You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When a fusion has inputs with different types, the maximum vectorization factor is bounded by the largest data type. For example, a fp16 batchnorm with fp32 weights is vectorized by 4 at maximum as that's the limit for fp32.
The text was updated successfully, but these errors were encountered:
When a fusion has inputs with different types, the maximum vectorization factor is bounded by the largest data type. For example, a fp16 batchnorm with fp32 weights is vectorized by 4 at maximum as that's the limit for fp32.
The text was updated successfully, but these errors were encountered: