Can't git clone submodules of training

[chiungliang]

Hi,

To run inference, some of codes are from mlcommons/training.
I bump into the same issue reported in mlcommons/training.
Here is the link of the question.
mlcommons/training#446

Is there anybody know how to solve the problem?
Lot of thanks.

[christ1ne]

@chiungliang can you share the exact steps you ran and the corresponding error messages? Thanks!

[chiungliang]

$$ git clone --recurse-submodules https://github.com/mlperf/training.git
Cloning into 'training'...
remote: Enumerating objects: 43, done.
remote: Counting objects: 100% (43/43), done.
remote: Compressing objects: 100% (41/41), done.
remote: Total 4034 (delta 6), reused 3 (delta 1), pack-reused 3991
Receiving objects: 100% (4034/4034), 79.01 MiB | 17.54 MiB/s, done.
Resolving deltas: 100% (1619/1619), done.
Submodule 'community' (https://github.com/mlperf/community.git) registered for path 'community'
Submodule 'compliance/verify_submission/mlperf_submission_helper/mlp_compliance' (https://github.com/bitfort/mlp_compliance) registered for path 'compliance/verify_submission/mlperf_submission_helper/mlp_compliance'
Submodule 'recommendation/dlrm' (https://github.com/facebookresearch/dlrm.git) registered for path 'recommendation/dlrm'
Cloning into '/work/xxx/try/training/community'...
Username for 'https://github.com':
Password for 'https://chiungliang@github.com':
remote: Invalid username or password.
fatal: Authentication failed for 'https://github.com/mlperf/community.git/'
fatal: clone of 'https://github.com/mlperf/community.git' into submodule path '/work/xxx/try/training/community' failed
Failed to clone 'community'. Retry scheduled
Cloning into '/work/xxx/try/training/compliance/verify_submission/mlperf_submission_helper/mlp_compliance'...
remote: Enumerating objects: 114, done.
remote: Total 114 (delta 0), reused 0 (delta 0), pack-reused 114
Receiving objects: 100% (114/114), 44.46 KiB | 1.20 MiB/s, done.
Resolving deltas: 100% (57/57), done.
Cloning into '/work/xxx/try/training/recommendation/dlrm'...
remote: Enumerating objects: 7, done.
remote: Counting objects: 100% (7/7), done.
remote: Compressing objects: 100% (6/6), done.
remote: Total 512 (delta 1), reused 1 (delta 1), pack-reused 505
Receiving objects: 100% (512/512), 1.33 MiB | 6.98 MiB/s, done.
Resolving deltas: 100% (299/299), done.
Cloning into '/work/xxx/try/training/community'...
Username for 'https://github.com':

[guschmue]

try
git clone --recursive https://github.com/mlcommons/training

[chiungliang]

Got the same messages.

Can't access training/community @ eebf719
https://github.com/mlperf/community/tree/eebf719c1b2174243c5d275dc57acc045bb3722d

Got 404 NOT FOUND

[guschmue]

Yes, community is a private repo. Looks it has not been updated for very long.
training wg would need to look at this.

[christ1ne]

tracked at: mlcommons/training#446

[EtoDemerzel0427]

@guschmue Hi, so how should we set up the inference environment with this problem unsolved?

[guschmue]

I think the problem is that:

[submodule "community"]
        path = community
        url = https://github.com/mlperf/community.git

does not exists anymore.

The following should be a work for the dlrm instructions:

git clone https://github.com/mlcommons/training.git
git submodule deinit -f community
git  rm -f community
git submodule update --init

[EtoDemerzel0427]

I think the problem is that:

[submodule "community"]
        path = community
        url = https://github.com/mlperf/community.git

does not exists anymore.

The following should be a work for the dlrm instructions:

git clone https://github.com/mlcommons/training.git
git submodule deinit -f community
git  rm -f community
git submodule update --init

@guschmue Thanks, it worked.

[vineetgarc]

Looks like the workaround is till needed. Can this be fixed please !