
|
||||
model_kwargs: Dict[str, Any] = Field(default_factory=dict)
|
||||
"""Holds any model parameters valid for `create` call not explicitly specified."""
|
||||
openai_api_key: Optional[str] = None
|
||||
openai_api_base: Optional[str] = None
|
||||
# to support explicit proxy for OpenAI
|
||||
openai_proxy: Optional[str] = None
|
||||
max_retries: int = 6
|
||||
"""Maximum number of retries to make when generating."""
|
||||
prefix_messages: List = Field(default_factory=list)
|
||||
"""Series of messages for Chat input."""
|
||||
streaming: bool = False
|
||||
"""Whether to stream the results or not."""
|
||||
allowed_special: Union[Literal["all"], AbstractSet[str]] = set()
|
||||
"""Set of special tokens that are allowed。"""
|
||||
disallowed_special: Union[Literal["all"], Collection[str]] = "all"
|
||||
"""Set of special tokens that are not allowed。"""
|
||||
|
||||
@root_validator(pre=True)
|
||||
def build_extra(cls, values: Dict[str, Any]) -> Dict[str, Any]:
|
||||
"""Build extra kwargs from additional params that were passed in."""
|
||||
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.")
|
||||
extra[field_name] = values.pop(field_name)
|
||||
values["model_kwargs"] = extra
|
||||
return values
|
||||
|
||||
@root_validator()
|
||||
def validate_environment(cls, values: Dict) -> Dict:
|
||||
"""Validate that api key and python package exists in environment."""
|
||||
openai_api_key = get_from_dict_or_env(
|
||||
values, "openai_api_key", "OPENAI_API_KEY"
|
||||
)
|
||||
openai_api_base = get_from_dict_or_env(
|
||||
values,
|
||||
"openai_api_base",
|
||||
"OPENAI_API_BASE",
|
||||
default="",
|
||||
)
|
||||
openai_proxy = get_from_dict_or_env(
|
||||
values,
|
||||
"openai_proxy",
|
||||
"OPENAI_PROXY",
|
||||
default="",
|
||||
)
|
||||
openai_organization = get_from_dict_or_env(
|
||||
values, "openai_organization", "OPENAI_ORGANIZATION", default=""
|
||||
)
|
||||
try:
|
||||
import openai
|
||||
|
||||
openai.api_key = openai_api_key
|
||||
if openai_api_base:
|
||||
openai.api_base = openai_api_base
|
||||
if openai_organization:
|
||||
openai.organization = openai_organization
|
||||
if openai_proxy:
|
||||
openai.proxy = {"http": openai_proxy, "https": openai_proxy} # type: ignore[assignment] # noqa: E501
|
||||
except ImportError:
|
||||
raise ImportError(
|
||||
"Could not import openai python package. "
|
||||
"Please install it with `pip install openai`."
|
||||
)
|
||||
try:
|
||||
values["client"] = openai.ChatCompletion
|
||||
except AttributeError:
|
||||
raise ValueError(
|
||||
"`openai` has no `ChatCompletion` attribute, this is likely "
|
||||
"due to an old version of the openai package. Try upgrading it "
|
||||
"with `pip install --upgrade openai`."
|
||||
)
|
||||
return values
|
||||
|
||||
@property
|
||||
def _default_params(self) -> Dict[str, Any]:
|
||||
"""Get the default parameters for calling OpenAI API."""
|
||||
return self.model_kwargs
|
||||
|
||||
def _get_chat_params(
|
||||
self, prompts: List[str], stop: Optional[List[str]] = None
|
||||
) -> Tuple:
|
||||
if len(prompts) > 1:
|
||||
raise ValueError(
|
||||
f"OpenAIChat currently only supports single prompt, got {prompts}"
|
||||
)
|
||||
messages = self.prefix_messages + [{"role": "user", "content": prompts[0]}]
|
||||
params: Dict[str, Any] = {**{"model": self.model_name}, **self._default_params}
|
||||
if stop is not None:
|
||||
if "stop" in params:
|
||||
raise ValueError("`stop` found in both the input and default params.")
