优化模型

管理精细调整工作，以使模型适应您特定的训练数据。相关指南：精细调整模型

创建优化任务

POST https://api.openai.com/v1/fine-tunes

创建一个任务，从给定的数据集中微调指定模型。

响应包括已入队的任务的详细信息，包括任务状态和完成后微调模型的名称。

了解更多关于微调的内容。

请求体

training_file string 必填项上传的包含训练数据的文件的ID。
请参阅上传文件以了解如何上传文件。
您的数据集必须格式化为JSONL文件，其中每个训练示例都是一个具有“prompt”和“completion”键的JSON对象。此外，您必须以微调目的上传文件。
请参阅微调指南以获取更多详细信息。
validation_file string 可选项上传的包含验证数据的文件的ID。
如果提供此文件，则在微调过程中会定期使用该数据生成验证指标。这些指标可以在微调结果文件中查看。您的训练和验证数据应该是互斥的。
您的数据集必须格式化为JSONL文件，其中每个验证示例都是一个具有“prompt”和“completion”键的JSON对象。此外，您必须以微调目的上传文件。
请参阅微调指南以获取更多详细信息。
model string 可选项默认为curie 要微调的基础模型的名称。您可以选择其中之一：“ada”，“babbage”，“curie”，“davinci”或2022-04-21之后创建的微调模型。要了解有关这些模型的更多信息，请参阅模型文档。
n_epochs integer 可选项默认为4 用于训练模型的时期数。时期是指完整遍历训练数据集的一个周期。
batch_size integer 可选项默认为null 用于训练的批处理大小。批处理大小是用于训练单个前向和后向传递的训练示例的数量。默认情况下，批处理大小将动态配置为训练集中示例数的约0.2％，上限为256-通常，我们发现较大的批处理大小 tend to work better for larger datasets。
learning_rate_multiplier number 可选项默认为null 用于训练的学习率乘数。微调学习率是用于预训练的原始学习率乘以此值。
默认情况下，学习率乘数为0.05、0.1或0.2，具体取决于最终批处理大小（较大的批处理大小 tend to perform better with larger batch sizes）。我们建议尝试0.02到0.2范围内的值，以查看哪些值可以产生最佳结果。
prompt_loss_weight number 可选项默认为0.01 用于提示令牌丢失的权重。这控制模型尝试学习生成提示的程度（与始终具有1.0权重的完成进行比较），并且可以在完成很短时为训练添加稳定效果。
如果提示非常长（相对于完成），则可能有意义降低此权重，以避免过度优先考虑学习提示。
compute_classification_metrics boolean 可选项默认为false 如果设置，我们使用验证集在每个时期结束时计算分类特定的指标，例如准确性和F-1分数。这些指标可以在结果文件中查看。
为了计算分类指标，您必须提供validation_file。此外，您必须为多类分类指定classification_n_classes或为二元分类指定classification_positive_class。
classification_n_classes integer 可选项默认为null 分类任务中的类数。
对于多类分类，此参数是必需的。
classification_positive_class string 可选项默认为null 二元分类中的正类。
进行二元分类时，此参数需要生成精度、召回率和F1指标。
classification_betas array 可选项默认为null 如果提供此项，则在指定的beta值上计算F-beta分数。F-beta分数是F-1分数的概括。仅用于二元分类。
对于beta值为1（即F-1分数），精度和召回率具有相同的权重。较大的beta分数将更多的权重放在召回上，而较少的beta分数将更多的权重放在精度上。
suffix string 可选项默认为null 最多40个字符的字符串，将添加到您的微调模型名称中。
例如，“custom-model-name”的后缀将产生类似ada:ft-your-org:custom-model-name-2022-02-15-04-21-04的模型名称。

请求示例

curl https://api.openai.com/v1/fine-tunes \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "training_file": "file-XGinujblHPwGLSztz8cPS8XY"
  }'

import os
import openai
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.FineTune.create(training_file="file-XGinujblHPwGLSztz8cPS8XY")

const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
const response = await openai.createFineTune({
  training_file: "file-XGinujblHPwGLSztz8cPS8XY",
});

