Upload create_scale.py with huggingface_hub

create_scale.py

@@ -0,0 +1,43 @@
+
+import os
+import numpy as np
+import math
+import glob
+import fire 
+import shutil
+
+def main(scale_folder_name):
+    all_scale_dir = glob.glob(f'{scale_folder_name}/*')
+
+    for scale in all_scale_dir:
+        scale_percent = scale.split('/')[-1].split('_')[-1]
+        scale_percent = float(scale_percent)/100
+        print(scale_percent)
+        split_dir = glob.glob(scale + '/*')
+        for split in split_dir:
+            # modify file auto_lang_
+            auto_lang_ann_path = os.path.join(split, "lang_annotations/auto_lang_ann.npy")
+            auto_lang_ann = np.load(auto_lang_ann_path, allow_pickle=True).item()
+            
+            lang_ann = auto_lang_ann['language']['ann']
+            lang_task = auto_lang_ann['language']['ann']
+            trajectories = auto_lang_ann['info']['indx']
+            
+            length = len(lang_ann)
+            assert len(lang_ann) == len(lang_task) == len(trajectories) == length
+            
+            auto_lang_ann['language']['ann'] = lang_ann[:math.ceil(length*scale_percent)]
+            auto_lang_ann['language']['task'] = lang_task[:math.ceil(length*scale_percent)]
+            auto_lang_ann['info']['indx'] = trajectories[:math.ceil(length*scale_percent)]
+            
+            np.save(auto_lang_ann_path, auto_lang_ann)
+            
+            # dummy .npz file to bypass lookup_naming_pattern
+            shutil.copy(
+                src= "calvin_debug_dataset/training/episode_0358484.npz",
+                dst= os.path.join(split)
+            )
+            
+
+if __name__ == '__main__':
+  fire.Fire(main)