Good day,
We managed to submit the docker image (Submission ID 9702596) but we did not get a score.
We suspect that it is something to do with the Tensorflow version or CUDA.
Below is a snippet of our Dockerfile :
```
FROM ubuntu
ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get update && apt-get install -y \
python3 \
python3-pip \
libsm6 libxext6 libxrender-dev
COPY requirements.txt /requirements.txt
RUN pip3 install --no-cache-dir -r requirements.txt
```
requirements.txt:
```
opencv-python
numpy
pandas
glob2
tensorflow-gpu
keras
```
run.sh:
```
#!/bin/bash
export CUDA_HOME=/cm/local/apps/cuda/libs/current
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:${CUDA_HOME}/lib64
export PATH=${CUDA_HOME}/bin:${PATH}
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/cm/shared/apps/cuda10.0/toolkit/10.0.130/lib64
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/share/apps/rc/software/cuDNN/7.6.2.24-CUDA-10.1.243/lib64
##Run your model:
python3 /RA2_script.py
```
Please advise if there is anything done wrongly in the codes.
Thank you!
Best Regards,
Raphael and @yanmingtan
Created by raphael_quek I've modified the instructions on the Docker submission page in the wiki for this challenge to help clarify the difference between submitting to these two queues.
Have a great weekend,
Robert
Hi @raphael_quek and @yanmingtan,
The submission ID you provided (9702596) is on the Fast Lane - the fast lane queue has a subset of the train and test data to help you test your container on our infrastructure and ensure that it produces a valid prediction file that can be scored. It does not provide a score, and it does not use up your submission quota.
Now that you know it is valid, you'll want to submit your to the main challenge queue (RA2 Challenge Submission, queue number 9614346) to run on the full leaderboard dataset.
Let me know if you have any questions about this!
Thanks,
Robert
Hey - my suggestion would be to use conda instead and install tensorflow via that
You can start from a miniconda base package & then just install it directly:
```
FROM continuumio/miniconda3@sha256:6c979670684d970f8ba934bf9b7bf42e77c30a22eb96af1f30a039b484719159
RUN /opt/conda/bin/conda install tensorflow-gpu
```
Note: You can install install packages via pip after that
Otherwise you can maybe check the error logs - when you submitted your image you should have received an email that your image was submitted. It also contains a link to some log files