获取词元数和计费字符数

生成式 AI 模型会将文本数据细分为单元(称为“词元”)以进行处理。文本数据转换为词元的方式取决于所使用的词元化器。词元可以是字符、字词或短语。每个模型具有在提示和响应中可以处理的词元数上限。本页面介绍了如何获取提示的词元数和计费字符数。

支持的模型

以下基础模型支持获取提示词元数:

  • text-bison
  • chat-bison
  • code-bison
  • codechat-bison
  • code-gecko
  • textembedding-gecko

如需了解如何获取 Gemini 模型的提示词元数,请参阅 Vertex AI Gemini API 获取词元数说明

获取提示的词元数

您可以使用 countTokens API 获取提示的词元数和计费字符数。countTokens 的输入格式取决于您使用的模型。每种输入格式都与 predict 输入格式相同。

text-bison

REST

如需使用 Vertex AI API 获取提示的词元数和计费字符数,请向发布者模型端点发送 POST 请求。

在使用任何请求数据之前,请先进行以下替换:

  • LOCATION:输入支持的区域。如需查看支持的区域的完整列表,请参阅可用位置
  • PROJECT_ID:您的项目 ID
  • PROMPT:要获取其词元数和计费字符数的提示。(请勿为此处的提示添加引号。)

HTTP 方法和网址:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/text-bison:countTokens

请求 JSON 正文:

{
  "instances": [
    { "prompt": "PROMPT" }
  ]
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

cat > request.json << 'EOF'
{
  "instances": [
    { "prompt": "PROMPT" }
  ]
}
EOF

然后,执行以下命令以发送 REST 请求:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/text-bison:countTokens"

PowerShell

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

@'
{
  "instances": [
    { "prompt": "PROMPT" }
  ]
}
'@  | Out-File -FilePath request.json -Encoding utf8

然后,执行以下命令以发送 REST 请求:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/text-bison:countTokens" | Select-Object -Expand Content

您应该收到类似以下内容的 JSON 响应。

{
  "totalTokens": 12,
  "totalBillableCharacters": 54
}

示例 curl 命令

PROJECT_ID="PROJECT_ID"

curl \
-X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/us-central1/publishers/google/models/text-bison:countTokens -d \
$'{
  "instances": [
    { "prompt": "Give me ten interview questions for the role of program manager." },
    { "prompt": "List some good qualities for a program manager." }
  ]
}'

chat-bison

REST

如需使用 Vertex AI API 获取提示的词元数和计费字符数,请向发布者模型端点发送 POST 请求。

在使用任何请求数据之前,请先进行以下替换:

  • LOCATION:输入支持的区域。如需查看支持的区域的完整列表,请参阅可用位置
  • PROJECT_ID:您的项目 ID
  • CONTEXT:可选。 上下文可以是您为模型提供的响应方式的说明,也可以是模型使用或引用来生成响应的信息。当您需要向模型提供信息时,请在提示中添加上下文信息,或者将响应的边界限制为仅上下文中的内容。
  • EXAMPLE_AUTHOR_1EXAMPLE_INPUT 的作者(用户)。
  • EXAMPLE_INPUT:消息的样本。
  • EXAMPLE_AUTHOR_2EXAMPLE_OUTPUT 的作者(聊天机器人)。
  • EXAMPLE_OUTPUT:理想响应的样本。
  • AUTHOR_1:消息的作者(用户)。
  • CONTENT:消息的内容。

HTTP 方法和网址:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/chat-bison:countTokens

请求 JSON 正文:

{
  "instances": [
    {
      "context": "CONTEXT",
      "examples": [
        {
          "input": {
            "author": "EXAMPLE_AUTHOR_1",
            "content": "EXAMPLE_INPUT"
          },
          "output": {
            "author": "EXAMPLE_AUTHOR_2",
            "content": "EXAMPLE_OUTPUT"
          }
        }
      ],
      "messages": [
        {
          "author": "AUTHOR_1",
          "content": "CONTENT"
        }
      ]
    }
  ]
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

cat > request.json << 'EOF'
{
  "instances": [
    {
      "context": "CONTEXT",
      "examples": [
        {
          "input": {
            "author": "EXAMPLE_AUTHOR_1",
            "content": "EXAMPLE_INPUT"
          },
          "output": {
            "author": "EXAMPLE_AUTHOR_2",
            "content": "EXAMPLE_OUTPUT"
          }
        }
      ],
      "messages": [
        {
          "author": "AUTHOR_1",
          "content": "CONTENT"
        }
      ]
    }
  ]
}
EOF

然后,执行以下命令以发送 REST 请求:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/chat-bison:countTokens"

