Jupyter中OpenAI对象无fine_tunes属性报错，求解决方案

报错信息 #

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[10], line 5
      2 dataset_path = r"*****"
      4 # Fine-tune the model with the examples from the dataset
----> 5 fine_tuned_model_id = fine_tune_model_with_examples(dataset_path)
      6 print(f"Fine-tuned model ID: {fine_tuned_model_id}")
      8 # Example source code to analyze

Cell In[9], line 34, in fine_tune_model_with_examples(dataset_path)
     26 # Launch fine-tuning
     27 fine_tuning_parameters = {
     28     "training_file": train_file_id,
     29     "validation_file": test_file_id,
     30     "model": "gpt-3.5-turbo",
     31     "n_epochs": 4
     32 }
---> 34 response = client.fine_tunes.create(**fine_tuning_parameters)
     35 return response['id']

AttributeError: 'OpenAI' object has no attribute 'fine_tunes'

引发报错的代码 #

# Function to fine-tune the model
def fine_tune_model_with_examples(dataset_path):
    # Load and balance the dataset
    balanced_dataset = load_and_balance_dataset(dataset_path)

    # Split the dataset into training and test sets
    train_set, test_set = split_dataset(balanced_dataset)

    # Prepare fine-tuning data in chat format
    train_examples = prepare_fine_tuning_data(train_set)
    test_examples = prepare_fine_tuning_data(test_set)

    # Save training and test data to JSONL files
    save_to_jsonl(train_examples, "train_fine_tuning_data.jsonl")
    save_to_jsonl(test_examples, "test_fine_tuning_data.jsonl")

    # Upload training and test data to OpenAI
    with open("train_fine_tuning_data.jsonl", 'rb') as train_file:
        train_file_response = client.files.create(file=train_file, purpose='fine-tune')
    with open("test_fine_tuning_data.jsonl", 'rb') as test_file:
        test_file_response = client.files.create(file=test_file, purpose='fine-tune')

    train_file_id = train_file_response.id
    test_file_id = test_file_response.id

    # Launch fine-tuning
    fine_tuning_parameters = {
        "training_file": train_file_id,
        "validation_file": test_file_id,
        "model": "gpt-3.5-turbo",
        "n_epochs": 4
    }

    response = client.fine_tunes.create(**fine_tuning_parameters)
    return response['id']

问题背景 #

在Jupyter Notebook 7.0.8环境下用Python开发，运行代码时出现上述报错。已更新所有依赖包，且按照OpenAI API更新说明将openai.FineTune.create()调整为client.fine_tunes.create()，但错误仍存在。

解决方案 #

核心原因 #

你使用的是OpenAI Python SDK v1.x版本，该版本对微调接口做了重构：旧版的fine_tunes模块被替换为fine_tuning.jobs，同时参数结构和返回值处理逻辑也有变化。

具体修改步骤 #

1. 确认SDK版本：执行pip show openai查看版本号，若为v1.x开头，则需按以下方式修改代码。
2. 调整接口调用路径：将client.fine_tunes.create()替换为client.fine_tuning.jobs.create()。
3. 修改参数结构：原参数n_epochs需要嵌套在hyperparameters字典中传递。
4. 更新返回值获取方式：v1.x版本中返回的是对象，需用response.id替代response['id']。

修改后的代码片段 #

# Launch fine-tuning
fine_tuning_parameters = {
    "training_file": train_file_id,
    "validation_file": test_file_id,
    "model": "gpt-3.5-turbo",
    "hyperparameters": {"n_epochs": 4}
}

response = client.fine_tuning.jobs.create(**fine_tuning_parameters)
return response.id

内容的提问来源于stack exchange，提问作者Codezilla