Hi,
Is it possible to submit both preprocessing & training images to the Express Lane stage (just like we submit both for the main dataset)? The current Express Lane allows only submission of the training docker image, not the submission.txt, like we do for the main challenge.
Thanks!
Created by Aashish Dattani zettata thanks. i see .i didn't know the credentials are stored. Hi Yuanfang,
Are you sure that your credentials are not saved in `$HOME/.docker/config.json`? I tried the following on a machine where none of my credentials are saved:
```
# docker push docker.synapse.org//my-dm-preprocess-png
The push refers to a repository [docker.synapse.org//my-dm-preprocess-png]
d1919a61ac90: Preparing
41ed3221fc30: Preparing
63e222febf30: Preparing
4c0ae271473c: Preparing
4dcfe770bead: Preparing
2a72c3ab4190: Waiting
97ca462ad9ee: Waiting
unauthorized: authentication required
# docker pull docker.synapse.org//dm-preprocess-helloworld
Using default tag: latest
Pulling repository docker.synapse.org//dm-preprocess-helloworld
Error: image /dm-preprocess-helloworld:latest not found
#
```
The first command shows that I cannot push to my private project. The second shows that it can not download the image from a private project (the Docker image dm-preprocess-helloworld:latest does exist).
Can you please confirm that your credentials are actually saved as this would be an important issue otherwise? By the way, is your project private?
Thanks!
PS: Please open a new thread each time you meet a significantly different issue. so. i am really confused on how the docker in general works, i found i could push in my repository, with any password and username, is that normal?
i pushed in this submission with the user ID 'aaa' : https://www.synapse.org/#!Synapse:syn7511482
> e.g. when encounter a '.' or when encounter a weird view not presenting in 1% or encountering a patient that does not have standard view, eg. cc, mcl.
The best approach would be to make your method ignore the cases that you don't want to consider (e.g. when at least one of the standard CC/MLO view is missing).
> i found i cannot access model state, as unreadable--not as said in the recent email.
We have only known for a few days that Group Health is allowing us to return you the trained models. We are currently still working on that feature. Yes, the model is limited to 1 GB. I've added a section to the wiki page of the Express Lane (the update will be deployed shortly) to mention a few updates, including the fact that you now have access to a 200 GB scratch space (/scratch). The content of this space is cleaned before running any container. e.g. when encounter a '.' or when encounter a weird view not presenting in 1% or encountering a patient that does not have standard view, eg. cc, mcl. i actually haven't tried anything or have any method, but just imagine a few potential errors. but i hope the same potential crash could occur in both set, as real data. of course time should be limited so that it is still express.
i found i cannot access model state, as unreadable--not as said in the recent email. and also that means model state maximal is 1G, instead of 5G, as was previously stated? > can express line also contain full data but cut at 20min?
We use a small "dummy" dataset for the express lane.
> error happens in the middle when some corner case is there.
What kind of error generated by your method are you referring to (crash or incorrect results)? Dear Aashish,
The express lane for preprocessing and training works just like the real queue, but the dataset is smaller and the size of the preprocessing folder is also smaller. Are you having difficulties with submitting like you do for the main challenge?
Best,
Thomas can express line also contain full data but cut at 20min? error happens in the middle when some corner case is there.
Drop files to upload
Express Lane for preprocessing and training? page is loading…