Skip to content

upload-kaggle

The upload-kaggle command publishes datasets to Kaggle.

Basic Usage

Upload a dataset file to Kaggle:

Basic upload
deepfabric upload-kaggle dataset.jsonl --handle username/dataset-name

Authentication Methods

The upload-kaggle command supports multiple authentication approaches:

Most secure approach for production environments:

export KAGGLE_USERNAME="your-kaggle-username"
export KAGGLE_KEY="your-kaggle-api-key"
deepfabric upload-kaggle dataset.jsonl --handle username/dataset-name

Credential specification directly in the command:

deepfabric upload-kaggle dataset.jsonl --handle username/dataset-name --username your-username --key your-api-key

Options

Option Description
--handle Kaggle dataset handle (required)
--username Kaggle username (or set KAGGLE_USERNAME env var)
--key Kaggle API key (or set KAGGLE_KEY env var)
--tags Tags for the dataset (can be specified multiple times)
--version-notes Version notes for the dataset update
--description Description for the dataset

Dataset Tagging

Customize dataset discoverability through tag specification:

Add tags
deepfabric upload-kaggle dataset.jsonl \
  --handle username/educational-content \
  --tags educational \
  --tags programming \
  --tags synthetic

Version Notes

When updating an existing dataset, provide version notes to document changes:

Version notes
deepfabric upload-kaggle dataset.jsonl \
  --handle username/my-dataset \
  --version-notes "Added 500 new samples for edge cases"

Dataset Description

Provide a description for new datasets:

With description
deepfabric upload-kaggle dataset.jsonl \
  --handle username/new-dataset \
  --description "Synthetic dataset for training code generation models"

Batch Upload Operations

Upload multiple related datasets:

Multiple uploads
# Upload training and validation sets
deepfabric upload-kaggle train_dataset.jsonl --handle myorg/comprehensive-dataset --tags training
deepfabric upload-kaggle val_dataset.jsonl --handle myorg/comprehensive-dataset --tags validation