Use bool type instead of fp16 for zero_mask in sparse embedding - Githubissues

numenta / nupic.research

Experimental algorithms. Unsupported.

https://nupicresearch.readthedocs.io

GNU Affero General Public License v3.0

104 stars 60 forks source link

Use bool type instead of fp16 for zero_mask in sparse embedding #533

Closed lscheinkman closed 3 years ago

lscheinkman commented 3 years ago

Requires pytorch 1.7+