With or without a paid subscription, Google now allows users to purchase additional Colab computing time on demand without affecting the current free tier plan.
The only major change is that users can purchase access in “Compute Units” — 100 units start at $9.99, and 500 units start at $49.99.
Google Colab product lead Chris Perry in a TensorFlow post on Thursdayblog postExplained in:
Paid users now have the flexibility to take full advantage of their compute quota at any rate they choose.
When the ‘computing unit’ balance is exhausted, the user can choose to purchase more ‘pay-as-you-go’ resources.
After the paid tier is consumed, the user’s Colab quota will fall back to the limited free tier.
With the introduction of the pay-as-you-go policy,Google also announced that paying users will now be able to choose between standard/premium GPU configurations.
The former usually refers to NVIDIA T4 / Tensor Core GPUs, while the latter usually refers to NVIDIA V100 or A100 Tensor Core hardware.
However, the company does not guarantee that customers will be allocated a specific GPU chip type, because it depends on many factors – including the availability of resources, and the user’s account balance.
Of course, advanced GPUs also consume ‘computing units’ faster than standard configurations.
A few weeks ago, Google started rolling out a pay-as-you-go option for Colab.At the time, the company emailed users informing them that it was offering subscribers the aforementioned “computing unit” configuration.
Officials say the change will make the service more transparent, allowing users to have more control over how and when they use Colab services.
On the other hand, some people think that Google is secretly trying to charge users more or limit the use of Colab resources.
Even so, a Google spokesperson responded in a statement that restrictions have always existed in the Colab tier plan for each gear.