Closed mhuen closed 3 years ago
Update to tensorflow 2.3 (higher versions should probably work as well) and support for float 64 precision. Added additional value clipping of dom_charges logits before exp operation to avoid too large numbers.
Update to tensorflow 2.3 (higher versions should probably work as well) and support for float 64 precision. Added additional value clipping of dom_charges logits before exp operation to avoid too large numbers.