Troubleshooting

Common error messages you may see and how to resolve them

Working With Aquarium in the Browser

The Embedding View Does Not Load/Button is Greyed Out

If in Anonymous Mode:

1. Did you upload custom embeddings?

   1. Embeddings must be uploaded for both your frames and ALL of your crops

   2. Is your embedding vector <=2048 in length?

      1. If you would like to use an embedding vector longer than 2048 please contact Aquarium

Non Anonymous or Anonymous Mode

1. Aquarium requires you upload at least 101 images in order for us to start an embedding generation process for your data.

2. Look at the Streaming Uploads tab under the Project Overview screen

   1. If the Embedding Status column does not say done, the embeddings are still processing. Depending on your dataset size the processing can take a while, if it takes longer than a few hours with a reasonable dataset size, feel free to reach out to Aquarium to see if something may be amiss

   

General Python Client Upload Errors

OSError: Attempting to flush dataset to disk with less than 1 GB of available disk space. Exiting...

The machine you are running the script from needs to have more than 1GB of available disk space before you run in order to complete an upload.

Uploading with Custom Embeddings Error: This upload would exceed the following quota(s): Maximum Frame Embedding Length, Maximum Crop Embedding Length.

If you are uploading custom embeddings, the vector length defaults to 2048. This error is likely due to your embedding vector length being greater than 2048. If you would like to use a larger embedding, please contact Aquarium to adjust your maximum embedding length allowance.

Waiting for inferences to finish processing... Waiting for dataset to finish processing...

If you see this error in your console when you

Semantic Segmentation Data Upload Errors

Mutable streaming datasets do not support mask_data for semseg masks yet.

If you upload your mask data as a ndarray of pixel mask data, you do have to explicitly upload your data using our BATCH methodology. If possible, we better support using mask_url to upload a greyscale image form of your mask data. Our BATCH methodology is only used for uploading semseg data via mask_data right now, please reach out to the Aquarium team for assistance.

Working with Aquarium from a Jupyter Notebook

InvalidResponse: ('Request failed with status code', 201, 'Expected one of', <HTTPStatus.OK: 200>)

This error most commonly manifests when running an upload for the first time, shortly after installing the aquariumlearning package in your notebook python environment. This occurs due to an older version of a package we depend on google-resumable-media, being sourced in your environment.

The Solution is to restart your notebook kernel after installing the aquariumlearning package for the first time. This can be done by restarting the kernel manually, or running the following in

import importlib
importlib.reload(al)