1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
|
# pylint: disable=line-too-long,useless-suppression
# ------------------------------------
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT License.
# ------------------------------------
"""
DESCRIPTION:
Given an AIProjectClient, this sample demonstrates how to use the asynchronous
`.fine_tuning_jobs` methods to create DPO (Direct Preference Optimization) fine-tuning jobs.
Supported OpenAI models: GPT-4o, GPT-4.1, GPT-4.1-mini, GPT-4.1-nano, and GPT-4o-mini.
USAGE:
python sample_finetuning_dpo_job_async.py
Before running the sample:
pip install azure-ai-projects>=2.0.0b1 azure-identity openai python-dotenv aiohttp
Set these environment variables with your own values:
1) AZURE_AI_PROJECT_ENDPOINT - Required. The Azure AI Project endpoint, as found in the overview page of your
Microsoft Foundry portal.
2) MODEL_NAME - Optional. The base model name to use for fine-tuning. Default to the `gpt-4o` model.
3) TRAINING_FILE_PATH - Optional. Path to the training data file. Default to the `data` folder.
4) VALIDATION_FILE_PATH - Optional. Path to the validation data file. Default to the `data` folder.
"""
import asyncio
import os
from dotenv import load_dotenv
from azure.identity.aio import DefaultAzureCredential
from azure.ai.projects.aio import AIProjectClient
from pathlib import Path
load_dotenv()
endpoint = os.environ["AZURE_AI_PROJECT_ENDPOINT"]
model_name = os.environ.get("MODEL_NAME", "gpt-4o-mini")
script_dir = Path(__file__).parent
training_file_path = os.environ.get("TRAINING_FILE_PATH", os.path.join(script_dir, "data", "dpo_training_set.jsonl"))
validation_file_path = os.environ.get(
"VALIDATION_FILE_PATH", os.path.join(script_dir, "data", "dpo_validation_set.jsonl")
)
async def main():
async with (
DefaultAzureCredential() as credential,
AIProjectClient(endpoint=endpoint, credential=credential) as project_client,
project_client.get_openai_client() as openai_client,
):
print("Uploading training file...")
with open(training_file_path, "rb") as f:
train_file = await openai_client.files.create(file=f, purpose="fine-tune")
print(f"Uploaded training file with ID: {train_file.id}")
print("Uploading validation file...")
with open(validation_file_path, "rb") as f:
validation_file = await openai_client.files.create(file=f, purpose="fine-tune")
print(f"Uploaded validation file with ID: {validation_file.id}")
print("Waits for the training and validation files to be processed...")
await openai_client.files.wait_for_processing(train_file.id)
await openai_client.files.wait_for_processing(validation_file.id)
print("Creating DPO fine-tuning job")
fine_tuning_job = await openai_client.fine_tuning.jobs.create(
training_file=train_file.id,
validation_file=validation_file.id,
model=model_name,
method={
"type": "dpo",
"dpo": {
"hyperparameters": {
"n_epochs": 3,
"batch_size": 1,
"learning_rate_multiplier": 1.0,
}
},
},
extra_body={
"trainingType": "Standard"
}, # Recommended approach to set trainingType. Omitting this field may lead to unsupported behavior.
)
print(fine_tuning_job)
if __name__ == "__main__":
asyncio.run(main())
|
