# Magic

Wherever you are, whatever strange function you are building up, `Magic Object` allows you to easily develop and interact with:skull:[Innocuous Book experiment](https://innocuous-book.gitbook.io/innocuous-book/experiment-1/experiment).

### Let's start:soon:

Instantiate `Magic Object` before use.

```
mj = MagicObj()
```

:arrow\_down: Then, enjoy Innocuous Book Magic power:sparkles:

### Methods:notebook:

#### get\_path

```
get_path(filename, path)
```

Search the data by `filename` from `path` , if exist, return the data path.

| Arguments |                                                                 |
| --------- | --------------------------------------------------------------- |
| filename  | String (name of data with extension which you want to download) |
| path      | String (directory where data put)                               |

Example: 

```
data_path = mj.get_path(
                filename = "boston_housing.csv",
                path = "/home/project/workspace/dataset")
data = pd.read_csv(data_path)
```

##

#### log

```
log(**kwargs)
```

Log metrics then you can choose one as optimal standard.

{% hint style="warning" %}
 Do not use `log` within a `Trainable` class.
{% endhint %}

| Arguments  |                                          |
| ---------- | ---------------------------------------- |
| \*\*kwargs | Arguments to log which must have **key** |

Example:

```
mj.log(loss=running_loss, eval_loss=eval_loss)
```

##

so u can read the data with that save ur checkpoints no matter what framework api for every single framework log you metrics

##

#### torch\_get\_checkpoint\_path

```
torch_get_checkpoint_path(path, epoch)
```

Return the path of checkpoints during each epoch or whenever you want.

| Arguments |                                                                                                                                |
| --------- | ------------------------------------------------------------------------------------------------------------------------------ |
| path      | String (directory where you wnat to save checkpoints)                                                                          |
| epoch     | integer (number of epochs to save the model as checkpoint. An epoch is an iteration over the entire `x` and `y` data provided) |

Example:

```
mj.torch_get_checkpoint_path(path='/home/project/workspace/results', epoch=epoch)
```

##

#### torch\_save

```
torch_save(checkpoint, path, epoch)
```

Save checkpoint during each epoch.

| Arguments  |                                                                                                                             |
| ---------- | --------------------------------------------------------------------------------------------------------------------------- |
| checkpoint | Dictionary or OrderDictionary (model, weights or anything to save as checkpoint)                                            |
| path       | String (directory where you wnat to save checkpoints)                                                                       |
| epoch      | Integer (number of epochs to save the model as checkpoint. `epoch` would iterate over the entire `x` and `y` data provided) |

Example:

```
mj.torch_save(
    checkpoint = model.state_dict(),
    path = '/home/project/workspace/results',
    epoch = epoch)
```

##

#### tensor\_save

```
tensor_save(checkpoint, path)
```

Save checkpoint whenever you want.

| Arguments  |                                                       |
| ---------- | ----------------------------------------------------- |
| checkpoint | Sequential object (model to save as checkpoint)       |
| path       | String (directory where you wnat to save checkpoints) |

Example:

```
mj.tensor_save(checkpoint=model, path='/home/project/workspace/results')
```

##

#### callback

```
callback(metrics, path, frequency, on, filename=None)
```

Save checkpoint with the opportunity by `on` during each epoch and s

| Arguments |                                                                                                                                                                                                                                                            |
| --------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| metrics   | Metrics in dictionary to be evaluated by the model during training and testing. Each can be a string (name of a built-in function), function or a [`tf.keras.metrics.Metric`](https://www.tensorflow.org/api_docs/python/tf/keras/metrics/Metric) instance |
| path      | String (directory where you wnat to save checkpoints)                                                                                                                                                                                                      |
| frequency | Integer or List (if an integer `n`, checkpoints are saved every `n` times of each hook. if a list, it specifies the checkpoint frequencies for each hook individually.)                                                                                    |
| on        | Integer or List (when to trigger checkpoint creations. must be one of the Keras event hooks (less the `on_`), e.g. "train\_start", or "predict\_end". defaults to "epoch\_end".)                                                                           |
| filename  | String (name of checkpoint you want to save)                                                                                                                                                                                                               |

Example:

```
model.fit(
	X_train, Y_train,
	initial_epoch=1, epochs=epochs, batch_size=16, verbose=0,
	callbacks=[m.callback(
		metrics={"accuracy":"accuracy"},
		filename="modelcheckpoint01",
		path="/home/pj/ray_results",
		frequency=1,
		on="epoch_end")]
		)
```