- Sort Score
- Result 10 results
- Languages All
Results 1 - 3 of 3 for nudge (0.03 sec)
-
tensorflow/compiler/mlir/quantization/common/ir/FakeQuantSupport.cc
// a consequence some values, which are supposed in the original [rmin, rmax] // range will be outside the shifted range and be clamped during quantization. // TODO: we should nudge the scale as well, but that requires the // fake quant op used in the training to use the nudged scale as well. static void getNudgedScaleAndZeroPoint(int64_t qmin, int64_t qmax, double rmin, double rmax, double &scale,
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Mar 21 11:52:27 UTC 2024 - 7.7K bytes - Viewed (0) -
subprojects/core/src/integTest/groovy/org/gradle/internal/environment/GradleBuildEnvironmentIntegrationTest.groovy
assert buildEnvironment.isLongLivingProcess() == ${daemon} } } task judge(type: DaemonJudge) """ // Force to start a new JVM file("gradle.properties") << "org.gradle.jvmargs=-Xmx32m" expect: succeeds('judge', arg) where: daemon | arg true | '--daemon' false | '--no-daemon' }
Registered: Wed Jun 12 18:38:38 UTC 2024 - Last Modified: Fri Sep 01 03:07:53 UTC 2023 - 1.9K bytes - Viewed (0) -
tensorflow/compiler/mlir/quantization/common/ir/FakeQuantSupport.h
// - numBits < 8 is promoted to uint8 or int8 // - "narrow_range" narrows the lower bound of the storage type's range by // 1 // - the specified min/max values are "nudged" so that the result has a zero // that can be exactly expressed // - min=max=0 implies scale=0 and zero_point=0 // // With the above assumptions applied, every conforming specified FakeQuant op
Registered: Sun Jun 16 05:45:23 UTC 2024 - Last Modified: Thu Mar 21 11:52:27 UTC 2024 - 3.7K bytes - Viewed (0)