Skip to content

SOTA Open Source TTS

Update docker container

Lengyue3ec16024dd8d
diff --git a/.dockerignore b/.dockerignore
index 78eb259..efcfce8 100644
--- a/.dockerignore
+++ b/.dockerignore
@@ -2,3 +2,4 @@ results
data
*.filelist
/data_server/target
+checkpoints
diff --git a/docker-compose.dev.yml b/docker-compose.dev.yml
index 08ddd7e..3054037 100644
--- a/docker-compose.dev.yml
+++ b/docker-compose.dev.yml
@@ -5,8 +5,7 @@ services:
build: .
container_name: fish-speech
volumes:
- - ./data:/exp/data
- - ./raw_data:/exp/raw_data
+ - ./:/exp
deploy:
resources:
reservations:
diff --git a/dockerfile b/dockerfile
index 7e6566f..25310ab 100644
--- a/dockerfile
+++ b/dockerfile
@@ -1,39 +1,42 @@
-FROM nvcr.io/nvidia/pytorch:23.09-py3
+FROM nvcr.io/nvidia/pytorch:24.02-py3
# Install system dependencies
ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get update && apt-get install -y git curl build-essential ffmpeg libsm6 libxext6 libjpeg-dev \
- zlib1g-dev aria2 zsh openssh-server sudo python3.10-venv && \
+ zlib1g-dev aria2 zsh openssh-server sudo python3.10-venv protobuf-compiler && \
apt-get clean && rm -rf /var/lib/apt/lists/*
-# Install s5cmd
-RUN curl -L https://github.com/peak/s5cmd/releases/download/v2.2.0/s5cmd_2.2.0_Linux-64bit.tar.gz | tar xvz -C /tmp && \
- mv /tmp/s5cmd /usr/local/bin/s5cmd && s5cmd --help
-
-# Install code server and zsh
-RUN wget -c https://github.com/coder/code-server/releases/download/v4.5.1/code-server_4.5.1_amd64.deb && \
- dpkg -i ./code-server_4.5.1_amd64.deb && \
- code-server --install-extension ms-python.python && \
- rm ./code-server_4.5.1_amd64.deb && \
- sh -c "$(curl https://raw.githubusercontent.com/robbyrussell/oh-my-zsh/master/tools/install.sh)" "" --unattended
+# Install zsh
+RUN sh -c "$(curl https://raw.githubusercontent.com/robbyrussell/oh-my-zsh/master/tools/install.sh)" "" --unattended
# Set zsh as default shell
RUN chsh -s /usr/bin/zsh
ENV SHELL=/usr/bin/zsh
+# Setup torchaudio
+RUN git clone https://github.com/pytorch/audio --recursive --depth 1 && \
+ cd audio && pip install -v --no-use-pep517 . && \
+ cd .. && rm -rf audio && python -c "import torchaudio; print(torchaudio.__version__)"
+
# Setup flash-attn
RUN pip3 install --upgrade pip && \
pip3 install ninja packaging && \
- pip3 install git+https://github.com/Dao-AILab/flash-attention.git
+ FLASH_ATTENTION_FORCE_BUILD=TRUE pip3 install git+https://github.com/Dao-AILab/flash-attention.git
+
+# Test flash-attn
+RUN python3 -c "from flash_attn import flash_attn_varlen_func"
# Project Env
WORKDIR /exp
-COPY . .
+COPY pyproject.toml ./
+COPY data_server ./data_server
+COPY fish_speech ./fish_speech
# Setup rust-data-server
RUN curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y && \
- cd data_server && $HOME/.cargo/bin/cargo build --release
+ cd data_server && $HOME/.cargo/bin/cargo build --release && cp target/release/data_server /usr/local/bin/ && \
+ cd .. && rm -rf data_server && data_server --help
-RUN pip3 install -e .
+RUN pip3 install -e . && pip uninstall -y fish-speech && rm -rf fish_speech
CMD /bin/zsh
diff --git a/pyproject.toml b/pyproject.toml
index 8cd9731..46cf2be 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -36,7 +36,8 @@ dependencies = [
"nnAudio>=0.3.2",
"loralib>=0.1.2",
"natsort>=8.4.0",
- "cn2an>=0.5.22"
+ "cn2an>=0.5.22",
+ "pyrootutils>=1.0.4"
]
[project.optional-dependencies]