Search Options

Results per page
Sort
Preferred Languages
Advance

Results 1 - 1 of 1 for GetAccumulationType (0.17 sec)

  1. tensorflow/compiler/mlir/tf2xla/transforms/legalize_tf.cc

    // Softmax op utilities.
    //===----------------------------------------------------------------------===//
    
    // Returns the type to use for accumulating the given type.
    static Type GetAccumulationType(Type ty) {
      // Upcast 16 bit sum reductions to 32 bit to reduce the precision loss from
      // repeated floating point additions.
      return (ty.isF16() || ty.isBF16()) ? FloatType::getF32(ty.getContext()) : ty;
    }
    
    Registered: Sun Jun 16 05:45:23 UTC 2024
    - Last Modified: Tue Jun 11 20:00:43 UTC 2024
    - 291.8K bytes
    - Viewed (0)
Back to top