PowerShell

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

@'
{
  "instances": [
    {
      "context": "CONTEXT",
      "examples": [
        {
          "input": {
            "author": "EXAMPLE_AUTHOR_1",
            "content": "EXAMPLE_INPUT"
          },
          "output": {
            "author": "EXAMPLE_AUTHOR_2",
            "content": "EXAMPLE_OUTPUT"
          }
        }
      ],
      "messages": [
        {
          "author": "AUTHOR_1",
          "content": "CONTENT"
        }
      ]
    }
  ]
}
'@  | Out-File -FilePath request.json -Encoding utf8

然后,执行以下命令以发送 REST 请求:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/chat-bison:countTokens" | Select-Object -Expand Content

您应该收到类似以下内容的 JSON 响应。

{
  "totalTokens": 43,
  "totalBillableCharacters": 182
}

示例 curl 命令

PROJECT_ID="PROJECT_ID"

curl \
-X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/us-central1/publishers/google/models/chat-bison:countTokens -d \
$'{
  "instances": [
    {
      "context": "You are Captain Bartholomew, the most feared pirate dog of the seven seas.",
      "examples": [
        {
          "input": {
            "author": "User",
            "content": "Hello!"
          },
          "output": {
            "author": "Captain Barktholomew",
            "content": "Argh! What brings ye to my ship?"
          }
        },
        {
          "input": {
            "author": "User",
            "content": "Who are you?"
          },
          "output": {
            "author": "Captain Barktholomew",
            "content": "I be Captain Barktholomew, the most feared pirate dog of the seven seas."
          }
        }
      ],
      "messages": [
        {
          "author": "User",
          "content": "Hello!"
        },
        {
          "author": "Captain Barktholomew",
          "content": "Ahoy there, landlubber! What brings ye to me ship?"
        },
        {
          "author": "User",
          "content": "Can you tell me a tale of your most recent adventure?"
        },
        {
          "author": "Captain Barktholomew",
          "content": "Aye, I\'ll spin ye a tale of me latest adventure....",
        },
        {
          "author": "User",
          "content": "I\'m listening."
        }
      ]
    }
  ]
}'

code-bison

REST

如需使用 Vertex AI API 获取提示的词元数和计费字符数,请向发布者模型端点发送 POST 请求。

在使用任何请求数据之前,请先进行以下替换:

  • LOCATION:输入支持的区域。如需查看支持的区域的完整列表,请参阅可用位置
  • PROJECT_ID:您的项目 ID
  • PREFIX:对于代码模型,prefix 表示一条有意义的编程代码的开头,或描述要生成的代码的自然语言提示。

HTTP 方法和网址:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/code-bison:countTokens

请求 JSON 正文:

{
  "instances": [
    {
      "prefix": "PREFIX"
    }
  ]
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

cat > request.json << 'EOF'
{
  "instances": [
    {
      "prefix": "PREFIX"
    }
  ]
}
EOF

然后,执行以下命令以发送 REST 请求:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/code-bison:countTokens"

PowerShell

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

@'
{
  "instances": [
    {
      "prefix": "PREFIX"
    }
  ]
}
'@  | Out-File -FilePath request.json -Encoding utf8

然后,执行以下命令以发送 REST 请求:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/code-bison:countTokens" | Select-Object -Expand Content

您应该收到类似以下内容的 JSON 响应。

{
  "totalTokens": 43,
  "totalBillableCharacters": 182
}

示例 curl 命令

PROJECT_ID=PROJECT_ID

curl \
-X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/us-central1/publishers/google/models/code-bison:countTokens -d \
$'{
  "instances": [
    {
      "prefix": "Write a function that checks if a year is a leap year."
    }
  ]
}'

codechat-bison

REST

如需使用 Vertex AI API 获取提示的词元数和计费字符数,请向发布者模型端点发送 POST 请求。

在使用任何请求数据之前,请先进行以下替换:

  • LOCATION:输入支持的区域。如需查看支持的区域的完整列表,请参阅可用位置
  • PROJECT_ID:您的项目 ID
  • AUTHOR:消息的作者。
  • CONTENT:消息的内容。

HTTP 方法和网址:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/codechat-bison:countTokens

请求 JSON 正文:

{
  "instances": {
    "messages": [
      {
         "author": "AUTHOR",
         "content": "CONTENT"
      }
    ]
  }
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

cat > request.json << 'EOF'
{
  "instances": {
    "messages": [
      {
         "author": "AUTHOR",
         "content": "CONTENT"
      }
    ]
  }
}
EOF

然后,执行以下命令以发送 REST 请求:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/codechat-bison:countTokens"

