17.10.24:编辑标题以便于搜索
原标题:OpenAI API 请求负载的哪一部分受到最大代币数量的限制?
我有点了解如何计算字符中的标记,但我实际上必须计算什么? 如果我有这样的有效负载:
{
"model": "gpt-3.5-turbo",
"temperature": 1,
"max_tokens": 400,
"presence_penalty": 0.85,
"frequency_penalty": 0.85,
"messages": [
{
"role": "system",
"content": "prompt"
},
{
"role": "assistant",
"content": "message"
},
// tens of messages
]
}
我必须从其中完全数出代币吗?还是我必须只计算"messages"
?如果是这样,我是否还必须计算所有 json 语法字符,例如空格键、括号和逗号?
"role"
和
"content"
键怎么样?
"role"
值怎么样?或者我必须简单地将所有
"content"
值连接到一个字符串中并仅根据它来计算标记?
"messages"
中提供的列表中的所有代币都被计算在内。这包括键“role”和“content”及其值,但不包括空格、括号、逗号和引号。我使用 OpenAI 提供的以下脚本来计算输入中的令牌数量。我修改了脚本来计算多条消息的输入(而不是输出响应)所涉及的成本,这对我来说相当准确。
import json
import os
import tiktoken
import numpy as np
from collections import defaultdict
def num_tokens_from_messages(messages, model="gpt-3.5-turbo-0613"):
"""Return the number of tokens used by a list of messages."""
try:
encoding = tiktoken.encoding_for_model(model)
except KeyError:
print("Warning: model not found. Using cl100k_base encoding.")
encoding = tiktoken.get_encoding("cl100k_base")
if model in {
"gpt-3.5-turbo-0613",
"gpt-3.5-turbo-16k-0613",
"gpt-4-0613",
"gpt-4-32k-0613",
}:
tokens_per_message = 3
tokens_per_name = 1
elif model == "gpt-3.5-turbo-0301":
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 "gpt-3.5-turbo" in model:
print("Warning: gpt-3.5-turbo may update over time. Returning num tokens assuming gpt-3.5-turbo-0613.")
return num_tokens_from_messages(messages, model="gpt-3.5-turbo-0613")
elif "gpt-4" in model:
print("Warning: gpt-4 may update over time. Returning num tokens assuming gpt-4-0613.")
return num_tokens_from_messages(messages, model="gpt-4-0613")
else:
raise NotImplementedError(
f"""num_tokens_from_messages() is not implemented for model {model}. See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens."""
)
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
convo_lens = []
for ex in dataset: #Your list of inputs
messages = ex["messages"]
convo_lens.append(num_tokens_from_messages(messages))
n_input_tokens_in_dataset = sum(min(4096, length) for length in convo_lens)
print(f"Input portion of the data has ~{n_input_tokens_in_dataset} tokens")
# costs as of Aug 29 2023.
costs = {
"gpt-4-0613": {
"input" : 0.03,
"output": 0.06
},
"gpt-4-32k-0613": {
"input" : 0.06,
"output": 0.12
},
"gpt-3.5-turbo-0613": {
"input": 0.0015,
"output": 0.002
},
"gpt-3.5-turbo-16k-0613": {
"input": 0.003,
"output": 0.004
}
}
# We select GPT 3.5 turbo here
print(f"Cost of inference: ${(n_input_tokens_in_dataset/1000) * costs['gpt-3.5-turbo-0613']['input']}")