fixed new dataset

README.md

@@ -16,7 +16,7 @@ git clone https://huggingface.co/datasets/DolphinNie/dungeon-dataset
 
 ## 1. Data Explanation
 
-This is the Map dataset from the open-sourced game [Brogue](https://github.com/tmewett/BrogueCE). It contains 40,000 train dataset, 10,000 test dataset and 10,000 validation dataset.
+This is the Map dataset from the open-sourced game [Brogue](https://github.com/tmewett/BrogueCE). It contains 49,000 train dataset, 14,000 test dataset and 7,000 validation dataset.
 
 Each map is stored in a `.csv` file. The map is a `(32x32)` array, which is the map size.
 
@@ -87,8 +87,17 @@ import matplotlib.pyplot as plt
 # Load dataset from hugging face
 dataset = load_dataset("DolphinNie/dungeon-dataset")
 
-# Dataset stored by huggingface is not in a correct format, we need to do further process
-def dataset_convert(dataset):
+
+def get_processed_dataset(load_dataset_from_pickle=False,
+                          save_dataset_to_pickle=False,
+                          pickle_save_path='dungeon-dataset.pkl'):
+    dataset = pull_hugging_face_dataset(load_dataset_from_pickle,
+                                        save_dataset_to_pickle,
+                                        pickle_save_path)
+    dataset_train, dataset_test, dataset_valid = convert_dataset(dataset)
+    return dataset_train, dataset_test, dataset_valid
+
+def convert_dataset(dataset):
     dataset_train = list()
     dataset_test = list()
     dataset_valid = list()
@@ -98,11 +107,11 @@ def dataset_convert(dataset):
         datapoint_num = int(dataset[name[i]].num_rows / 32)
         dataset_tf = dataset[name[i]].to_pandas()
         for n in range(datapoint_num):
-            env_num = tf_train[n*32:(n+1)*32]
+            env_num = dataset_tf[n * 32:(n + 1) * 32]
             datasets[i].append(env_num)
     return dataset_train, dataset_test, dataset_valid
 
-dataset_train, dataset_test, dataset_valid = dataset_convert(dataset)
+dataset_train, dataset_test, dataset_valid = get_processed_dataset(load_dataset_from_pickle, save_dataset_to_pickle)
 
 # Visualize the datapoints if you want
 def visualize_map(dungeon_map):

env_num_test.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dcb44c2b4168f12e5d514b0df8b5352f320c82da0372ea0f9eea7e210bbe92e7
-size 4256874
+oid sha256:96a40d89fee5643190fb4c0279fcb2ddf8163c2f7f48e91c78740790f979f8e8
+size 5966510

env_num_train.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c1bf4e80a23630fc026ff6a20f1b1e0fef6b4e107da8848d33093d311d2f28da
-size 17194333
+oid sha256:b70df308db028325a2c98de635bc88e3acd1d9956f9a73c9dbcdc7f1f4e2cb96
+size 21061633

env_num_valid.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6f3293a3269fc6ce99c509654b48037a7d529b749a1c45022bbfa2f50ae04629
-size 4303217
+oid sha256:7301cd6db76f25ea986244771d57d8c63f23ac94149c438c5954593905495ff2
+size 3011923