We have written the core PyTorch operations in WGSL, which speedup the calculations.
Also for devices/browsers which don't have GPU support. We have implemented fallback to CPU with NumPy polyfills implementing the PyTorch functions.
We have written the core PyTorch operations in WGSL, which speedup the calculations.
Also for devices/browsers which don't have GPU support. We have implemented fallback to CPU with NumPy polyfills implementing the PyTorch functions.