Timing out when attempting to make a prediction?

First off - thank you so much for creating this amazing resource!

Expected Behavior

Receive a response from the MSA server.

Current Behavior

Timeouts when trying to query the MSA server for any protein sequence.

Steps to Reproduce (for bugs)

When trying to run colabfold-batch for any sequence, I get the below timeout error(s).

ColabFold Output (for bugs)

2024-08-29 11:03:36,940 Running colabfold 1.5.5
2024-08-29 11:03:37,350 Unable to initialize backend 'rocm': NOT_FOUND: Could not find registered platform with name: "rocm". Available platform names are: Interpreter CUDA
2024-08-29 11:03:37,351 Unable to initialize backend 'tpu': module 'jaxlib.xla_extension' has no attribute 'get_tpu_client'
2024-08-29 11:03:40,213 Running on GPU
2024-08-29 11:03:40,985 Found 9 citations for tools or databases
2024-08-29 11:03:40,986 Query 1/1: cd6a78b82fede215cdb80a797f228098a41178fdde0753728f3781e6221fb1ee (length 1167)
2024-08-29 11:03:56,000 Timeout while fetching result from MSA server. Retrying...
2024-08-29 11:05:16,019 Error while fetching result from MSA server. Retrying... (1/5)
2024-08-29 11:05:16,023 Error: HTTPSConnectionPool(host='api.colabfold.com', port=443): Read timed out.

Context

curl https://api.colabfold.com/queue returns

{"queued":0}

Your Environment

I am currently running ColabFold locally using the docker container (https://github.com/sokrypton/ColabFold/wiki/Running-ColabFold-in-Docker) on a cluster system running CentOS7, with a Tesla T4 GPU. The cluster does have access to the internet.

Start with the Docker setup described in the Running-ColabFold-in-Docker wiki and reproduce the colabfold-batch request against api.colabfold.com. Compare the MSA timeout output with the /queue response and the CentOS7 environment. Done would require identifying whether the failure is in the container, network path, or MSA service and documenting an actionable resolution.