'Can PyTorch / XLA be used for the Coral dev board TPUs?
The last discussion of this I found was before [this article]https://cloud.google.com/blog/products/ai-machine-learning/pytorch-is-now-ga-on-google-cloud-tpus came out.
Has anyone had any experience or know if it is possible to use a pytorch model directly on the devboard directly with XLA instead of converting to a TFLite model first?
Solution 1:[1]
To run the models on Coral TPU models need to be compile with compiler, which maps the operations to TPU/CPU. As compiler accept only TFLite model as input it is not possible to use a pytorch model directly on the devboard.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|---|
| Solution 1 | Coral So Support |