---
title: "Quick Revision"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Quick Revision}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
library(mlr3hf)
```

# Quick Revision

This vignette provides a quick overview of how to use **mlr3hf**.

Suppose you are working on **neuromorphic computing** and need a dataset for your machine learning project.

### Step 1: Search for a Dataset

First, search for relevant datasets on the Hugging Face Hub.

```{r}
library(mlr3hf)
library(mlr3)

list_datasets(num_of_datasets = 5, search = "neuromorphic computing")
```

This returns a list of dataset repositories matching your search term.
**Note:** If search dataset can return less that the number of datasets specified in `num_of_datasets`, it means that there are no more datasets available for the search term.

### Step 2: Create an `HFData` Object

Next, create an `HFData` object using a `repo_id` obtained from `list_datasets()` above.

```{r}
dt = hfdt(repo_id = "Shoriful025/neuromorphic_edge_computing_telemetry")
```

To retrieve the link to the repository on Hugging Face:

```{r}
dt$repo_link
```

You can also inspect the available dataset configurations and files:

```{r}
dt$configs
dt$siblings
```

### Step 3: Select a Configuration

Once you know which configuration you want, recreate the object by specifying it. Here, the configuration is one of the values returned by `dt$configs` above.

```{r}
dt = hfdt(
  repo_id = "Shoriful025/neuromorphic_edge_computing_telemetry",
  config = "default"
)
```

The dataset is downloaded locally, a backend is created, and the dataset can now be inspected.

To view the column names or load the data:

```{r}
dt$colnames
dt$data[1:5]
```

### Step 4: Create an `mlr3` Task

To create an `mlr3` task, specify a target column — one of the values returned by `dt$colnames` above.

```{r}
task = as_task(
  dt,
  target_names = "temperature_c"
)
```

### Shortcut: Specify Everything at Once

If you already know the repository, configuration, and target column, you can create everything directly:

```{r}
dt = hfdt(
  repo_id = "Shoriful025/neuromorphic_edge_computing_telemetry",
  config = "default",
  target = "temperature_c"
)

task = as_task(dt)
task
```

Alternatively, `tsk()` can be used to create the task directly, without going through an intermediate `HFData` object:

```{r}
task = tsk("hf",
  repo_id = "Shoriful025/neuromorphic_edge_computing_telemetry",
  config = "default",
  target = "temperature_c"
)
task
```

This covers the essential workflow for using `mlr3hf` to search, load, and convert Hugging Face datasets into `mlr3` tasks.