Implementation:BerriAI Litellm Create Fine Tuning Job

Knowledge Sources	Domains	Last Updated
BerriAI/litellm	Fine-Tuning, API Integration, Model Customization	2026-02-15

Overview

Concrete tool for creating fine-tuning jobs across multiple LLM providers provided by LiteLLM.

Description

The create_fine_tuning_job function in LiteLLM initiates a fine-tuning process that customizes a base model using an uploaded training dataset. It accepts a model name, training file ID, optional hyperparameters, and a provider specification, then routes the request to the appropriate provider-specific handler (OpenAI, Azure OpenAI, or Vertex AI). The function constructs a typed FineTuningJobCreate payload, resolves API credentials through a priority chain, and returns a normalized LiteLLMFineTuningJob object containing the job ID, status, and metadata. Both synchronous (create_fine_tuning_job) and asynchronous (acreate_fine_tuning_job) variants are available.

Usage

Use create_fine_tuning_job when:

Starting a fine-tuning job after uploading training data via create_file.
Needing to customize a base model with domain-specific examples.
Building provider-agnostic fine-tuning workflows that can target OpenAI, Azure, or Vertex AI.

Code Reference

Source Location

litellm/fine_tuning/main.py (lines 37-285)

Signature

# Async version
async def acreate_fine_tuning_job(
    model: str,
    training_file: str,
    hyperparameters: Optional[dict] = {},
    suffix: Optional[str] = None,
    validation_file: Optional[str] = None,
    integrations: Optional[List[str]] = None,
    seed: Optional[int] = None,
    custom_llm_provider: Literal["openai", "azure", "vertex_ai"] = "openai",
    extra_headers: Optional[Dict[str, str]] = None,
    extra_body: Optional[Dict[str, str]] = None,
    **kwargs,
) -> LiteLLMFineTuningJob:

# Sync version
def create_fine_tuning_job(
    model: str,
    training_file: str,
    hyperparameters: Optional[dict] = {},
    suffix: Optional[str] = None,
    validation_file: Optional[str] = None,
    integrations: Optional[List[str]] = None,
    seed: Optional[int] = None,
    custom_llm_provider: Literal["openai", "azure", "vertex_ai"] = "openai",
    extra_headers: Optional[Dict[str, str]] = None,
    extra_body: Optional[Dict[str, str]] = None,
    **kwargs,
) -> Union[LiteLLMFineTuningJob, Coroutine[Any, Any, LiteLLMFineTuningJob]]:

Import

from litellm.fine_tuning.main import create_fine_tuning_job, acreate_fine_tuning_job

I/O Contract

Inputs

Parameter	Type	Required	Description
model	`str`	Yes	The name of the base model to fine-tune (e.g., "gpt-3.5-turbo", "gpt-4o-mini-2024-07-18").
training_file	`str`	Yes	The ID of an uploaded file containing training data (obtained from `create_file`).
hyperparameters	`Optional[dict]`	No	Dictionary with keys "batch_size", "learning_rate_multiplier", "n_epochs". Defaults to empty dict.
suffix	`Optional[str]`	No	A string (up to 18 characters) appended to the fine-tuned model name for identification.
validation_file	`Optional[str]`	No	The ID of an uploaded file containing validation data for evaluating training quality.
integrations	`Optional[List[str]]`	No	A list of integrations to enable for the fine-tuning job.
seed	`Optional[int]`	No	Seed for reproducibility of the training job.
custom_llm_provider	`Literal["openai", "azure", "vertex_ai"]`	No	The LLM provider to use. Defaults to "openai".
extra_headers	`Optional[Dict[str, str]]`	No	Additional HTTP headers to include in the request.
extra_body	`Optional[Dict[str, str]]`	No	Additional fields to include in the request body.
**kwargs	various	No	Additional parameters including `api_key`, `api_base`, `api_version`, `timeout`, `organization`, and provider-specific options.

Outputs

Return Type	Description
`LiteLLMFineTuningJob`	A normalized fine-tuning job object (extends OpenAI's `FineTuningJob`) containing: id (job identifier), model (base model), status (e.g., "validating_files", "queued", "running", "succeeded", "failed"), training_file, created_at, hyperparameters, fine_tuned_model (available upon completion), and error (if failed). Includes a `_hidden_params` dict for internal metadata and an optional `seed` field.

Usage Examples

Create a basic fine-tuning job with OpenAI

from litellm.fine_tuning.main import create_fine_tuning_job

job = create_fine_tuning_job(
    model="gpt-3.5-turbo",
    training_file="file-abc123",
    custom_llm_provider="openai",
)

print(job.id)      # e.g., "ftjob-abc123"
print(job.status)  # e.g., "validating_files"
print(job.model)   # "gpt-3.5-turbo"

Create a job with hyperparameters and validation

from litellm.fine_tuning.main import create_fine_tuning_job

job = create_fine_tuning_job(
    model="gpt-4o-mini-2024-07-18",
    training_file="file-abc123",
    hyperparameters={
        "batch_size": "auto",
        "learning_rate_multiplier": 0.1,
        "n_epochs": 3,
    },
    suffix="my-custom-model",
    validation_file="file-xyz789",
    seed=42,
    custom_llm_provider="openai",
)

print(job.id)
print(job.hyperparameters)

Async job creation with Azure OpenAI

import asyncio
from litellm.fine_tuning.main import acreate_fine_tuning_job

async def start_fine_tuning():
    job = await acreate_fine_tuning_job(
        model="gpt-35-turbo",
        training_file="file-abc123",
        custom_llm_provider="azure",
        api_base="https://my-resource.openai.azure.com/",
        api_key="my-azure-api-key",
        api_version="2024-02-01",
    )
    return job

result = asyncio.run(start_fine_tuning())
print(result.id)
print(result.status)

Create a Vertex AI fine-tuning job

from litellm.fine_tuning.main import create_fine_tuning_job

job = create_fine_tuning_job(
    model="gemini-1.0-pro-002",
    training_file="file-vertex-123",
    custom_llm_provider="vertex_ai",
    vertex_project="my-gcp-project",
    vertex_location="us-central1",
)

print(job.id)
print(job.status)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment