Nebari deployment

[jourdain]

The log seems to report the following

[I 241010 19:47:05 proxyhandlers:760] Process mainprocess failed with stdout: Some weights of the model checkpoint at facebook/detr-resnet-50 were not used when initializing DetrForObjectDetection: ['model.backbone
.conv_encoder.model.layer1.0.downsample.1.num_batches_tracked', 'model.backbone.conv_encoder.model.layer2.0.downsample.1.num_batches_tracked', 'model.backbone.conv_encoder.model.layer3.0.downsample.1.num_batches_tr
acked', 'model.backbone.conv_encoder.model.layer4.0.downsample.1.num_batches_tracked']
    - This IS expected if you are initializing DetrForObjectDetection from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model fro
m a BertForPreTraining model).
    - This IS NOT expected if you are initializing DetrForObjectDetection from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSeq
uenceClassification model).
    ERROR:asyncio:Task exception was never retrieved
    future: <Task finished name='Task-9' coro=<debounce.<locals>.decorator.<locals>.wrapper.<locals>.debounced() done, defined at /Users/aktech/miniconda3/envs/jhub-apps-dev/lib/python3.12/site-packages/nrtk_explor
er/library/debounce.py:17> exception=TypeError("object NoneType can't be used in 'await' expression")>
    Traceback (most recent call last):
      File "/Users/aktech/miniconda3/envs/jhub-apps-dev/lib/python3.12/site-packages/nrtk_explorer/library/debounce.py", line 22, in debounced
        await func(*args, **kwargs)
    TypeError: object NoneType can't be used in 'await' expression
====================================================

[jourdain]

@PaulHax @vicentebolea any idea on what could be happening?

[vicentebolea]

It looks like a libraries version issue, maybe moving to an older version of transforms could "fix" this

[PaulHax]

Not 100% sure this gets it, but I get the same error when I change debounce.py to use create_task from trame.app.asynchronous rather than from asyncio

https://github.com/Kitware/nrtk-explorer/pull/132/files#diff-d7955bbc308df5cde56ee6be7255b7d3bd6b4da234963e107f0ef9eea7b948c8L28-R

[jourdain]

trame.app.asynchronous attach a callback to be aware of a failure on the task that will be hidden if you just call asyncio.create_task

[jourdain]

@PaulHax can you double check that is truly resolved?

It worked fine without the proxy and I suspected there is some issue with overlapping aysnc servers from proxy and your app, but
before digging too deep, I wanted to share it with you to find out if there is something app related you can fix.

Also, I don't see the error 100% of the time, sometimes it works (in the proxy) and sometimes it throws the above error.

The command I run to test quickly with proxy (CLI, outside of jhub-apps):

python -m jhsingle_native_proxy.main --authtype=none --no-force-alive --conda-env= python {-}m nrtk_explorer.app.main {--}server {--}host 0.0.0.0 {--}port {port}

[PaulHax]

I'm not reproducing the error =/

We have not release the hypothedical fix 8841d82 to PyPI or Conda. Is Nebari pulling nrtk-explorer from main branch or using package manager?

I'm happy to remove the one place we are debouncing: https://github.com/Kitware/nrtk-explorer/blob/main/src/nrtk_explorer/app/core.py#L102

Was not much of an UX improvment.

[jourdain]

Once we have that fix in a release, I should try it on Nebari to see if we fixed the issue. Nebari is using a conda environment, but we do a pip install inside that env.