add lr and optimizer params

scripts/cnocr_train.py

@@ -80,7 +80,13 @@ def parse_args():
         default=2,
     )
     parser.add_argument(
-        "--gpu", help="Number of GPUs for training [Default 0]", type=int
+        "--gpu", help="Number of GPUs for training [Default 0]", type=int, default=0
+    )
+    parser.add_argument(
+        "--optimizer",
+        help="optimizer for training [Default: Adam]",
+        type=str,
+        default='Adam',
     )
     parser.add_argument(
         '--epoch', type=int, default=20, help='train epochs [Default: 20]'
@@ -90,6 +96,12 @@ def parse_args():
         type=int,
         help='load the model on an epoch using the model-load-prefix [Default: no trained model will be loaded]',
     )
+    parser.add_argument(
+        '--lr',
+        type=float,
+        default=None,
+        help='learning rate [Default: None, means lr from hp will be used]',
+    )
     parser.add_argument(
         "--prefix",
         help="Checkpoint prefix [Default '{}']".format(default_model_prefix),
@@ -98,36 +110,16 @@ def parse_args():
     parser.add_argument(
         "--loss", help="'ctc' or 'warpctc' loss [Default 'ctc']", default='ctc'
     )
-    parser.add_argument(
-        "--num_proc",
-        help="Number CAPTCHA generating processes [Default 4]",
-        type=int,
-        default=4,
-    )
-    parser.add_argument(
-        "--font_path", help="Path to ttf font file or directory containing ttf files"
-    )
     return parser.parse_args()
 
 
-def get_fonts(path):
-    fonts = list()
-    if os.path.isdir(path):
-        for filename in os.listdir(path):
-            if filename.endswith('.ttf') or filename.endswith('.ttc'):
-                fonts.append(os.path.join(path, filename))
-    else:
-        fonts.append(path)
-    return fonts
-
-
-def run_cn_ocr(args):
+def train_cnocr(args):
     head = '%(asctime)-15s %(message)s'
     logging.basicConfig(level=logging.DEBUG, format=head)
     args.prefix = '{}-{}'.format(args.prefix, args.model_name)
 
     hp = CnHyperparams()
-    hp._num_epoch = args.epoch
+    hp = _update_hp(hp, args)
 
     network, hp = gen_network(args.model_name, hp)
     metrics = CtcMetrics(hp.seq_length)
@@ -147,6 +139,14 @@ def run_cn_ocr(args):
     )
 
 
+def _update_hp(hp, args):
+    hp._num_epoch = args.epoch
+    hp.optimizer = args.optimizer
+    if args.lr is not None:
+        hp._learning_rate = args.lr
+    return hp
+
+
 def _gen_iters(hp, train_fp_prefix, val_fp_prefix, use_train_image_aug):
     height, width = hp.img_height, hp.img_width
     augs = None
@@ -192,4 +192,4 @@ def _gen_iters(hp, train_fp_prefix, val_fp_prefix, use_train_image_aug):
 
 if __name__ == '__main__':
     args = parse_args()
-    run_cn_ocr(args)
+    train_cnocr(args)