原书出处
Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)
set random seeds or setting temperature to zero and using greedy sampling. With all inputs held constant, we expect repeatable outputs, even with generative models. Some lower-level optimizations, such as parallelism in compute or model quantization, may still produce nondeterministic behavior. But, in general, it is still possible to make execution deterministic through disabling these optimizations, if this is important enough to test. Even if
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)/_epub-src/OEBPS/ch06.html
EPUB SHA-256:d98501f374a474910f195f0adfd53c1d895ae8ddc5822476c05cd7ba006ac63d