Oops something went wrong, there might be issues with this model

Community Help
1

Hello i’m working on a crater detector - Sign in to Roboflow

I keep getting the Oops something went wrong error.

Any ideas what could cause this?

J

2

Hi @John_Tapsfield

Sorry to hear you’re having issues. Can you provide more details as to when and where this error is occurring, as well as the following information:

  • URL where the error occurred
  • Project Type
  • Operating System & Browser
  • Project Universe Link or Workspace/Project ID

Also, please attach screenshots of relevant Developer Tools error messages.

These details will be immensely helpful for us to help you solve your issues.

3

Hi there,

Thanks for getting back to me.

image.png
image.png1284×836 142 KB

4

Hi @John_Tapsfield

We’re taking a look into this issue for you and I’ll post any updates here.

5

Ok, thank you. Much appreciated.

I otherwise like the product

1 Like
6

@stellqsphere.

Hello, I don’t suppose you have any updates on why or how the model is failing?

It’s not a verbose error, thus hard to know if the issue is within the user or if you’re able to triage this.

Sadly sicne I’ve hit this snag, as an sme I’m reticent to spend a lot more money or time on this model.

Kind regards

John

Many thanks

John

7

I’m gonna try out the model and recreate this error, were you able to recreate the error? @leo

8

@John_Tapsfield Okay so check this out:

I went myself and saw firsthand what was happening. I think its something wrong with the deployment. In the forum post I linked, someone else had that problem. A field engineer instructed them to: “generate a new version and follow the process outlined in the linked post” the link being: ANNOUNCEMENT: Custom Model Weights Upload Overview It’s worth giving a shot.

If that does not work I’d recommend checking out the health check tab. something crucial im not sure of is if your images are being resized and fed properly. I looked in the health check tab of your project and over 90% of the images you used are critically large, which may cause a problem, it’s certainly setting off alarms in the health check section. Just something to note.

9

A fix went out fairly recently relating to a bug report that we filed based on this post. Could you try again?

Thanks for pitching in @Alejandro_Bermudez, were you able to recreate the error?

10

@leo @John_Tapsfield
The problem seems to occur when trying to access the “try pretrained model” via the website, however accessing the model via API calls and I’m assuming local hosting via weights should be no problem.

I was not able to recreate the problem, I think that this is a deployment issue because only the website-deployed access is causing trouble. I would try to recreate it using his public datasets but I don’t have any training credits to spare right now, I think something went wrong during the deployment and im not sure what. The model seems to work, I went ahead and tested out the model they made with a test image from his sample and was able to get back predictions no problem.

test
test2078×1246 191 KB

11

Hi both,

Really appreciate your input here - can you confirm that you think I should try to retrain the model as the next step here?
Happy to spend a little money on some more credits if you can direct how best to apply this.

Kind regards

John

12

@John_Tapsfield
I think its important to find out why the error is happening and get Roboflow’s opinion before spending any more money, if you were to simply retrain the model, and the problem had something to do with the data used, then retraining would reproduce the same problem. If it is not your fault you likely wont have to pay again. Additionally it might just be something they can fix on their end and you wont have to retrain the model at all.

Its important to note that the model seems to be working fine, just not on Roboflow Universe, which leads me to believe that the issue may be server-sided. This in mind, this I wouldn’t suggest spending money on retraining until you hear from Roboflow as if the website’s showcasing is the only issue, you should have no problems using the model for the intended usage.
@leo

13

Hi all,

We had a fix deployed on our open source inference server but it took a bit longer to get that fix on our hosted inference. We plan to have this fix out for our hosted inference today!

Thank you everyone for your patience!

1 Like
14

Thank you for letting us know!

We have been very impressed with Roboflow thus far - and can see a lot of utility in it going forwards.

Thanks for all your help and support - hopefully the patch / update resolves this.

Thanks

John

1 Like
15

Thank you for your kind words! Feel free to let us know if you run into any more issues!

16

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.