|
||||
params["stop"] = stop
|
||||
if params.get("max_tokens") == -1:
|
||||
# for ChatGPT api, omitting max_tokens is equivalent to having no limit
|
||||
del params["max_tokens"]
|
||||
return messages, params
|
||||
|
||||
def _stream(
|
||||
self,
|
||||
prompt: str,
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[CallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> Iterator[GenerationChunk]:
|
||||
messages, params = self._get_chat_params([prompt], stop)
|
||||
params = {**params, **kwargs, "stream": True}
|
||||
for stream_resp in completion_with_retry(
|
||||
self, messages=messages, run_manager=run_manager, **params
|
||||
):
|
||||
token = stream_resp["choices"][0]["delta"].get("content", "")
|
||||
chunk = GenerationChunk(text=token)
|
||||
yield chunk
|
||||
if run_manager:
|
||||
run_manager.on_llm_new_token(token, chunk=chunk)
|
||||
|
||||
async def _astream(
|
||||
self,
|
||||
prompt: str,
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> AsyncIterator[GenerationChunk]:
|
||||
messages, params = self._get_chat_params([prompt], stop)
|
||||
params = {**params, **kwargs, "stream": True}
|
||||
async for stream_resp in await acompletion_with_retry(
|
||||
self, messages=messages, run_manager=run_manager, **params
|
||||
):
|
||||
token = stream_resp["choices"][0]["delta"].get("content", "")
|
||||
chunk = GenerationChunk(text=token)
|
||||
yield chunk
|
||||
if run_manager:
|
||||
await run_manager.on_llm_new_token(token, chunk=chunk)
|
||||
|
||||
def _generate(
|
||||
self,
|
||||
prompts: List[str],
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[CallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> LLMResult:
|
||||
if self.streaming:
|
||||
generation: Optional[GenerationChunk] = None
|
||||
for chunk in self._stream(prompts[0], stop, run_manager, **kwargs):
|
||||
if generation is None:
|
||||
generation = chunk
|
||||
else:
|
||||
generation += chunk
|
||||
assert generation is not None
|
||||
return LLMResult(generations=[[generation]])
|
||||
|
||||
messages, params = self._get_chat_params(prompts, stop)
|
||||
params = {**params, **kwargs}
|
||||
full_response = completion_with_retry(
|
||||
self, messages=messages, run_manager=run_manager, **params
|
||||
)
|
||||
llm_output = {
|
||||
"token_usage": full_response["usage"],
|
||||
"model_name": self.model_name,
|
||||
}
|
||||
return LLMResult(
|
||||
generations=[
|
||||
[Generation(text=full_response["choices"][0]["message"]["content"])]
|
||||
],
|
||||
llm_output=llm_output,
|
||||
)
|
||||
|
||||
async def _agenerate(
|
||||
self,
|
||||
prompts: List[str],
|
||||
stop: Optional[List[str]] = None,
|
||||
run_manager: Optional[AsyncCallbackManagerForLLMRun] = None,
|
||||
**kwargs: Any,
|
||||
) -> LLMResult:
|
||||
if self.streaming:
|
||||
generation: Optional[GenerationChunk] = None
|
||||
async for chunk in self._astream(prompts[0], stop, run_manager, **kwargs):
|
||||
if generation is None:
|
||||
generation = chunk
|
||||
else:
|
||||
generation += chunk
|
||||
assert generation is not None
|
||||
return LLMResult(generations=[[generation]])
|
||||
|
||||
messages, params = self._get_chat_params(prompts, stop)
|
||||
params = {**params, **kwargs}
|
||||
full_response = await acompletion_with_retry(
|
||||
self, messages=messages, run_manager=run_manager, **params
|
||||
)
|
||||
llm_output = {
|
||||
"token_usage": full_response["usage"],
|
||||
"model_name": self.model_name,
|
||||
}
|
||||
return LLMResult(
|
||||
generations=[
|
||||
[Generation(text=full_response["choices"][0]["message"]["content"])]
|
||||
],
|
||||
llm_output=llm_output,
|
||||
)
|
||||
|
||||
@property
|
||||
def _identifying_params(self) -> Mapping[str, Any]:
|
||||
"""Get the identifying parameters."""
