Open utterances-bot opened 9 months ago
Hi there! test model is not working properly here although it passes all the tests. The link to Qin You is also broken.
@qin-yu could you please confirm whether you received an email from utterances when I first opened this issue?
@esgomezm No there was only one email:
@qin-yu could you please confirm whether you received an email from utterances when I first opened this issue?
Hi there! test model is not working properly here although it passes all the tests. The link to Qin You is also broken.
There is one test about reproducing a test output using the test input. If this one passed, how can it not work properly 🤔 where can I find the log and errors?
Maybe there is a typo which caused the link to break? My family name is just Yu. Or do you mean this github_user: qin-yu
caused error? If you are talking about git repo not found that's because the paper and code are not published yet.
@esgomezm
Hi @esgomezm
I just tried to click "Test the model" on all three StarDist models available on Bioimage.IO, none of them finishes "initialising" phase, i.e. they just disappeared after showing "initialising". So I guess it's not a news for you that StarDist doesn't run on Bioimage.IO.
I still don't understand what do you mean by "not working properly here"
hi @qin-yu sorry that we have a server outage at the moment, the disk is full, will be restored soon.
hi @qin-yu sorry that we have a server outage at the moment, the disk is full, will be restored soon.
Thanks @oeway and @esgomezm ~
I just checked again on my phone, "Test the model" shows the interface after initialisation. I'll make the repo public once we submit the paper.
Hi @oeway I need your help
I tried to run a 1 x 192 x 192 x 192 x 1
float32
small crop in npy
format from my training data, and the test of run stopped at this error. It looks like bioengine-model-runner
cannot find modest-octopus
though my model has been listed on https://uk1s3.embassy.ebi.ac.uk/model-repository
I would greatly appreciate it if you could take a moment to look into this. Should I create an issue somewhere?
Plugin mi-core_ryq051697132421341: Failed to run the model (10.5281/zenodo.8421755) in the BioEngine, error: Traceback (most recent call last):
File "/models/bioengine-model-runner/1/model.py", line 120, in _process_request
model_adapter=TritonModelAdapter(
File "/models/bioengine-model-runner/1/triton_model_adapter.py", line 20, in __init__
response = client.get_model_config(self._model_id)
File "/tmp/python_env_KTuXRi/1/lib/python3.8/site-packages/tritonclient/http/__init__.py", line 587, in get_model_config
_raise_if_error(response)
File "/tmp/python_env_KTuXRi/1/lib/python3.8/site-packages/tritonclient/http/__init__.py", line 65, in _raise_if_error
raise error
tritonclient.utils.InferenceServerException: Request for unknown model: 'modest-octopus' is not found
<Contents>
<Key>modest-octopus/1/model.savedmodel/saved_model.pb</Key>
<LastModified>2023-10-12T01:41:58.917Z</LastModified>
<ETag>"b9efbf6c5cfd8073aac9aa790623d7e9"</ETag>
<Size>110737</Size>
<StorageClass>STANDARD</StorageClass>
<Owner>
<ID>imjoy-bia-47050780</ID>
<DisplayName>Wei Ouyang</DisplayName>
</Owner>
<Type>Normal</Type>
</Contents>
<Contents>
<Key>modest-octopus/1/model.savedmodel/variables/variables.data-00000-of-00001</Key>
<LastModified>2023-10-12T01:42:11.583Z</LastModified>
<ETag>"488f28cf03c55cce7244971f891b1e99-3"</ETag>
<Size>18168328</Size>
<StorageClass>STANDARD</StorageClass>
<Owner>
<ID>imjoy-bia-47050780</ID>
<DisplayName>Wei Ouyang</DisplayName>
</Owner>
<Type>Normal</Type>
</Contents>
<Contents>
<Key>modest-octopus/1/model.savedmodel/variables/variables.index</Key>
<LastModified>2023-10-12T01:42:11.900Z</LastModified>
<ETag>"50fecc1e01d6d4d23a876ab08827f715"</ETag>
<Size>1687</Size>
<StorageClass>STANDARD</StorageClass>
<Owner>
<ID>imjoy-bia-47050780</ID>
<DisplayName>Wei Ouyang</DisplayName>
</Owner>
<Type>Normal</Type>
</Contents>
<Contents>
<Key>modest-octopus/config.pbtxt</Key>
<LastModified>2023-10-12T01:41:58.495Z</LastModified>
<ETag>"443b5f595a8e0f9a790ce06693fea11c"</ETag>
<Size>160</Size>
<StorageClass>STANDARD</StorageClass>
<Owner>
<ID>imjoy-bia-47050780</ID>
<DisplayName>Wei Ouyang</DisplayName>
</Owner>
<Type>Normal</Type>
</Contents>
<Contents>
<Key>modest-octopus/rdf.yaml</Key>
<LastModified>2023-10-12T01:41:58.248Z</LastModified>
<ETag>"af180c69887d83c1cda1ca44dcd03b80"</ETag>
<Size>4352</Size>
<StorageClass>STANDARD</StorageClass>
<Owner>
<ID>imjoy-bia-47050780</ID>
<DisplayName>Wei Ouyang</DisplayName>
</Owner>
<Type>Normal</Type>
</Contents>
Hi @oeway I was checking again the website, it's getting worse now, the cover images don't even load anymore...
