I'm training my model on Kaggle GPU, but it's taking too long. So, I decided to switch to TPU for faster training.

I enabled TPU in the notebook settings and ran the following code to initialize it:

import tensorflow as tf
    try:
        tpu = tf.distribute.cluster_resolver.TPUClusterResolver()  # Detect TPU
        tf.config.experimental_connect_to_cluster(tpu)
        tf.tpu.experimental.initialize_tpu_system(tpu)
        strategy = tf.distribute.TPUStrategy(tpu)
        print("TPU initialized")
    except ValueError:
        strategy = tf.distribute.get_strategy()  # Default strategy for CPU and single GPU
        print("TPU not found, using default strategy")

However, I keep getting this output:

TPU not found, using default strategy

What I Have Checked:

• TPU is enabled in notebook settings
• I still have TPU quota available this week
• Restarted the session and tried again

Additionally, the Draft Session says:

Utilization is not currently available for TPU VMs.

enter image description here

Has Kaggle stopped providing TPU access, or is there a missing step in my setup? If TPU is still available, how can I get it to work properly?