How to download training logs to view in tensorflow?

[the-it-weirdo]

Search before asking

• I have searched the HUB issues and discussions and found no similar questions.

Question

Hello, I trained a model with my own dataset. I used cloud training. I can see the graphs in the Ultralytics Hub. However, I am unable to add any axis title to the graph. My searches tell me that I can add axis title using tensorboard. How do I access the tranining logs for the same? How can I download the logs to open with tensorboard?

Additional

No response

[sergiuwaxmann]

@the-it-weirdo Hello!

Unfortunately, I am not very familiar with TensorBoard but if it helps, you might be able to get the data you need from the PyTorch export.

Alternatively, you can get the metrics using Ultralytics HUB SDK:

from hub_sdk import HUBClient

# Initialize the HUBClient
credentials = {"api_key": "<API-KEY>"}
client = HUBClient(credentials)
print("Authenticated:", client.authenticated)

# Get model by ID
model = client.model("<MODEL-ID>")

# Get metrics
metrics = model.get_metrics()
print("Metrics:", metrics)

# Get weights
weight_url = model.get_weights_url("best")
print("Weight URL:", weight_url)

Make sure to replace <API-KEY> and <MODEL-ID> with your actual API key and model ID. This script will authenticate you, fetch the training metrics, and provide a direct link to download the best weights.