CPU-only environment

[dkurt]

When do import without CUDA:

import seqdist.sparse
from seqdist.ctc_simple import logZ_cupy

the following error appears.

original error: libcublas.so.10: cannot open shared object file: No such file or directory

To resolve it, we need a CPU/NumPY version of seqdist.sparse.logZ

[apredeus]

I'm trying to use @dkurt 's version on a CPU-only cluster and getting the same error. It can't be done yet, can it?

[dkurt]

@apredeus, for now #52 only works with v2 model (the latest is v3):

bonito basecaller dna_r9.4.1@v2 --use_openvino --device=cpu /data/reads > basecalls.fasta

I'll try to port logZ to NumPy in case of CPU device and update a pull request.

[apredeus]

Thank you! I'm getting the following error with v2

NameError: name 'IECore' is not defined

[dkurt]

@apredeus, please type export LD_LIBRARY_PATH=$(pwd)/venv3/lib:$LD_LIBRARY_PATH

[apredeus]

I did, but it didn't help.

[dkurt]

@apredeus, I guess that this is OS/Python version limitation. Can you please share your configuration?

[apredeus]

I've made a conda env with python 3.7 and virtual env within it. It's a 64-core server running Debian 8. Here's the output of pip list:

Package Version Location

---

alembic 1.4.3
attrs 20.3.0
cached-property 1.5.2
certifi 2020.11.8
chardet 3.0.4
cliff 3.5.0
cmd2 1.4.0
colorama 0.4.4
colorlog 4.6.2
crf-beam 0.0.1a0
cupy-cuda102 8.1.0
Cython 0.29.21
fast-ctc-decode 0.2.5
fastrlock 0.5
future 0.18.2
h5py 3.1.0
idna 2.8
importlib-metadata 2.0.0
joblib 0.17.0
Mako 1.1.3
mappy 2.17
MarkupSafe 1.1.1
numpy 1.16.3
ont-bonito 0.3.1 /pub37/alexp/bonito
ont-fast5-api 3.1.6
openvino-python 2021.1
optuna 1.1.0
packaging 20.4
parasail 1.2
pbr 5.5.1
pip 20.2.4
prettytable 0.7.2
progressbar33 2.4
pyparsing 2.4.7
pyperclip 1.8.1
python-dateutil 2.8.1
python-editor 1.0.4
PyYAML 5.3.1
requests 2.22.0
scipy 1.5.4
seqdist 0.0.1
setuptools 47.1.0
six 1.15.0
SQLAlchemy 1.3.20
stevedore 3.2.2
tbb 2020.3.254
toml 0.10.0
torch 1.5.0
tqdm 4.31.1
urllib3 1.25.11
wcwidth 0.2.5
wheel 0.35.1
zipp 3.4.0

echo $LD_LIBRARY_PATH

/pub37/alexp/bonito/venv3/lib:/pub37/alexp/miniconda3/lib:

[dkurt]

@apredeus, Are you able to use Docker? If so, I can recommend this approach due there is no binaries of OpenVINO for Debian 8.

1. Create a file with name "Dockerfile" and the following content:

Dockerfile

FROM ubuntu:18.04

USER root

RUN apt-get update && \
    apt-get install -y --no-install-recommends \
      python3-minimal \
      python3-pip \
      git \
      build-essential \
      python3-dev \
      libz-dev

RUN git clone -b openvino --depth 1 https://github.com/dkurt/bonito /bonito

RUN cd /bonito && \
    python3 -m pip install --upgrade pip && \
    python3 -m pip install setuptools && \
    python3 -m pip install -r requirements.txt && \
    python3 setup.py develop && \
    bonito download --models

RUN useradd -ms /bin/bash -G users bonito && \
    chown bonito -R /home/bonito

USER bonito

ENV LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH

2. Build image

docker build . -t bonito

3. Run

export INPUT_DIR=$(pwd)/reads
export OUTPUT_DIR=$(pwd)/output

mkdir -p ${OUTPUT_DIR}

docker run \
  -v "${INPUT_DIR}":"/reads" \
  -v "${OUTPUT_DIR}":"/output" \
  bonito:latest \
  /bin/sh -c "bonito basecaller dna_r9.4.1@v2 /reads --use_openvino --device=cpu > /output/basecalls.fasta"

[apredeus]

@dkurt thank you, but I can't use Docker on our cluster. Thank you for your help though, I really appreciate it. Perhaps I'll look for some GPU options in a little while.

[dkurt]

@apredeus, I'll let you know if there is a chance to make it work without Docker.

[dkurt]

I'll try to port logZ to NumPy in case of CPU device and update a pull request.

I've ported LogZ to NumPy so now v3 model also work on CPU with OpenVINO. See updated #52.

@apredeus, I'll let you know if there is a chance to make it work without Docker.

I can recommend to build OpenVINO with Python binding from source instead of using pip distro:

git clone --depth 1 https://github.com/openvinotoolkit/openvino/
cd openvino
git submodule update --init --recursive
mkdir build && cd build

cmake \
  -DCMAKE_BUILD_TYPE=Release \
  -DENABLE_CLDNN=OFF \
  -DENABLE_VPU=OFF \
  -DENABLE_TESTS=OFF \
  -DENABLE_PYTHON=ON \
  -DENABLE_OPENCV=OFF \
  -DENABLE_SAMPLES=OFF \
  -DPYTHON_EXECUTABLE=`which python3` \
  -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.6m.so \
  -DPYTHON_INCLUDE_DIR=/usr/include/python3.6 ..

make -j$(nproc)

(not sure which Python version you have so replace 3.6 to appropriate)

then

export LD_LIBRARY_PATH=/path/to/openvino/bin/intel64/Release/lib/:$LD_LIBRARY_PATH
export PYTHONPATH=/path/to/openvino/bin/intel64/Release/lib/python_api/python3.6/:$PYTHONPATH

[colindaven]

Thanks @dkurt ! That was really helpful.

I followed the docker recipe and bonito is now running.

Changes for others:

You might need to add your proxy to ENV if behind a firewall.

# Add my proxy
ENV http_proxy 'http://YOURPROXYIP:8080'
ENV https_proxy 'http://YOURPROXYIP:8080'

# Create user bonito, same id gid as YOUR_EXTERNAL_NFS_USER (here 9415). To solve "permission denied" problem
RUN groupadd -g 9415 bonito
RUN useradd -m -u 9415 -g 9415 -s /bin/bash bonito
RUN chown bonito -R /home/bonito