reformat readme

Author: EloiZ

README.md

@@ -32,15 +32,15 @@ OCTET/checkpoints/encoder_256x512.pt
 
 To use OCTET, you will need to instantiate a `Config` object with the desired parameters. The `Config` object should contain the following parameters:
 
-`blobgan_weights`: path to the weights file for the blobGan model
-`encoder_weights`: path to the weights file for the image encoder
-`decision_model_weights`: path to the weights file for the decision model to be explained
-`device`: torch device to use for computation
-`output_dir`: directory where generated images will be saved
-`real_images`: a boolean indicating whether real images should be used (True) or generated images should be used (False)
-`dataset_path`: path to the directory containing the images to use (if `real_images` is True)
-`bs`: batch size to use when loading images from the dataset
-`num_imgs`: number of images to generate (if `real_images` is False)
+* `blobgan_weights`: path to the weights file for the blobGan model
+* `encoder_weights`: path to the weights file for the image encoder
+* `decision_model_weights`: path to the weights file for the decision model to be explained
+* `device`: torch device to use for computation
+* `output_dir`: directory where generated images will be saved
+* `real_images`: a boolean indicating whether real images should be used (True) or generated images should be used (False)
+* `dataset_path`: path to the directory containing the images to use (if `real_images` is True)
+* `bs`: batch size to use when loading images from the dataset
+* `num_imgs`: number of images to generate (if `real_images` is False)
 
 Once you have created a `Config` object, you can instantiate an OCTET object with `octet = OCTET(config)`. You can then call the following methods on the OCTET object: `octet.invert_and_cf()` that will invert the image to get the latent code and generate the counterfactual explanation.