e.g.
GET https://zenodo.org/api/files/70891daf-e746-4552-8824-9e0120ff962d/stardist_raw_and_segmentation.jpg 404 (NOT FOUND)
zenodo server is down 😓
zenodo server is down 😓
Thanks for checking. Didn't know it can happen like this (only api has been down
zenodo server is down 😓
Thanks for checking. Didn't know it can happen like this (only api has been down
To be exact, zenodo has just replaced its backend entirely and this make everything fails in the model zoo: https://github.com/bioimage-io/collection-bioimage-io/issues/653
We will look into the fix, sorry about this.
Ohh I see. Many thanks~
@qin-yu I just tested your model, it works but a bit slow, maybe due to the large volume in the test image:
The results seems a bit wired, maybe some issue in the display? Could you test it with python directly, we have some code snippet here: https://github.com/bioimage-io/bioengine
The results seems a bit wired, maybe some issue in the display? Could you test it with python directly, we have some code snippet here: https://github.com/bioimage-io/bioengine
Thanks for testing @oeway Does the visualisation take the 97 channel output from the ResNet or an actual segmentation?
Does the bioimage.io pipeline also doing the StarDist postprocessing before the visualisation? @FynnBe
no, the stardist postprocessing is not applied here
no, the stardist postprocessing is not applied here
@oeway 👆 I guess the visualisation on bioimage.io is trying to show a 97 channel image, which looks unnatural to us. For models like StarDist, the online inference only make sense if at least a probability map is obtained by some postprocessing steps. But good to hear the model runs.
Hi @oeway @FynnBe do you know what's wrong with the import for me:
When I do:
from bioimageio.spec import load_raw_resource_description
load_raw_resource_description(
"https://bioimage-io.github.io/collection-bioimage-io/rdfs/10.5281/zenodo.8421755/8432366/rdf.yaml",
update_to_format="latest"
)
The error is:
Exception has occurred: ValidationError
{'documentation': ['Errors in all options for this field. Fix any of the following errors:', ['Empty URI scheme component'], ['Invalid suffix (content-4): Replacement index 0 out of range for positional args tuple']]}
File "/g/kreshuk/yu/repositories/run-stardist/bioimageio/final-for2581-stardist/debug_import.py", line 3, in <module>
model = import_bioimageio("https://bioimage-io.github.io/collection-bioimage-io/rdfs/10.5281/zenodo.8421755/8432366/rdf.yaml", "/g/kreshuk/yu/repositories/run-stardist/bioimageio/stardist_model_download")
marshmallow.exceptions.ValidationError: {'documentation': ['Errors in all options for this field. Fix any of the following errors:', ['Empty URI scheme component'], ['Invalid suffix (content-4): Replacement index 0 out of range for positional args tuple']]}
while this is the returned 'document'
with an empty URI scheme:
yes, could you please check that you are using the latest bioimageio.spec 0.4.9post4 ? the post4 release addresses exactly this issue.
yes, could you please check that you are using the latest bioimageio.spec 0.4.9post4 ? the post4 release addresses exactly this issue.
That's a new version from yesterday 🤣 I'll update my environment
StarDist Plant Nuclei 3D ResNet
Bioimage.io -- an AI model repository for deep learning.
https://bioimage.io/?partner=stardist&type=model&id=10.5281%2Fzenodo.8421755