返回结果

{
  "id": "ft-AF1WoRqd3aJAHsqc9NY7iL8F",
  "object": "fine-tune",
  "model": "curie",
  "created_at": 1614807352,
  "events": [
    {
      "object": "fine-tune-event",
      "created_at": 1614807352,
      "level": "info",
      "message": "Job enqueued. Waiting for jobs ahead to complete. Queue number: 0."
    }
  ],
  "fine_tuned_model": null,
  "hyperparams": {
    "batch_size": 4,
    "learning_rate_multiplier": 0.1,
    "n_epochs": 4,
    "prompt_loss_weight": 0.1,
  },
  "organization_id": "org-...",
  "result_files": [],
  "status": "pending",
  "validation_files": [],
  "training_files": [
    {
      "id": "file-XGinujblHPwGLSztz8cPS8XY",
      "object": "file",
      "bytes": 1547276,
      "created_at": 1610062281,
      "filename": "my-data-train.jsonl",
      "purpose": "fine-tune-train"
    }
  ],
  "updated_at": 1614807352,
}

优化任务列表

GET https://api.openai.com/v1/fine-tunes

列出您组织的微调任务。

请求示例

curl https://api.openai.com/v1/fine-tunes \
  -H "Authorization: Bearer $OPENAI_API_KEY"

import os
import openai
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.FineTune.list()

const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
const response = await openai.listFineTunes();

返回结果

{
  "object": "list",
  "data": [
    {
      "id": "ft-AF1WoRqd3aJAHsqc9NY7iL8F",
      "object": "fine-tune",
      "model": "curie",
      "created_at": 1614807352,
      "fine_tuned_model": null,
      "hyperparams": { ... },
      "organization_id": "org-...",
      "result_files": [],
      "status": "pending",
      "validation_files": [],
      "training_files": [ { ... } ],
      "updated_at": 1614807352,
    },
    { ... },
    { ... }
  ]
}

获取优化任务信息

GET https://api.openai.com/v1/fine-tunes/{fine_tune_id}

获取有关微调工作的信息。

了解更多关于微调的内容。

路径参数

fine_tune_id 字符串必填项微调作业的ID

请求示例

curl https://api.openai.com/v1/fine-tunes/ft-AF1WoRqd3aJAHsqc9NY7iL8F \
  -H "Authorization: Bearer $OPENAI_API_KEY"

import os
import openai
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.FineTune.retrieve(id="ft-AF1WoRqd3aJAHsqc9NY7iL8F")

const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
const response = await openai.retrieveFineTune("ft-AF1WoRqd3aJAHsqc9NY7iL8F");

返回结果

{
  "id": "ft-AF1WoRqd3aJAHsqc9NY7iL8F",
  "object": "fine-tune",
  "model": "curie",
  "created_at": 1614807352,
  "events": [
    {
      "object": "fine-tune-event",
      "created_at": 1614807352,
      "level": "info",
      "message": "Job enqueued. Waiting for jobs ahead to complete. Queue number: 0."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807356,
      "level": "info",
      "message": "Job started."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807861,
      "level": "info",
      "message": "Uploaded snapshot: curie:ft-acmeco-2021-03-03-21-44-20."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807864,
      "level": "info",
      "message": "Uploaded result files: file-QQm6ZpqdNwAaVC3aSz5sWwLT."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807864,
      "level": "info",
      "message": "Job succeeded."
    }
  ],
  "fine_tuned_model": "curie:ft-acmeco-2021-03-03-21-44-20",
  "hyperparams": {
    "batch_size": 4,
    "learning_rate_multiplier": 0.1,
    "n_epochs": 4,
    "prompt_loss_weight": 0.1,
  },
  "organization_id": "org-...",
  "result_files": [
    {
      "id": "file-QQm6ZpqdNwAaVC3aSz5sWwLT",
      "object": "file",
      "bytes": 81509,
      "created_at": 1614807863,
      "filename": "compiled_results.csv",
      "purpose": "fine-tune-results"
    }
  ],
  "status": "succeeded",
  "validation_files": [],
  "training_files": [
    {
      "id": "file-XGinujblHPwGLSztz8cPS8XY",
      "object": "file",
      "bytes": 1547276,
      "created_at": 1610062281,
      "filename": "my-data-train.jsonl",
      "purpose": "fine-tune-train"
    }
  ],
  "updated_at": 1614807865,
}

