This article will tell you how to use deepvariant from nvidia driver.
Sponsor:
We are using the D52BV provided by Quanta Computer.
Environment inside:
- 80 vCPU
- 4 T4 GPU
- 512 GiB RAM
Environmental configuration
- Ubuntu 18.04 LTS
- Nvidia driver 410.129
- CUDA 10.0
- cudnn 7.6.3
- Bazel 0.19.2
- Tensorflow 1.13
- DeepVariant 0.8.0
- Python2.7
- Java 1.8
Install Nvidia driver
Before you install nvidia driver. You need go to this page download program.
sudo chmod u+x NVIDIA-Linux-x86_64-410.129-diagnostic.run
sudo ./NVIDIA-Linux-x86_64-410.129-diagnostic.run
If you install fail or you lack some tools, this command is uninstall driver.
sudo ./NVIDIA-Linux-x86_64-410.129-diagnostic.run --uninstall
Then, follow below command.
nvidia-smi
You can see this figure.
Congratulations, you did it.
By the way. I delete GPU UUID, so you can’t see it.
Install CUDA 10.0
First you need go to this page download CUDA 10.0
And follow below command.
sudo dpkg -i cuda-repo-ubuntu1804-10-0-local-10.0.130-410.48_1.0-1_amd64.debsudo apt-key add /var/cuda-repo-10-0-local-10.0.130-410.48/7fa2af80.pubsudo apt updatesudo apt install cuda-10.0
Add environment variables
vim ~/.bashrc
export PATH=$PATH:/usr/local/cuda/bin/;
source ~/.bashrc
Then, input this command you can see CUDA version.
This means that CUDA was successfully installed.
nvcc -V
Install Cudnn 7.6.3
You need go to this page download three file.
- libcudnn7_7.6.3.30–1+cuda10.0_amd64.deb
- libcudnn7-dev_7.6.3.30–1+cuda10.0_amd64.deb
- libcudnn7-doc_7.6.3.30–1+cuda10.0_amd64.deb
sudo dpkg -i libcudnn7_7.6.3.30-1+cuda10.0_amd64.deb
sudo dpkg -i libcudnn7-dev_7.6.3.30-1+cuda10.0_amd64.deb
sudo dpkg -i libcudnn7-doc_7.6.3.30-1+cuda10.0_amd64.deb
Install Bazel 0.19.2
Install required packages
sudo apt-get install pkg-config zip g++ zlib1g-dev unzip python
Download bazel 0.19.2
https://github.com/bazelbuild/bazel/releases/tag/0.19.2bazel-0.19.2-installer-linux-x86_64.shchmod +x bazel-0.19.2-installer-linux-x86_64.sh4./bazel-0.19.2-installer-linux-x86_64.sh --uservim ~/.bashrcexport PATH=”$PATH:$HOME/bin”source ~/.bashrc
Install Tensorflow GPU 1.13
First you need use Python2
If you are use Python3 use can change python version.
ls /usr/local/python*Python --versionalias python='/usr/bin/python2.7'
Install tensorflow-gpu 1.13.1
sudo apt-get install python3-distutilspip install tensorflow-gpu==1.13.1pip install setuptoolspip show tensorflow-gpu
Download DeepVariant
You need to go to deepvariant github page.
Download the deepvariant version you want.
Github. We use deepvariant 0.8.0
wget https://github.com/google/deepvariant/releases/download/v0.8.0/deepvariant.zipunzip deepvariant.zip
Deepvariant have three steps:
- Make examples
- Call Variants
- Postprocess Variant
Make example
You need go to this path:
You can find make example.
deepvariant/binaries/DeepVariant/0.8.0/DeepVariant-0.8.0/make_examples.zip
How to use it ?
python2 deepvariant/binaries/DeepVariant/0.8.0/DeepVariant-0.8.0/make_examples.zip \
--mode calling \
--ref {fasta file} \
--reads "{bam file}" \
--examples "{output_make_example}"
Call Variants
Call variants path:
deepvariant/binaries/DeepVariant/0.8.0/DeepVariant-0.8.0/call_variants.zip
command:
Postprocess Variant
Postprocess Variant path:
deepvariant/binaries/DeepVariant/0.8.0/DeepVariant-0.8.0/postprocess_variants.zip
command:
python2 /deepvariant/binaries/DeepVariant/0.8.0/DeepVariant-0.8.0/postprocess_variants.zip \
--ref "{fasta file}" \
--infile "{call_variant_file}" \
--outfile "{Postprocess_Variant_file}"
Finish
Please kindly review it and any question is welcome
Thanks,
Jacky
One More thing
If you feel that you don’t understand how to deploy environment ?
You can use this approach.
This method could be use.
However I didn’t try it. Maybe you can try.
git clone https://github.com/google/deepvariant.git
cd deepvariant#This will take a bit longer.
./build-prereq.sh
Because very tools always using Java 1.8.
However, if you use Ubuntu 18.04 defaults version is 1.11 so you need change version.
update-alternatives --config java
java -version