PowerShell

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

@'
{
  "instances": {
    "messages": [
      {
         "author": "AUTHOR",
         "content": "CONTENT"
      }
    ]
  }
}
'@  | Out-File -FilePath request.json -Encoding utf8

然后,执行以下命令以发送 REST 请求:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/codechat-bison:countTokens" | Select-Object -Expand Content

您应该收到类似以下内容的 JSON 响应。

{
  "totalTokens": 43,
  "totalBillableCharacters": 182
}

示例 curl 命令

PROJECT_ID=PROJECT_ID

curl \
-X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/us-central1/publishers/google/models/codechat-bison:countTokens -d \
$'{
  "instances": {
    "messages": [
      {
        "author": "user",
        "content": "Hi, how are you?"
      },
      {
        "author": "system",
        "content": "I am doing good. What Can I help you with in the coding world?"
      },
      {
        "author": "user",
        "content": "Please help write a function to calculate the min of two numbers"
      }
    ]
  }
}'

code-gecko

REST

如需使用 Vertex AI API 获取提示的词元数和计费字符数,请向发布者模型端点发送 POST 请求。

在使用任何请求数据之前,请先进行以下替换:

  • LOCATION:输入支持的区域。如需查看支持的区域的完整列表,请参阅可用位置
  • PROJECT_ID:您的项目 ID
  • PREFIX:对于代码模型,prefix 表示一条有意义的编程代码的开头,或描述要生成的代码的自然语言提示。模型尝试在 prefixsuffix 之间填充代码。
  • SUFFIX:对于代码补全,suffix 表示一条有意义的编程代码的结尾。模型尝试在 prefixsuffix 之间填充代码。

HTTP 方法和网址:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/code-gecko:countTokens

请求 JSON 正文:

{
  "instances": [
    {
      "prefix": "PREFIX",
      "suffix": "SUFFIX"
    }
  ]
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

cat > request.json << 'EOF'
{
  "instances": [
    {
      "prefix": "PREFIX",
      "suffix": "SUFFIX"
    }
  ]
}
EOF

然后,执行以下命令以发送 REST 请求:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/code-gecko:countTokens"

PowerShell

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

@'
{
  "instances": [
    {
      "prefix": "PREFIX",
      "suffix": "SUFFIX"
    }
  ]
}
'@  | Out-File -FilePath request.json -Encoding utf8

然后,执行以下命令以发送 REST 请求:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/code-gecko:countTokens" | Select-Object -Expand Content

您应该收到类似以下内容的 JSON 响应。

{
  "totalTokens": 43,
  "totalBillableCharacters": 182
}

示例 curl 命令

PROJECT_ID=PROJECT_ID

curl \
-X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/us-central1/publishers/google/models/code-gecko:countTokens -d \
$'{
  "instances": [
    {
      "prefix": "def reverse_string(s):",
      "suffix": ""
    }
  ]
}'

textembedding-gecko

REST

如需使用 Vertex AI API 获取提示的词元数和计费字符数,请向发布者模型端点发送 POST 请求。

在使用任何请求数据之前,请先进行以下替换:

  • LOCATION:输入支持的区域。如需查看支持的区域的完整列表,请参阅可用位置
  • PROJECT_ID:您的项目 ID
  • TEXT:您要为其生成嵌入的文本。

HTTP 方法和网址:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/textembedding-gecko:countTokens

请求 JSON 正文:

{
  "instances": [
    {
      "content": "TEXT"
    }
  ]
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

cat > request.json << 'EOF'
{
  "instances": [
    {
      "content": "TEXT"
    }
  ]
}
EOF

然后,执行以下命令以发送 REST 请求:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/textembedding-gecko:countTokens"

PowerShell

将请求正文保存在名为 request.json 的文件中。在终端中运行以下命令,在当前目录中创建或覆盖此文件:

@'
{
  "instances": [
    {
      "content": "TEXT"
    }
  ]
}
'@  | Out-File -FilePath request.json -Encoding utf8

然后,执行以下命令以发送 REST 请求:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/publishers/google/models/textembedding-gecko:countTokens" | Select-Object -Expand Content

您应该收到类似以下内容的 JSON 响应。

{
  "totalTokens": 43,
  "totalBillableCharacters": 182
}

示例 curl 命令

PROJECT_ID=PROJECT_ID

curl \
-X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json" \
https://us-central1-aiplatform.googleapis.com/v1/projects/${PROJECT_ID}/locations/us-central1/publishers/google/models/textembedding-gecko:countTokens -d \
$'{
  "instances": [
    {
      "content": "What is life?"
    }
  ]
}'

价格和配额

CountTokens API 可免费使用。CountTokens API 和 ComputeTokens API 的最大配额为每分钟 3000 个请求。

后续步骤