取消优化任务

POST https://api.openai.com/v1/fine-tunes/{fine_tune_id}/cancel

立即取消微调作业。

路径参数

fine_tune_id 字符串必填要取消的fine-tune作业的ID

请求示例

curl https://api.openai.com/v1/fine-tunes/ft-AF1WoRqd3aJAHsqc9NY7iL8F/cancel \
  -H "Authorization: Bearer $OPENAI_API_KEY"

import os
import openai
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.FineTune.cancel(id="ft-AF1WoRqd3aJAHsqc9NY7iL8F")

const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
const response = await openai.cancelFineTune("ft-AF1WoRqd3aJAHsqc9NY7iL8F");

返回结果

{
  "id": "ft-xhrpBbvVUzYGo8oUO1FY4nI7",
  "object": "fine-tune",
  "model": "curie",
  "created_at": 1614807770,
  "events": [ { ... } ],
  "fine_tuned_model": null,
  "hyperparams": { ... },
  "organization_id": "org-...",
  "result_files": [],
  "status": "cancelled",
  "validation_files": [],
  "training_files": [
    {
      "id": "file-XGinujblHPwGLSztz8cPS8XY",
      "object": "file",
      "bytes": 1547276,
      "created_at": 1610062281,
      "filename": "my-data-train.jsonl",
      "purpose": "fine-tune-train"
    }
  ],
  "updated_at": 1614807789,
}

获取优化任务事件状态

GET https://api.openai.com/v1/fine-tunes/{fine_tune_id}/events

获取优化任务的详细状态更新。

路径参数

fine_tune_id 字符串必需的获取事件的微调作业的ID。

查询参数

stream 布尔值可选的默认为false 是否为微调作业流式传输事件。如果设置为true，则事件将作为数据传输服务器推送事件随时可用。当作业完成（成功、取消或失败）时，流将以数据：[DONE]消息终止。
如果设置为false，则仅返回到目前为止生成的事件。

请求示例

curl https://api.openai.com/v1/fine-tunes/ft-AF1WoRqd3aJAHsqc9NY7iL8F/events \
  -H "Authorization: Bearer $OPENAI_API_KEY"

import os
import openai
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.FineTune.list_events(id="ft-AF1WoRqd3aJAHsqc9NY7iL8F")

const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
const response = await openai.listFineTuneEvents("ft-AF1WoRqd3aJAHsqc9NY7iL8F");

返回结果

{
  "object": "list",
  "data": [
    {
      "object": "fine-tune-event",
      "created_at": 1614807352,
      "level": "info",
      "message": "Job enqueued. Waiting for jobs ahead to complete. Queue number: 0."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807356,
      "level": "info",
      "message": "Job started."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807861,
      "level": "info",
      "message": "Uploaded snapshot: curie:ft-acmeco-2021-03-03-21-44-20."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807864,
      "level": "info",
      "message": "Uploaded result files: file-QQm6ZpqdNwAaVC3aSz5sWwLT."
    },
    {
      "object": "fine-tune-event",
      "created_at": 1614807864,
      "level": "info",
      "message": "Job succeeded."
    }
  ]
}

删除优化的模型

DELETE https://api.openai.com/v1/models/{model}

删除一个精调过的模型。您必须在组织中具有所有者角色。

路径参数

model 字符串必填项要删除的模型

请求示例

curl https://api.openai.com/v1/models/curie:ft-acmeco-2021-03-03-21-44-20 \
  -X DELETE \
  -H "Authorization: Bearer $OPENAI_API_KEY"

import os
import openai
openai.api_key = os.getenv("OPENAI_API_KEY")
openai.Model.delete("curie:ft-acmeco-2021-03-03-21-44-20")

const { Configuration, OpenAIApi } = require("openai");
const configuration = new Configuration({
  apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration);
const response = await openai.deleteModel('curie:ft-acmeco-2021-03-03-21-44-20');

返回结果

{
  "id": "curie:ft-acmeco-2021-03-03-21-44-20",
  "object": "model",
  "deleted": true
}

最后更新于1年前

这有帮助吗？

const { Configuration, OpenAIApi } = require("openai"); const configuration = new Configuration({ apiKey: process.env.OPENAI_API_KEY, }); const openai = new OpenAIApi(configuration); const response = await openai.createFineTune({ training_file: "file-XGinujblHPwGLSztz8cPS8XY", });

