ChatDev/camel/model_backend.py

199 lines
7.0 KiB
Python

# =========== Copyright 2023 @ 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 @ CAMEL-AI.org. All Rights Reserved. ===========
from abc import ABC, abstractmethod
from typing import Any, Dict
import openai
import tiktoken
from camel.typing import ModelType
from chatdev.statistics import prompt_cost
from chatdev.utils import log_visualize
try:
from openai.types.chat import ChatCompletion
openai_new_api = True # new openai api version
except ImportError:
openai_new_api = False # old openai api version
import os
OPENAI_API_KEY = os.environ['OPENAI_API_KEY']
if 'BASE_URL' in os.environ:
BASE_URL = os.environ['BASE_URL']
else:
BASE_URL = None
class ModelBackend(ABC):
r"""Base class for different model backends.
May be OpenAI API, a local LLM, a stub for unit tests, etc."""
@abstractmethod
def run(self, *args, **kwargs):
r"""Runs the query to the backend model.
Raises:
RuntimeError: if the return value from OpenAI API
is not a dict that is expected.
Returns:
Dict[str, Any]: All backends must return a dict in OpenAI format.
"""
pass
class OpenAIModel(ModelBackend):
r"""OpenAI API in a unified ModelBackend interface."""
def __init__(self, model_type: ModelType, model_config_dict: Dict) -> None:
super().__init__()
self.model_type = model_type
self.model_config_dict = model_config_dict
def run(self, *args, **kwargs):
string = "\n".join([message["content"] for message in kwargs["messages"]])
encoding = tiktoken.encoding_for_model(self.model_type.value)
num_prompt_tokens = len(encoding.encode(string))
gap_between_send_receive = 15 * len(kwargs["messages"])
num_prompt_tokens += gap_between_send_receive
if openai_new_api:
# Experimental, add base_url
if BASE_URL:
client = openai.OpenAI(
api_key=OPENAI_API_KEY,
base_url=BASE_URL,
)
else:
client = openai.OpenAI(
api_key=OPENAI_API_KEY
)
num_max_token_map = {
"gpt-3.5-turbo": 4096,
"gpt-3.5-turbo-16k": 16384,
"gpt-3.5-turbo-0613": 4096,
"gpt-3.5-turbo-16k-0613": 16384,
"gpt-4": 8192,
"gpt-4-0613": 8192,
"gpt-4-32k": 32768,
"gpt-4-1106-preview": 4096,
"gpt-4-1106-vision-preview": 4096,
}
num_max_token = num_max_token_map[self.model_type.value]
num_max_completion_tokens = num_max_token - num_prompt_tokens
self.model_config_dict['max_tokens'] = num_max_completion_tokens
response = client.chat.completions.create(*args, **kwargs, model=self.model_type.value,
**self.model_config_dict)
cost = prompt_cost(
self.model_type.value,
num_prompt_tokens=response.usage.prompt_tokens,
num_completion_tokens=response.usage.completion_tokens
)
log_visualize(
"**[OpenAI_Usage_Info Receive]**\nprompt_tokens: {}\ncompletion_tokens: {}\ntotal_tokens: {}\ncost: ${:.6f}\n".format(
response.usage.prompt_tokens, response.usage.completion_tokens,
response.usage.total_tokens, cost))
if not isinstance(response, ChatCompletion):
raise RuntimeError("Unexpected return from OpenAI API")
return response
else:
num_max_token_map = {
"gpt-3.5-turbo": 4096,
"gpt-3.5-turbo-16k": 16384,
"gpt-3.5-turbo-0613": 4096,
"gpt-3.5-turbo-16k-0613": 16384,
"gpt-4": 8192,
"gpt-4-0613": 8192,
"gpt-4-32k": 32768,
}
num_max_token = num_max_token_map[self.model_type.value]
num_max_completion_tokens = num_max_token - num_prompt_tokens
self.model_config_dict['max_tokens'] = num_max_completion_tokens
response = openai.ChatCompletion.create(*args, **kwargs, model=self.model_type.value,
**self.model_config_dict)
cost = prompt_cost(
self.model_type.value,
num_prompt_tokens=response["usage"]["prompt_tokens"],
num_completion_tokens=response["usage"]["completion_tokens"]
)
log_visualize(
"**[OpenAI_Usage_Info Receive]**\nprompt_tokens: {}\ncompletion_tokens: {}\ntotal_tokens: {}\ncost: ${:.6f}\n".format(
response["usage"]["prompt_tokens"], response["usage"]["completion_tokens"],
response["usage"]["total_tokens"], cost))
if not isinstance(response, Dict):
raise RuntimeError("Unexpected return from OpenAI API")
return response
class StubModel(ModelBackend):
r"""A dummy model used for unit tests."""
def __init__(self, *args, **kwargs) -> None:
super().__init__()
def run(self, *args, **kwargs) -> Dict[str, Any]:
ARBITRARY_STRING = "Lorem Ipsum"
return dict(
id="stub_model_id",
usage=dict(),
choices=[
dict(finish_reason="stop",
message=dict(content=ARBITRARY_STRING, role="assistant"))
],
)
class ModelFactory:
r"""Factory of backend models.
Raises:
ValueError: in case the provided model type is unknown.
"""
@staticmethod
def create(model_type: ModelType, model_config_dict: Dict) -> ModelBackend:
default_model_type = ModelType.GPT_3_5_TURBO
if model_type in {
ModelType.GPT_3_5_TURBO,
ModelType.GPT_3_5_TURBO_NEW,
ModelType.GPT_4,
ModelType.GPT_4_32k,
ModelType.GPT_4_TURBO,
ModelType.GPT_4_TURBO_V,
None
}:
model_class = OpenAIModel
elif model_type == ModelType.STUB:
model_class = StubModel
else:
raise ValueError("Unknown model")
if model_type is None:
model_type = default_model_type
# log_visualize("Model Type: {}".format(model_type))
inst = model_class(model_type, model_config_dict)
return inst