Upload compress_to_h5.py with huggingface_hub

compress_to_h5.py

@@ -0,0 +1,64 @@
+import h5py
+import os
+import numpy as np
+
+# Directory containing your .npz files
+input_directory = "task_ABC_D/training"
+output_h5_file = "training_abc_d.h5"
+error_log_file = "error_log.txt"  # New text file to store error filenames  
+
+# Function to get list of already processed files
+def get_processed_files(h5_file):
+    processed_files = set()
+    if os.path.exists(h5_file):
+        with h5py.File(h5_file, "r") as h5f:
+            processed_files = set(h5f.keys())
+    return processed_files
+
+# Get list of already processed files
+processed_files = get_processed_files(output_h5_file)
+
+# Create an HDF5 file
+with h5py.File(output_h5_file, "a") as h5f:
+    # Loop through all files in the directory
+    for filename in os.listdir(input_directory):
+        if filename.endswith('.npz'):  # Only process .npz files
+            if filename in processed_files:
+                print("Skipping already processed file:", filename)
+                continue
+            filepath = os.path.join(input_directory, filename)
+            print(filename)
+            
+            # Load the .npz file
+            try:
+                npz_data = np.load(filepath)
+                # Create a group for this .npz file
+                file_group = h5f.create_group(filename)
+                
+                # Store each array from the .npz file as a separate dataset
+                for array_name in npz_data.files:
+                    # Create dataset with compression
+                    file_group.create_dataset(
+                        array_name, 
+                        data=npz_data[array_name], 
+                        compression="lzf", 
+                        # compression_opts=4
+                    )
+                
+                # Close the npz file
+                npz_data.close()
+            except Exception as e:
+                error_message = "Error processing " + filename
+                print(error_message)
+                # Append the filename to the error log file
+                with open(error_log_file, "a") as error_log:
+                    error_log.write(filename + "\n")
+
+print('done')
+
+# with h5py.File("combined_npz_files.h5", "r") as h5f:
+#     for group_name in h5f.keys():
+#         print(f"Group: {group_name}")
+#         group = h5f[group_name]
+#         for dataset_name in group.keys():
+#             print(f"  Dataset: {dataset_name}, Shape: {group[dataset_name].shape}, Dtype: {group[dataset_name].dtype}")