add gen ck

configs/_base_/datasets/coco_detection.py

 dataset_type = 'CocoDataset'
-data_root = 'data/coco/'
+data_root = '/opt/Dataset/COCO2017/'
 img_norm_cfg = dict(
     mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
 train_pipeline = [

configs/_base_/datasets/coco_instance.py

 dataset_type = 'CocoDataset'
-data_root = 'data/coco/'
+data_root = '/opt/Dataset/COCO2017/'
 img_norm_cfg = dict(
     mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
 train_pipeline = [

configs/_base_/datasets/coco_instance_semantic.py

 dataset_type = 'CocoDataset'
-data_root = 'data/coco/'
+data_root = '/opt/Dataset/COCO2017/'
 img_norm_cfg = dict(
     mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
 train_pipeline = [

tools/gen_checkpoint.py

+import argparse
+import os
+import warnings
+import torch
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(
+        description='generate model')
+    parser.add_argument('--backbone', help='the backbone checkpoint file')
+    parser.add_argument('--head', help='the head checkpoint file')
+    parser.add_argument('--out', help='output result file in pickle format')
+    args = parser.parse_args()
+    return args
+
+def get_sd(filename, return_sd=True):
+    print("loading {}".format(filename))
+    ck = torch.load(filename)
+    if return_sd:
+        return ck['state_dict']
+    else:
+        return ck
+
+def merge(target, backbone, head):
+    tsd = target['state_dict']
+    bsd = target['state_dict']
+    hsd = target['state_dict']
+    for key in tsd.keys():
+        if 'backbone' in key:
+            assert key in bsd
+            tsd[key] = bsd[key]
+        else:
+            assert key in hsd
+            tsd[key] = hsd[key]
+    return target
+
+def main():
+    args = parse_args()
+    print("generate checkpoint")
+
+    backbone = get_sd(args.backbone, return_sd=False)
+    head = get_sd(args.head, return_sd=False)
+
+    target = backbone.copy()
+    #target = head.copy()
+
+    target = merge(target, backbone, head)
+    os.makedirs(os.path.basename(args.out), exist_ok=True)
+    torch.save(target, args.out)
+    print("saved checkpoint in {}".format(args.out))
+    
+
+
+if __name__ == '__main__':
+    main()