experiments/STT
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fixes for Dockerfile.{train,build} and adjust instructions for new image

Browse Source
This commit is contained in:
Reuben Morais 2021-03-29 21:05:49 +02:00
parent 1029d06a23
commit 214a150c19
6 changed files with 69 additions and 92 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.gitignore vendored
View File

@ -32,8 +32,6 @@
/doc/.build/ /doc/.build/
/doc/xml-c/ /doc/xml-c/
/doc/xml-java/ /doc/xml-java/
Dockerfile.build
Dockerfile.train
doc/xml-c doc/xml-c
doc/xml-java doc/xml-java
doc/xml-dotnet doc/xml-dotnet

4
Dockerfile.build.tmpl → Dockerfile.build
View File

@ -3,8 +3,8 @@
# Need devel version cause we need /usr/include/cudnn.h # Need devel version cause we need /usr/include/cudnn.h
FROM nvidia/cuda:10.1-cudnn7-devel-ubuntu18.04 FROM nvidia/cuda:10.1-cudnn7-devel-ubuntu18.04
ENV STT_REPO=#STT_REPO# ARG STT_REPO=https://github.com/coqui-ai/STT.git
ENV STT_SHA=#STT_SHA# ARG STT_SHA=origin/main
# >> START Install base software # >> START Install base software

63
Dockerfile.train Normal file
View File

@ -0,0 +1,63 @@
# Please refer to the TRAINING documentation, "Basic Dockerfile for training"
FROM tensorflow/tensorflow:1.15.4-gpu-py3
ENV DEBIAN_FRONTEND=noninteractive
# We need to purge python3-xdg because it's breaking STT install later with
# weird errors about setuptools
#
# libopus0 and libsndfile1 are dependencies for audio augmentation
#
RUN apt-get update && \
apt-get install -y --no-install-recommends \
build-essential \
cmake \
curl \
git \
libboost-all-dev \
libbz2-dev \
libopus0 \
libsndfile1 \
unzip \
wget && \
apt-get purge -y python3-xdg && \
rm -rf /var/lib/apt/lists/
# Make sure pip and its deps are up-to-date
RUN pip3 install --upgrade pip wheel setuptools
WORKDIR /code
# Tool to convert output graph for inference
RUN wget https://github.com/coqui-ai/STT/releases/download/v0.9.3/convert_graphdef_memmapped_format.linux.amd64.zip -O temp.zip && \
unzip temp.zip && rm temp.zip
COPY native_client /code/native_client
COPY .git /code/.git
COPY training/coqui_stt_training/VERSION /code/training/coqui_stt_training/VERSION
COPY training/coqui_stt_training/GRAPH_VERSION /code/training/coqui_stt_training/GRAPH_VERSION
# Build CTC decoder first, to avoid clashes on incompatible versions upgrades
RUN cd native_client/ctcdecode && make NUM_PROCESSES=$(nproc) bindings
RUN pip3 install --upgrade native_client/ctcdecode/dist/*.whl
# Install STT
# - No need for the decoder since we did it earlier
# - There is already correct TensorFlow GPU installed on the base image,
# we don't want to break that
COPY setup.py /code/setup.py
COPY VERSION /code/VERSION
COPY training /code/training
RUN DS_NODECODER=y DS_NOTENSORFLOW=y pip3 install --upgrade -e .
# Build KenLM to generate new scorers
COPY kenlm /code/kenlm
RUN cd /code/kenlm && \
mkdir -p build && \
cd build && \
cmake .. && \
make -j $(nproc)
# Copy rest of the code and test training
COPY . /code
RUN ./bin/run-ldc93s1.sh && rm -rf ~/.local/share/stt