|
||||
return {**{"model_name": self.model_name}, **self._default_params}
|
||||
|
||||
@property
|
||||
def _llm_type(self) -> str:
|
||||
"""Return type of llm."""
|
||||
return "openai-chat"
|
||||
|
||||
def get_token_ids(self, text: str) -> List[int]:
|
||||
"""Get the token IDs using the tiktoken package."""
|
||||
# tiktoken NOT supported for Python < 3.8
|
||||
if sys.version_info[1] < 8:
|
||||
return super().get_token_ids(text)
|
||||
try:
|
||||
import tiktoken
|
||||
except ImportError:
|
||||
raise ImportError(
|
||||
"Could not import tiktoken python package. "
|
||||
"This is needed in order to calculate get_num_tokens. "
|
||||
"Please install it with `pip install tiktoken`."
|
||||
)
|
||||
|
||||
enc = tiktoken.encoding_for_model(self.model_name)
|
||||
return enc.encode(
|
||||
text,
|
||||
allowed_special=self.allowed_special,
|
||||
disallowed_special=self.disallowed_special,
|
||||
)
|
||||
@ -1,73 +0,0 @@
|
||||
# props to shroominic
|
||||
from swarms.tools.base import Tool, ToolException
|
||||
from typing import Any, List
|
||||
from codeinterpreterapi import CodeInterpreterSession, File, ToolException
|
||||
|
||||
|
||||
class CodeInterpreter(Tool):
|
||||
def __init__(self, name: str, description: str):
|
||||
super().__init__(name, description, self.run)
|
||||
|
||||
def run(self, user_request: str, file_paths: List[str] = []) -> Any:
|
||||
# create a session
|
||||
session = CodeInterpreterSession()
|
||||
session.start()
|
||||
|
||||
# create files from paths
|
||||
files = [File.from_path(file_path) for file_path in file_paths]
|
||||
|
||||
try:
|
||||
# generate a response based on user input
|
||||
response = session.generate_response(user_request, files=files)
|
||||
|
||||
# output the response (text + image)
|
||||
print("AI: ", response.content)
|
||||
for file in response.files:
|
||||
file.show_image()
|
||||
except Exception as e:
|
||||
raise ToolException(f"Error running CodeInterpreter: {e}")
|
||||
finally:
|
||||
# terminate the session
|
||||
session.stop()
|
||||
|
||||
async def arun(self, user_request: str, file_paths: List[str] = []) -> Any:
|
||||
# create a session
|
||||
session = CodeInterpreterSession()
|
||||
await session.astart()
|
||||
|
||||
# create files from paths
|
||||
files = [File.from_path(file_path) for file_path in file_paths]
|
||||
|
||||
try:
|
||||
# generate a response based on user input
|
||||
response = await session.generate_response(user_request, files=files)
|
||||
|
||||
# output the response (text + image)
|
||||
print("AI: ", response.content)
|
||||
for file in response.files:
|
||||
file.show_image()
|
||||
except Exception as e:
|
||||
raise ToolException(f"Error running CodeInterpreter: {e}")
|
||||
finally:
|
||||
# terminate the session
|
||||
await session.astop()
|
||||
|
||||
|
||||
"""
|
||||
|
||||
tool = CodeInterpreter("Code Interpreter", "A tool to interpret code and generate useful outputs.")
|
||||
tool.run("Plot the bitcoin chart of 2023 YTD")
|
||||
|
||||
# Or with file inputs
|
||||
tool.run("Analyze this dataset and plot something interesting about it.", ["examples/assets/iris.csv"])
|
||||
|
||||
|
||||
|
||||
import asyncio
|
||||
|
||||
tool = CodeInterpreter("Code Interpreter", "A tool to interpret code and generate useful outputs.")
|
||||
asyncio.run(tool.arun("Plot the bitcoin chart of 2023 YTD"))
|
||||
|
||||
# Or with file inputs
|
||||
asyncio.run(tool.arun("Analyze this dataset and plot something interesting about it.", ["examples/assets/iris.csv"]))
|
||||
"""
|
||||
Loading…
Reference in new issue