update for supporting variation of coco-text like datasets

Summary:
1. Add changes to support variation of datasets
2. Fix runner to support torchscript export

Reviewed By: wat3rBro

Differential Revision: D26871461

fbshipit-source-id: ec46f7e0d2c14c9b802aec22d78b2a089e962a2f

d2go/data/extended_coco.py

@@ -79,7 +79,7 @@ def convert_coco_text_to_coco_detection_json(
     For COCOText see: https://vision.cornell.edu/se3/coco-text-2/
     For COCODetection see: http://cocodataset.org/#overview
     """
-    with open(source_json) as f:
+    with open(source_json, "r") as f:
         coco_text_json = json.load(f)
 
     coco_text_json["annotations"] = list(coco_text_json["anns"].values())
@@ -115,6 +115,7 @@ def convert_coco_text_to_coco_detection_json(
         for x in coco_text_json["images"]:
             x["id"] = image_id_remap[x["id"]]
         for x in coco_text_json["annotations"]:
+            if x["image_id"] in image_id_remap:
                 x["image_id"] = image_id_remap[x["image_id"]]
 
     os.makedirs(os.path.dirname(target_json), exist_ok=True)