121 lines
4.2 KiB
Python
121 lines
4.2 KiB
Python
|
# from transformers import GPT2TokenizerFast
|
||
|
|
import tiktoken
|
||
|
|
from openai import AzureOpenAI
|
||
|
|
import openai
|
||
|
|
import os
|
||
|
|
from time import sleep
|
||
|
|
import dotenv
|
||
|
|
# loads .env file with your OPENAI_API_KEY
|
||
|
|
dotenv.load_dotenv()
|
||
|
|
|
||
|
|
|
||
|
|
def set_environment_variables(engine=os.getenv("Engine_0613_16k")):
|
||
|
|
if engine.startswith('gpt4') or engine.startswith('gpt-4'):
|
||
|
|
openai.api_base = os.getenv("OPENAI_API_BASE_DC")
|
||
|
|
openai.api_key = os.getenv("OPENAI_API_KEY_GPT4")
|
||
|
|
elif engine.startswith('modc-stg-gpt4'):
|
||
|
|
openai.api_base = os.getenv("OPENAI_API_BASE_GPT4_MODC")
|
||
|
|
openai.api_key = os.getenv("OPENAI_API_KEY_GPT4_MODC")
|
||
|
|
elif engine.upper() == 'ENGINE_GPT4_TURBO':
|
||
|
|
openai.api_base = os.getenv("OPENAI_API_BASE_GPT4_TURBO")
|
||
|
|
openai.api_key = os.getenv("OPENAI_API_KEY_GPT4_TURBO")
|
||
|
|
elif engine.startswith('modc-stg-gpt35turbo16k'):
|
||
|
|
openai.api_base = os.getenv("OPENAI_API_BASE_GPT3_MODC")
|
||
|
|
openai.api_key = os.getenv("OPENAI_API_KEY_GPT3_MODC")
|
||
|
|
else:
|
||
|
|
openai.api_base = os.getenv("OPENAI_API_BASE")
|
||
|
|
openai.api_key = os.getenv("OPENAI_API_KEY")
|
||
|
|
openai.Engine = engine
|
||
|
|
openai.api_type = os.getenv("OPENAI_API_TYPE")
|
||
|
|
openai.api_version = os.getenv("OPENAI_API_VERSION")
|
||
|
|
|
||
|
|
# tokenizer = GPT2TokenizerFast.from_pretrained("gpt2")
|
||
|
|
tokenizer = tiktoken.get_encoding("cl100k_base")
|
||
|
|
|
||
|
|
|
||
|
|
def get_embedding(text, engine=os.getenv("EMBEDDING_ENGINE")):
|
||
|
|
count = 0
|
||
|
|
error = ''
|
||
|
|
while count < 5:
|
||
|
|
try:
|
||
|
|
if count > 0:
|
||
|
|
print(f'retrying the {count} time for getting text embedding...')
|
||
|
|
return openai.Embedding.create(input=text, engine=engine)['data'][0]['embedding']
|
||
|
|
except Exception as e:
|
||
|
|
error = str(e)
|
||
|
|
print(error)
|
||
|
|
count += 1
|
||
|
|
sleep(1)
|
||
|
|
|
||
|
|
|
||
|
|
def num_tokens_from_string(string: str) -> int:
|
||
|
|
"""Returns the number of tokens in a text string."""
|
||
|
|
num_tokens = len(tokenizer.encode(string))
|
||
|
|
return num_tokens
|
||
|
|
|
||
|
|
|
||
|
|
def num_tokens_from_messages(messages, model="gpt-35-turbo-16k"):
|
||
|
|
"""Returns the number of tokens used by a list of messages."""
|
||
|
|
encoding = tiktoken.get_encoding("cl100k_base")
|
||
|
|
if model == "gpt-35-turbo-16k":
|
||
|
|
tokens_per_message = 4 # every message follows <|start|>{role/name}\n{content}<|end|>\n
|
||
|
|
tokens_per_name = -1 # if there's a name, the role is omitted
|
||
|
|
elif model == "gpt-4-32k":
|
||
|
|
tokens_per_message = 3
|
||
|
|
tokens_per_name = 1
|
||
|
|
else:
|
||
|
|
tokens_per_message = 3
|
||
|
|
tokens_per_name = 1
|
||
|
|
num_tokens = 0
|
||
|
|
for message in messages:
|
||
|
|
num_tokens += tokens_per_message
|
||
|
|
for key, value in message.items():
|
||
|
|
num_tokens += len(encoding.encode(value))
|
||
|
|
if key == "name":
|
||
|
|
num_tokens += tokens_per_name
|
||
|
|
num_tokens += 3 # every reply is primed with <|start|>assistant<|message|>
|
||
|
|
return num_tokens
|
||
|
|
|
||
|
|
|
||
|
|
def chat(prompt: str,
|
||
|
|
engine = os.getenv("Engine_GPT4o"),
|
||
|
|
azure_endpoint=os.getenv("OPENAI_API_BASE_GPT4o"),
|
||
|
|
api_key=os.getenv("OPENAI_API_KEY_GPT4o"),
|
||
|
|
api_version=os.getenv("OPENAI_API_VERSION_GPT4o"),
|
||
|
|
temperature: float = 0.0):
|
||
|
|
client = AzureOpenAI(
|
||
|
|
azure_endpoint=azure_endpoint,
|
||
|
|
api_key=api_key,
|
||
|
|
api_version=api_version
|
||
|
|
)
|
||
|
|
|
||
|
|
count = 0
|
||
|
|
error = ''
|
||
|
|
max_tokens = 4000
|
||
|
|
request_timeout = 120
|
||
|
|
while count < 8:
|
||
|
|
try:
|
||
|
|
if count > 0:
|
||
|
|
print(f'retrying the {count} time...')
|
||
|
|
response = client.chat.completions.create(
|
||
|
|
model=engine,
|
||
|
|
temperature=temperature,
|
||
|
|
max_tokens=max_tokens,
|
||
|
|
top_p=0.95,
|
||
|
|
frequency_penalty=0,
|
||
|
|
presence_penalty=0,
|
||
|
|
timeout=request_timeout,
|
||
|
|
stop=None,
|
||
|
|
messages=[
|
||
|
|
{"role": "user", "content": prompt}
|
||
|
|
]
|
||
|
|
)
|
||
|
|
return response.choices[0].message.content, False
|
||
|
|
except Exception as e:
|
||
|
|
error = str(e)
|
||
|
|
print(f"error message: {error}")
|
||
|
|
if 'maximum context length' in error:
|
||
|
|
return error, True
|
||
|
|
count += 1
|
||
|
|
sleep(3)
|
||
|
|
return error, True
|