Implementation:Cohere ai Cohere python FinetuneDatasetMetrics Model

Knowledge Sources	Cohere Python SDK
Domains	SDK, Fine-tuning
Last Updated	2026-02-15 14:00 GMT

Overview

FinetuneDatasetMetrics is a Pydantic model representing statistical metrics about a fine-tuning dataset, including token counts, example counts, and byte sizes for training and evaluation splits.

Description

The FinetuneDatasetMetrics class provides quantitative information about a dataset used for fine-tuning a Cohere model. All fields are optional, as metrics may not be fully computed at all stages of the fine-tuning pipeline. The metrics cover:

• trainable_token_count: The number of tokens from valid examples that can actually be used for training
• total_examples: The total number of examples in the dataset
• train_examples: The number of examples allocated to the training split
• train_size_bytes: The total byte size of training examples
• eval_examples: The number of examples allocated to the evaluation split
• eval_size_bytes: The total byte size of evaluation examples

These metrics are useful for understanding dataset composition, estimating training costs, and verifying that train/eval splits are properly balanced.

The class extends UncheckedBaseModel and is auto-generated by the Fern API definition toolchain.

Usage

Use FinetuneDatasetMetrics when inspecting the details of a fine-tuning job or dataset to understand the volume and distribution of training data. This model is typically accessed as a nested field within fine-tuning job or dataset response objects.

Code Reference

Source Location

• Repository: Cohere Python SDK
• File: src/cohere/types/finetune_dataset_metrics.py

Signature

class FinetuneDatasetMetrics(UncheckedBaseModel):
    trainable_token_count: typing.Optional[int] = None
    total_examples: typing.Optional[int] = None
    train_examples: typing.Optional[int] = None
    train_size_bytes: typing.Optional[int] = None
    eval_examples: typing.Optional[int] = None
    eval_size_bytes: typing.Optional[int] = None

Import

from cohere.types import FinetuneDatasetMetrics

I/O Contract

Fields

Field	Type	Required	Default	Description
trainable_token_count	Optional[int]	No	None	The number of tokens of valid examples that can be used for training
total_examples	Optional[int]	No	None	The overall number of examples in the dataset
train_examples	Optional[int]	No	None	The number of training examples
train_size_bytes	Optional[int]	No	None	The size in bytes of all training examples
eval_examples	Optional[int]	No	None	The number of evaluation examples
eval_size_bytes	Optional[int]	No	None	The size in bytes of all evaluation examples

Usage Examples

Inspecting Fine-tune Dataset Metrics

import cohere

co = cohere.Client()

# Get fine-tuned model details
finetuned_model = co.finetuning.get_finetuned_model(id="my-finetune-id")

# Access dataset metrics from the fine-tuning job
metrics = finetuned_model.settings.dataset_metrics
if metrics:
    print(f"Total examples: {metrics.total_examples}")
    print(f"Training examples: {metrics.train_examples}")
    print(f"Evaluation examples: {metrics.eval_examples}")
    print(f"Trainable token count: {metrics.trainable_token_count}")
    if metrics.train_size_bytes:
        print(f"Training data size: {metrics.train_size_bytes / 1024:.1f} KB")
    if metrics.eval_size_bytes:
        print(f"Evaluation data size: {metrics.eval_size_bytes / 1024:.1f} KB")

Constructing Metrics Directly

from cohere.types import FinetuneDatasetMetrics

metrics = FinetuneDatasetMetrics(
    trainable_token_count=150000,
    total_examples=1000,
    train_examples=800,
    train_size_bytes=2048000,
    eval_examples=200,
    eval_size_bytes=512000,
)

train_eval_ratio = metrics.train_examples / metrics.total_examples
print(f"Train/eval split: {train_eval_ratio:.0%} / {1 - train_eval_ratio:.0%}")
print(f"Avg tokens per example: {metrics.trainable_token_count / metrics.total_examples:.0f}")

Related Pages

• Environment:Cohere_ai_Cohere_python_Python_SDK_Runtime