Implementation:BerriAI Litellm Fine Tuning Types

Knowledge Sources	Domains	Last Updated
BerriAI/litellm	Machine Learning, API Types, Data Validation	2026-02-15

Overview

Concrete type definitions for fine-tuning job creation and hyperparameter configuration provided by LiteLLM.

Description

LiteLLM defines a set of Pydantic models that represent the data structures required to create fine-tuning jobs across multiple LLM providers. These types enforce schema validation at the Python level before any API call is made, ensuring that training file references, hyperparameter values, and provider-specific options conform to expected formats. The type hierarchy includes a base FineTuningJobCreate model aligned with the OpenAI API specification, a Hyperparameters model for training control parameters, and LiteLLMFineTuningJobCreate which extends the base with a custom_llm_provider field and permissive extra-field handling for cross-provider compatibility.

Usage

Import and use these types when:

• Constructing fine-tuning job request payloads programmatically.
• Validating hyperparameter values before submitting a job.
• Building provider-agnostic fine-tuning workflows that need to specify which LLM provider to target.

Code Reference

Source Location

• Hyperparameters: litellm/types/llms/openai.py (lines 949-956)
• FineTuningJobCreate: litellm/types/llms/openai.py (lines 959-995)
• LiteLLMFineTuningJobCreate: litellm/types/llms/openai.py (lines 998-1003)
• OpenAIFineTuningHyperparameters: litellm/types/fine_tuning.py (lines 1-5)

Signature

class Hyperparameters(BaseModel):
    batch_size: Optional[Union[str, int]] = None
    learning_rate_multiplier: Optional[Union[str, float]] = None
    n_epochs: Optional[Union[str, int]] = None


class FineTuningJobCreate(BaseModel):
    model: str
    training_file: str
    hyperparameters: Optional[Hyperparameters] = None
    suffix: Optional[str] = None
    validation_file: Optional[str] = None
    integrations: Optional[List[str]] = None
    seed: Optional[int] = None


class LiteLLMFineTuningJobCreate(FineTuningJobCreate):
    custom_llm_provider: Optional[Literal["openai", "azure", "vertex_ai"]] = None
    model_config = {"extra": "allow"}


class OpenAIFineTuningHyperparameters(Hyperparameters):
    model_config = {"extra": "allow"}

Import

from litellm.types.llms.openai import (
    Hyperparameters,
    FineTuningJobCreate,
    LiteLLMFineTuningJobCreate,
)
from litellm.types.fine_tuning import OpenAIFineTuningHyperparameters

I/O Contract

Inputs (FineTuningJobCreate)

Parameter	Type	Required	Description
model	str	Yes	The name of the base model to fine-tune (e.g., "gpt-3.5-turbo").
training_file	str	Yes	The ID of an uploaded file containing training data.
hyperparameters	Optional[Hyperparameters]	No	Training hyperparameters (batch_size, learning_rate_multiplier, n_epochs).
suffix	Optional[str]	No	A string of up to 18 characters appended to the fine-tuned model name.
validation_file	Optional[str]	No	The ID of an uploaded file containing validation data.
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.

Inputs (Hyperparameters)

Parameter	Type	Required	Description
batch_size	Optional[Union[str, int]]	No	Number of examples per batch, or "auto" for provider default.
learning_rate_multiplier	Optional[Union[str, float]]	No	Scaling factor for the learning rate.
n_epochs	Optional[Union[str, int]]	No	Number of training epochs, or "auto" for provider default.

Outputs

These are Pydantic model instances used as input data containers. They do not directly produce API responses. They are serialized via model_dump(exclude_none=True) to produce dictionaries sent to provider APIs.

Usage Examples

Creating a basic fine-tuning job request

from litellm.types.llms.openai import (
    FineTuningJobCreate,
    Hyperparameters,
)

hyperparams = Hyperparameters(
    batch_size="auto",
    learning_rate_multiplier=0.1,
    n_epochs=3,
)

job_request = FineTuningJobCreate(
    model="gpt-3.5-turbo",
    training_file="file-abc123",
    hyperparameters=hyperparams,
    suffix="my-custom-model",
    validation_file="file-xyz789",
    seed=42,
)

# Serialize for API submission
payload = job_request.model_dump(exclude_none=True)

Creating a provider-specific request

from litellm.types.llms.openai import LiteLLMFineTuningJobCreate

job_request = LiteLLMFineTuningJobCreate(
    model="gpt-4o-mini-2024-07-18",
    training_file="file-abc123",
    custom_llm_provider="openai",
    suffix="domain-expert",
)

Related Pages

• Principle:BerriAI_Litellm_Training_Data_Preparation