Skip to content

Commit

Permalink
merge with master
Browse files Browse the repository at this point in the history
Signed-off-by: Jason <[email protected]>
  • Loading branch information
blisc committed May 6, 2020
2 parents 579be8e + 8025d3d commit cffee88
Show file tree
Hide file tree
Showing 33 changed files with 1,466 additions and 236 deletions.
1 change: 1 addition & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -73,6 +73,7 @@ To release a new version, please update the changelog as followed:
- Added multi-dataset data-layer and dataset.
([PR #538](https://github.com/NVIDIA/NeMo/pull/538)) - @yzhang123
- Online Data Augmentation for ASR Collection. ([PR #565](https://github.com/NVIDIA/NeMo/pull/565)) - @titu1994
- Speed augmentation on CPU, TimeStretch augmentation on CPU+GPU ([PR #594](https://github.com/NVIDIA/NeMo/pull/565)) - @titu1994

### Changed

Expand Down
24 changes: 15 additions & 9 deletions Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -30,14 +30,23 @@ RUN apt-get update && \
python-dev && \
rm -rf /var/lib/apt/lists/*

# build trt open source plugins
# install trt
ENV PATH=$PATH:/usr/src/tensorrt/bin
WORKDIR /tmp/trt-oss
RUN git clone --recursive --branch release/7.0 https://github.com/NVIDIA/TensorRT.git && cd TensorRT && \
mkdir build && cd build && \
cmake .. -DCMAKE_BUILD_TYPE=Release -DTRT_LIB_DIR=/usr/lib/x86_64-linux-gnu/ -DTRT_BIN_DIR=`pwd` \
-DBUILD_PARSERS=OFF -DBUILD_SAMPLES=OFF -DBUILD_PLUGINS=ON -DGPU_ARCHS="70 75" && \
make -j nvinfer_plugin
ARG NV_REPO=https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64

RUN cd /tmp/trt-oss
ARG DEB=libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
RUN curl -sL --output ${DEB} ${NV_REPO}/${DEB}
ARG DEB=libnvinfer7_7.0.0-1+cuda10.2_amd64.deb
RUN curl -sL --output ${DEB} ${NV_REPO}/${DEB}
ARG DEB=libnvinfer-plugin7_7.0.0-1+cuda10.2_amd64.deb
RUN curl -sL --output ${DEB} ${NV_REPO}/${DEB}
ARG DEB=libnvonnxparsers7_7.0.0-1+cuda10.2_amd64.deb
RUN curl -sL --output ${DEB} ${NV_REPO}/${DEB}
ARG DEB=python-libnvinfer_7.0.0-1+cuda10.2_amd64.deb
RUN curl -sL --output ${DEB} ${NV_REPO}/${DEB}
RUN dpkg -i *.deb && cd ../.. && rm -rf /tmp/trt-oss

# install nemo dependencies
WORKDIR /tmp/nemo
Expand All @@ -53,9 +62,6 @@ FROM nemo-deps as nemo
ARG NEMO_VERSION
ARG BASE_IMAGE

# copy oss trt plugins
COPY --from=nemo-deps /tmp/trt-oss/TensorRT/build/libnvinfer_plugin.so* /usr/lib/x86_64-linux-gnu/

# Check that NEMO_VERSION is set. Build will fail without this. Expose NEMO and base container
# version information as runtime environment variable for introspection purposes
RUN /usr/bin/test -n "$NEMO_VERSION" && \
Expand Down
30 changes: 30 additions & 0 deletions docs/sources/source/asr/installation.rst
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
Installation
============

Neural Modules and their corresponding collections have certain requirements that can be optionally installed to
improve performance of operations.

Torch Audio
-----------

The `torchaudio` library is used for certain audio pre-processing Neural Modules. Primarily,

- AudioToMFCCPreprocessor
- TimeStretchAugmentation

Official installation directions are provided at the `torchaudio github page <https://github.com/pytorch/audio>`_. It is recommended to follow
the conda installation procedure and install the latest version of the library available on conda.

Numba
-----

The `numba` library is used for optimized execution of certain data augmentation procedures that can be used during
data pre-processing. It can substantially reduce execution time during training, and is a recommended installation for
Neural Modules.

Official installation directions are provided at the `numba github page <https://github.com/numba/numba>`_. It is recommended to follow
the conda installation procedure and install the latest version of the library available on conda.

.. code-block:: bash
conda install numba
1 change: 1 addition & 0 deletions docs/sources/source/asr/intro.rst
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ Speech Recognition
.. toctree::
:maxdepth: 8

installation
tutorial
datasets
models
Expand Down
1 change: 1 addition & 0 deletions docs/sources/source/speech_command/installation_link.rst
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
.. include:: ../asr/installation.rst
1 change: 1 addition & 0 deletions docs/sources/source/speech_command/intro.rst
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,7 @@ Speech Commands
.. toctree::
:maxdepth: 8

installation_link
tutorial
datasets
models
Expand Down
Loading

0 comments on commit cffee88

Please sign in to comment.