Using ijson to avoid loading full json in memory.

Using ijson to load item by item, so it is possible to load dataset
using dataset/convert_cc_sbu.py and dataset/convert_laion.py on machines
with low RAM.

dataset/convert_cc_sbu.py

@@ -1,20 +1,22 @@
-import json
-import csv
+import ijson
 
 # specify input and output file paths
 input_file = 'ccs_synthetic_filtered_large.json'
 output_file = 'ccs_synthetic_filtered_large.tsv'
 
-# load JSON data from input file
-with open(input_file, 'r') as f:
-    data = json.load(f)
+# set header to None
+headers = None
 
-# extract header and data from JSON
-header = data[0].keys()
-rows = [x.values() for x in data]
-
-# write data to TSV file
-with open(output_file, 'w') as f:
-    writer = csv.writer(f, delimiter='\t')
-    writer.writerow(header)
-    writer.writerows(rows)
+# load JSON data from input file and open the output file at same time
+with open(input_file, 'r') as in_file, open(output_file, 'w') as out_file:
+    objects = ijson.items(in_file, 'item')
+    
+    for obj in objects:
+        # extract header and data from JSON
+        if headers is None:
+            headers = list(obj.keys())
+            out_file.write('\t'.join(headers) + '\n')
+        
+        # write data to TSV file line by line
+        row = '\t'.join(str(obj[key]) for key in headers)
+        out_file.write(row + '\n')

dataset/convert_laion.py

@@ -1,20 +1,22 @@
-import json
-import csv
+import ijson
 
 # specify input and output file paths
 input_file = 'laion_synthetic_filtered_large.json'
 output_file = 'laion_synthetic_filtered_large.tsv'
 
-# load JSON data from input file
-with open(input_file, 'r') as f:
-    data = json.load(f)
+# set header to None
+headers = None
 
-# extract header and data from JSON
-header = data[0].keys()
-rows = [x.values() for x in data]
-
-# write data to TSV file
-with open(output_file, 'w') as f:
-    writer = csv.writer(f, delimiter='\t')
-    writer.writerow(header)
-    writer.writerows(rows)
+# load JSON data from input file and open the output file at same time
+with open(input_file, 'r') as in_file, open(output_file, 'w') as out_file:
+    objects = ijson.items(in_file, 'item')
+    
+    for obj in objects:
+        # extract header and data from JSON
+        if headers is None:
+            headers = list(obj.keys())
+            out_file.write('\t'.join(headers) + '\n')
+        
+        # write data to TSV file line by line
+        row = '\t'.join(str(obj[key]) for key in headers)
+        out_file.write(row + '\n')