If my keras model compile failed due to out of memory. does changing the data set from float32 to float16 might fix the problem? if not, what is the advantage of decreasing float32 to float16?
Thanks
1.4m articles
1.4m replys
5 comments
56.8k users