update README and command line help msg

examples/deepvoice3/README.md

@@ -112,6 +112,7 @@ tensorboard --logdir=runs/ --host=$HOSTNAME --port=8000
 usage: synthesize from a checkpoint [-h] --config CONFIG --input INPUT
                                     --output OUTPUT --checkpoint CHECKPOINT
                                     --monotonic_layers MONOTONIC_LAYERS
+                                    [--vocoder {griffin-lim,waveflow}]
 
 optional arguments:
   -h, --help            show this help message and exit
@@ -121,11 +122,14 @@ optional arguments:
   --checkpoint CHECKPOINT
                         data path of the checkpoint
   --monotonic_layers MONOTONIC_LAYERS
-                        monotonic decoder layer, index starts friom 1
+                        monotonic decoder layers' indices(start from 1)
+  --vocoder {griffin-lim,waveflow}
+                        vocoder to use
 ```
 
 `synthesize.py` is used to synthesize several sentences in a text file.
 `--monotonic_layers` is the index of the decoders layer that manifest monotonic diagonal attention. You can get monotonic layers by inspecting tensorboard logs. Mind that the index starts from 1. The layers that manifest monotonic diagonal attention are stable for a model during training and synthesizing, but differ among different runs. So once you get the indices of monotonic layers by inspecting tensorboard log, you can use them at synthesizing. Note that only decoder layers that show strong diagonal attention should be considerd.
+`--vocoder` is the vocoder to use. Current supported values are "waveflow" and "griffin-lim". Default value is "waveflow".
 
 example code:
 
@@ -135,5 +139,6 @@ CUDA_VISIBLE_DEVICES=2 python synthesize.py \
     --input sentences.txt \
     --output outputs/ \
     --checkpoint runs/Jul07_09-39-34_instance-mqcyj27y-4/step-1320000 \
-    --monotonic_layers "5,6"
+    --monotonic_layers "5,6" \
+    --vocoder waveflow
 ```

examples/deepvoice3/sentences.txt

@@ -1,5 +0,0 @@
-Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition
-in being comparatively modern.
-For although the Chinese took impressions from wood blocks engraved in relief for centuries before the woodcutters of the Netherlands, by a similar process
-produced the block books, which were the immediate predecessors of the true printed book,
-the invention of movable metal letters in the middle of the fifteenth century may justly be considered as the invention of the art of printing.

examples/deepvoice3/synthesize.py

@@ -86,7 +86,7 @@ if __name__ == "__main__":
     parser.add_argument("--input", type=str, required=True, help="text file to synthesize")
     parser.add_argument("--output", type=str, required=True, help="path to save audio")
     parser.add_argument("--checkpoint", type=str, required=True, help="data path of the checkpoint")
-    parser.add_argument("--monotonic_layers", type=str, required=True, help="monotonic decoder layer, index starts friom 1")
+    parser.add_argument("--monotonic_layers", type=str, required=True, help="monotonic decoder layers' indices(start from 1)")
     parser.add_argument("--vocoder", type=str, default="waveflow", choices=['griffin-lim', 'waveflow'], help="vocoder to use")
     args = parser.parse_args()
     with open(args.config, 'rt') as f: