CLI Reference
Reference of all kagglex commands, flags, and options.
Global Options
| Option | Description |
|---|---|
-v, --verbose |
Enable debug logging output |
-q, --quiet |
Suppress non-error log output |
--help |
Show help message and exit |
kagglex run
Stage, submit, monitor, and retrieve outputs for a Kaggle experiment.
Options
| Flag | Type | Default | Description |
|---|---|---|---|
--file PATH |
Path | None | Standalone Python script to execute |
--dir PATH |
Path | Current Dir | Root directory of project to package |
--command CMD |
String | None | Command to run remotely (e.g. python -m pkg.train) |
--title TITLE |
String | Auto-generated | Title of the Kaggle notebook/kernel |
--slug SLUG |
String | Auto-generated | Slug for the Kaggle kernel (5-50 chars) |
--gpu TYPE |
String | t4-2x |
Hardware accelerator: t4-2x, p100, v3-8, none |
--multi-gpu |
Flag | False | Enable multi-GPU distributed training |
--dataset DATASET |
String (repeatable) | [] |
Kaggle dataset reference (username/slug) |
--include-data PATH |
Path (repeatable) | [] |
Local file or directory to package |
--extra-deps PKG |
String (repeatable) | [] |
Extra pip dependencies to install remotely |
--env KEY=VAL |
String (repeatable) | [] |
Remote environment variable |
--env-file PATH |
Path | None | Path to .env file |
--secret NAME |
String (repeatable) | [] |
Kaggle User Secret name |
--parent-kernel ID |
String (repeatable) | [] |
Kernel output to mount as input |
--auto-dataset |
Flag | False | Auto-publish large payload as a Kaggle dataset |
--auto-dataset-slug SLUG |
String | None | Custom slug for auto-dataset payload |
--no-internet |
Flag | False | Disable internet access in Kaggle environment |
--output-dir PATH |
Path | ./outputs |
Directory to save downloaded run outputs |
--include-outputs PATTERN |
String (repeatable) | [] |
Glob pattern for files to download |
--exclude-outputs PATTERN |
String (repeatable) | [] |
Glob pattern for files to ignore |
--no-wait |
Flag | False | Return immediately after submitting kernel |
--stream |
Flag | False | Stream remote logs in real-time |
--no-pull |
Flag | False | Skip downloading outputs after completion |
--dry-run |
Flag | False | Stage project locally without submitting to Kaggle |
kagglex exec
Interact with a live running Kaggle Jupyter proxy session.
Options
| Flag | Type | Default | Description |
|---|---|---|---|
CODE |
String | None | Inline Python code to execute remotely |
--url URL |
String | $KAGGLE_JUPYTER_URL |
Full URL of Kaggle Jupyter Proxy |
--file PATH |
Path | None | Local Python script to execute remotely |
--test |
Flag | False | Test connection and kernel health |
--gpu-info |
Flag | False | Display remote GPU device details and VRAM |
--upload PATH |
Path | None | Upload local file to /kaggle/working |
--download NAME |
String | None | Download file from /kaggle/working |
-o, --output PATH |
Path | None | Destination path for downloaded file |
--list-files |
Flag | False | List files in /kaggle/working |
--timeout SEC |
Integer | 120 |
Request timeout in seconds |
kagglex quota
Display estimated GPU and TPU accelerator consumption over a rolling window.
Options
| Flag | Type | Default | Description |
|---|---|---|---|
--days DAYS |
Integer | 7 |
Number of days in the rolling calculation window |
--gpu-limit HOURS |
Float | 30.0 |
Weekly GPU quota limit in hours |
--tpu-limit HOURS |
Float | 20.0 |
Weekly TPU quota limit in hours |
kagglex list
List recent experiment runs recorded in ~/.kagglex/runs.json.
Options
| Flag | Type | Default | Description |
|---|---|---|---|
--limit N |
Integer | 20 |
Maximum number of run records to display |
kagglex cancel
Cancel an active Kaggle kernel.
kagglex dataset push
Create or update a private or public Kaggle dataset.
Options
| Flag | Type | Default | Description |
|---|---|---|---|
--dir PATH |
Path | Required | Directory containing files to upload |
--title TITLE |
String | Required | Dataset title |
--slug SLUG |
String | Auto-generated | Dataset slug |
--public |
Flag | False | Make dataset publicly accessible |