Merge branch 'master' of github.com:mozilla/TTS

Dockerfile

@@ -0,0 +1,26 @@
+FROM nvidia/cuda:9.0-base-ubuntu16.04 as base
+
+WORKDIR /srv/app
+
+RUN apt-get update && \
+	apt-get install -y git software-properties-common wget vim build-essential libsndfile1 && \
+	add-apt-repository ppa:deadsnakes/ppa && \
+	apt-get update && \
+	apt-get install -y python3.6 python3.6-dev python3.6-tk && \
+	# Install pip manually
+	wget https://bootstrap.pypa.io/get-pip.py && \
+	python3.6 get-pip.py && \
+	rm get-pip.py && \
+	# Used by the server in server/synthesizer.py
+	pip install soundfile
+
+ADD . /srv/app
+
+# Setup for development
+RUN python3.6 setup.py develop
+
+# http://bugs.python.org/issue19846
+# > At the moment, setting "LANG=C" on a Linux system *fundamentally breaks Python 3*, and that's not OK.
+ENV LANG C.UTF-8
+
+CMD python3.6 server/server.py -c server/conf.json

README.md

@@ -23,6 +23,14 @@ Or you can use ```requirements.txt``` to install the requirements only.
 
 ```pip install -r requirements.txt```
 
+### Docker
+A barebone `Dockerfile` exists at the root of the project, which should let you quickly setup the environment. By default, it will start the server and let you query it. Make sure to use `nvidia-docker` to use your GPUs. Make sure you follow the instructions in the [`server README`](server/README.md) before you build your image so that the server can find the model within the image.
+
+```
+docker build -t mozilla-tts .
+nvidia-docker run -it --rm -p 5002:5002 mozilla-tts
+```
+
 ## Checkpoints and Audio Samples
 Checkout [here](https://mycroft.ai/blog/available-voices/#the-human-voice-is-the-most-perfect-instrument-of-all-arvo-part) to compare the samples (except the first) below.
 
@@ -62,6 +70,8 @@ TTS provides a generic dataloder easy to use for new datasets. You need to write
 - [Nancy](http://www.cstr.ed.ac.uk/projects/blizzard/2011/lessac_blizzard2011/)
 
 ## Training and Fine-tuning LJ-Speech
+[Click Here](https://gist.github.com/erogol/8f39174c3f0475221c8978aeb10d4fdc) for hands on **Notebook example**, training LJSpeech.
+
 Split ```metadata.csv``` into train and validation subsets respectively ```metadata_train.csv``` and ```metadata_val.csv```. Note that having a validation split does not work well as oppose to other ML problems since at the validation time model generates spectrogram slices without "Teacher-Forcing" and that leads misalignment between the ground-truth and the prediction. Therefore, validation loss does not really show the model performance. Rather, you might use the all data for training and check the model performance by relying on human inspection.
 
 ```

extract_features.py

@@ -82,13 +82,13 @@ if __name__ == "__main__":
             linear_len = linear.shape[1]
             np.save(linear_path, linear, allow_pickle=False)
             output.insert(2, linear_path+".npy")
+            assert mel_len == linear_len
         if args.process_audio:
             audio_file = file_name + "_audio"
             audio_path = os.path.join(CACHE_PATH, 'audio', audio_file)
             np.save(audio_path, x, allow_pickle=False)
             del output[0]
             output.insert(0, audio_path+".npy")
-        assert mel_len == linear_len
         return output
 
 

requirements.txt

@@ -10,3 +10,4 @@ Pillow
 flask
 scipy==0.19.0
 lws
+tqdm

train.py

@@ -478,7 +478,7 @@ if __name__ == '__main__':
         default=False,
         help='Do not verify commit integrity to run training.')
     parser.add_argument(
-        '--data_path', type=str, help='dataset path.', default='Defines the data path. It overwrites config.json.')
+        '--data_path', type=str, help='Defines the data path. It overwrites config.json.', default='')
     args = parser.parse_args()
 
     # setup output paths and read configs