68
Dockerfile.train.tmpl
View File

@ -1,68 +0,0 @@
# Please refer to the TRAINING documentation, "Basic Dockerfile for training"
FROM tensorflow/tensorflow:1.15.4-gpu-py3
ENV DEBIAN_FRONTEND=noninteractive
ENV STT_REPO=#STT_REPO#
ENV STT_SHA=#STT_SHA#
RUN apt-get update && apt-get install -y --no-install-recommends \
apt-utils \
bash-completion \
build-essential \
cmake \
curl \
git \
libboost-all-dev \
libbz2-dev \
locales \
python3-venv \
unzip \
wget
# We need to remove it because it's breaking STT install later with
# weird errors about setuptools
RUN apt-get purge -y python3-xdg
# Install dependencies for audio augmentation
RUN apt-get install -y --no-install-recommends libopus0 libsndfile1
# Try and free some space
RUN rm -rf /var/lib/apt/lists/*
WORKDIR /
RUN git clone $STT_REPO STT
WORKDIR /STT
RUN git checkout $STT_SHA
# Build CTC decoder first, to avoid clashes on incompatible versions upgrades
RUN cd native_client/ctcdecode && make NUM_PROCESSES=$(nproc) bindings
RUN pip3 install --upgrade native_client/ctcdecode/dist/*.whl
# Prepare deps
RUN pip3 install --upgrade pip==20.2.2 wheel==0.34.2 setuptools==49.6.0
# Install STT
# - No need for the decoder since we did it earlier
# - There is already correct TensorFlow GPU installed on the base image,
# we don't want to break that
RUN DS_NODECODER=y DS_NOTENSORFLOW=y pip3 install --upgrade -e .
# Tool to convert output graph for inference
RUN python3 util/taskcluster.py --source tensorflow --branch r1.15 \
--artifact convert_graphdef_memmapped_format --target .
# Build KenLM to generate new scorers
WORKDIR /STT/native_client
RUN rm -rf kenlm && \
git clone https://github.com/kpu/kenlm && \
cd kenlm && \
git checkout 87e85e66c99ceff1fab2500a7c60c01da7315eec && \
mkdir -p build && \
cd build && \
cmake .. && \
make -j $(nproc)
WORKDIR /STT
RUN ./bin/run-ldc93s1.sh

11
doc/DEPLOYMENT.rst
View File

@ -181,17 +181,12 @@ Dockerfile for building from source
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
We provide ``Dockerfile.build`` to automatically build ``libstt.so``, the C++ native client, Python bindings, and KenLM. We provide ``Dockerfile.build`` to automatically build ``libstt.so``, the C++ native client, Python bindings, and KenLM.
You need to generate the Dockerfile from the template using:
If you want to specify a different repository or branch, you can specify the ``STT_REPO`` or ``STT_SHA`` arguments:
.. code-block:: bash .. code-block:: bash
make Dockerfile.build docker build . -f Dockerfile.build --build-arg STT_REPO=git://your/fork --build-arg STT_SHA=origin/your-branch
If you want to specify a different repository or branch, you can pass ``STT_REPO`` or ``STT_SHA`` parameters:
.. code-block:: bash
make Dockerfile.build STT_REPO=git://your/fork STT_SHA=origin/your-branch
.. _runtime-deps: .. _runtime-deps:

13
doc/TRAINING.rst
View File

@ -88,18 +88,7 @@ Setting the ``TF_FORCE_GPU_ALLOW_GROWTH`` environment variable to ``true`` seems
Basic Dockerfile for training Basic Dockerfile for training
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
We provide ``Dockerfile.train`` to automatically set up a basic training environment in Docker. You need to generate the Dockerfile from the template using: We provide ``Dockerfile.train`` to automatically set up a basic training environment in Docker. This should ensure that you'll re-use the upstream Python 3 TensorFlow GPU-enabled Docker image. The image can be used with ``FROM ghcr.io/coqui-ai/stt-train``.
This should ensure that you'll re-use the upstream Python 3 TensorFlow GPU-enabled Docker image.
.. code-block:: bash
make Dockerfile.train
If you want to specify a different 🐸STT repository / branch, you can pass ``STT_REPO`` or ``STT_SHA`` parameters:
.. code-block:: bash
make Dockerfile.train STT_REPO=git://your/fork STT_SHA=origin/your-branch
Common Voice training data Common Voice training data
^^^^^^^^^^^^^^^^^^^^^^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^^^^