{ "object": "list", "data": [ { "id": "ft-AF1WoRqd3aJAHsqc9NY7iL8F", "object": "fine-tune", "model": "curie", "created_at": 1614807352, "fine_tuned_model": null, "hyperparams": { ... }, "organization_id": "org-...", "result_files": [], "status": "pending", "validation_files": [], "training_files": [ { ... } ], "updated_at": 1614807352, }, { ... }, { ... } ] }

{ "id": "ft-AF1WoRqd3aJAHsqc9NY7iL8F", "object": "fine-tune", "model": "curie", "created_at": 1614807352, "events": [ { "object": "fine-tune-event", "created_at": 1614807352, "level": "info", "message": "Job enqueued. Waiting for jobs ahead to complete. Queue number: 0." }, { "object": "fine-tune-event", "created_at": 1614807356, "level": "info", "message": "Job started." }, { "object": "fine-tune-event", "created_at": 1614807861, "level": "info", "message": "Uploaded snapshot: curie:ft-acmeco-2021-03-03-21-44-20." }, { "object": "fine-tune-event", "created_at": 1614807864, "level": "info", "message": "Uploaded result files: file-QQm6ZpqdNwAaVC3aSz5sWwLT." }, { "object": "fine-tune-event", "created_at": 1614807864, "level": "info", "message": "Job succeeded." } ], "fine_tuned_model": "curie:ft-acmeco-2021-03-03-21-44-20", "hyperparams": { "batch_size": 4, "learning_rate_multiplier": 0.1, "n_epochs": 4, "prompt_loss_weight": 0.1, }, "organization_id": "org-...", "result_files": [ { "id": "file-QQm6ZpqdNwAaVC3aSz5sWwLT", "object": "file", "bytes": 81509, "created_at": 1614807863, "filename": "compiled_results.csv", "purpose": "fine-tune-results" } ], "status": "succeeded", "validation_files": [], "training_files": [ { "id": "file-XGinujblHPwGLSztz8cPS8XY", "object": "file", "bytes": 1547276, "created_at": 1610062281, "filename": "my-data-train.jsonl", "purpose": "fine-tune-train" } ], "updated_at": 1614807865, }

{ "id": "ft-xhrpBbvVUzYGo8oUO1FY4nI7", "object": "fine-tune", "model": "curie", "created_at": 1614807770, "events": [ { ... } ], "fine_tuned_model": null, "hyperparams": { ... }, "organization_id": "org-...", "result_files": [], "status": "cancelled", "validation_files": [], "training_files": [ { "id": "file-XGinujblHPwGLSztz8cPS8XY", "object": "file", "bytes": 1547276, "created_at": 1610062281, "filename": "my-data-train.jsonl", "purpose": "fine-tune-train" } ], "updated_at": 1614807789, }

const { Configuration, OpenAIApi } = require("openai"); const configuration = new Configuration({ apiKey: process.env.OPENAI_API_KEY, }); const openai = new OpenAIApi(configuration); const response = await openai.listFineTuneEvents("ft-AF1WoRqd3aJAHsqc9NY7iL8F");

{ "object": "list", "data": [ { "object": "fine-tune-event", "created_at": 1614807352, "level": "info", "message": "Job enqueued. Waiting for jobs ahead to complete. Queue number: 0." }, { "object": "fine-tune-event", "created_at": 1614807356, "level": "info", "message": "Job started." }, { "object": "fine-tune-event", "created_at": 1614807861, "level": "info", "message": "Uploaded snapshot: curie:ft-acmeco-2021-03-03-21-44-20." }, { "object": "fine-tune-event", "created_at": 1614807864, "level": "info", "message": "Uploaded result files: file-QQm6ZpqdNwAaVC3aSz5sWwLT." }, { "object": "fine-tune-event", "created_at": 1614807864, "level": "info", "message": "Job succeeded." } ] }

const { Configuration, OpenAIApi } = require("openai"); const configuration = new Configuration({ apiKey: process.env.OPENAI_API_KEY, }); const openai = new OpenAIApi(configuration); const response = await openai.deleteModel('curie:ft-acmeco-2021-03-03-21-44-20');