Problems while building docker

[IceCapriccio]

When I ran sudo docker build . -t hmnet，errors occurred in Step 10/35 : RUN apt-get update && apt-get install -y --allow-change-held-packages --no-install-recommends software-properties-common openssh-client openssh-server pdsh curl sudo net-tools vim iputils-ping wget perl libxml-parser-perl libcudnn7=${CUDNN_VERSION} libnccl2=${NCCL_VERSION} libnccl-dev=${NCCL_VERSION} --allow-downgrades:

E: Unable to locate package libcudnn7
E: Version '2.4.7-1+cuda10.0' for 'libnccl2' was not found
E: Version '2.4.7-1+cuda10.0' for 'libnccl-dev' was not found

It seems that in the docker apt can't find the package. Is it my fault somewhere? or the docker may exist some bugs?

[araabovyan]

@IceCapriccio have you found the solution?

[IceCapriccio]

No. Because I'm not familiar with docker, so I run this project in the physical machine directly. Now everything is going well.

[araabovyan]

@IceCapriccio can you share, how exactly did you install without docker? And it would be so much help if you share the versions of packages. Thanks in advance!

[IceCapriccio]

Of course. I record all what I do to launch this project, and saved some links I referred, but most pages are in Chinese.
All version of dependencies is same to the version in links or the requirements.txt I uploaded.
By the way, I don't have sudo permission so all what I do goes in my user directory.

1. MPI: install OpenMPI.
2. install perl, refer to this
3. install python packages by this requirements.requirements.txt
4. softlink from en-core-web-sm to en :spacy link en_core_web_sm en
5. add x permision to ROUGE-1.5.5.pl
6. perl dependencies installation refer to this article. But not all work well, there is some problems in UserAgent and DB_File module.
7. Search and Download perl module still missing in UserAgent from cpan (you can simply run ROUGE-1.5.5.pl to get which module is missing) and install it manually.
8. BerkeleyDB is needed while installing DB_File module (not found db.h), so I followed this page to install it.
9. finally, my gpu memory is not enough to run this model, so I edit some hyper-params, like MAX_SENT_NUM and MAX_GEN_LENGTH to compress the model. (most finally I run this model in A100 and got very closed results to the results in paper)

@IceCapriccio can you share, how exactly did you install without docker? And it would be so much help if you share the versions of packages. Thanks in advance!