From e2597bdde354cbe5a8414f8532e7d3baa83dffaa Mon Sep 17 00:00:00 2001
From: Brandon Thomas <echelon@gmail.com>
Date: Tue, 4 Apr 2023 02:20:40 -0700
Subject: [PATCH] Dockerfile for VITS

---
 tts/vits/Dockerfile | 69 +++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 69 insertions(+)
 create mode 100644 tts/vits/Dockerfile

diff --git a/tts/vits/Dockerfile b/tts/vits/Dockerfile
new file mode 100644
index 0000000..c5e0089
--- /dev/null
+++ b/tts/vits/Dockerfile
@@ -0,0 +1,69 @@
+# ==================== BASE CONTAINER + REQUIREMENTS ====================
+
+FROM nvidia/cuda:12.1.0-devel-ubuntu22.04
+
+# See: https://github.com/opencontainers/image-spec/blob/master/annotations.md
+LABEL org.opencontainers.image.title='Storyteller ML : VITS'
+LABEL org.opencontainers.image.description='Storyteller ML : VITS'
+LABEL org.opencontainers.image.documentation='https://github.com/storytold/storyteller-ml'
+LABEL org.opencontainers.image.source='https://github.com/storytold/storyteller-ml'
+LABEL org.opencontainers.image.url='https://github.com/storytold/storyteller-ml'
+
+# tzdata install hangs without presetting the time zone
+ENV TZ=UTC
+RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone
+
+# Notes on packages:
+#  * espeak-ng is critical for speech
+#  * screen: not sure why included; I threw in tmux and curl for good measure
+#  * unzip: not sure why included, but perhaps downloading models
+RUN DEBIAN_FRONTEND=noninteractive apt-get update \
+  && apt-get install -y --no-install-recommends \
+        build-essential \
+        curl \
+        espeak-ng \
+        python3-dev \
+        python3-pip \
+        python3.10 \
+        python3.10-venv \
+        screen \
+        tmux \
+        unzip
+
+RUN python3 --version
+
+# ==================== INSTALL MORE VITS REQUIREMENTS ====================
+
+WORKDIR /
+RUN mkdir -p /vits
+COPY ./requirements.txt /vits
+
+WORKDIR /vits
+
+RUN pip3 install --upgrade pip setuptools \
+  && pip3 install -r requirements.txt \
+  && pip3 install \
+     transformers \
+     g2p-en \
+     scipy \
+     text_unidecode \
+     emoji \
+     phonemizer \
+     numpy \
+     matplotlib \
+     librosa==0.9.2 \
+     Cython \
+     Unidecode \
+     tensorboard \
+     gdown
+
+# ==================== VITS CODE + FINAL BUILD ====================
+
+COPY ./ ./
+
+RUN cd monotonic_align \
+  && mkdir monotonic_align \
+  && python3 setup.py build_ext --inplace \
+  && cd ..
+
+WORKDIR /