MCPcopy Create free account
hub / github.com/tensorflow/lingvo / Xavier

Method Xavier

lingvo/core/py_utils.py:1121–1123  ·  view source on GitHub ↗

Xavier initialization (x = sqrt(6. / (in + out)); [-x, x]).

(scale=1.0, seed=None)

Source from the content-addressed store, hash-verified

source not stored for this graph (policy: none)

Callers 15

testIsDefaultParamInitMethod · 0.80
testMultiHeadedCrossAttentionDotProductMethod · 0.80
testMultiHeadedAttentionDotProductSegmentMaskMethod · 0.80
testContextOutSegmentMaskMethod · 0.80
testFPropCrossAttentionMethod · 0.80
testExtendStepAsyncTimeStepSelfAttentionMethod · 0.80
testMultipleExtendStepAsyncTimeStepSelfAttentionMethod · 0.80
testExtendStepMethod · 0.80
testFPropCrossAttentionMethod · 0.80
testFPropSelfAttentionMethod · 0.80
testExtendStepAsyncTimeStepSelfAttentionMethod · 0.80
testExtendStepSelfAttentionMethod · 0.80

Calls 1

_ParamsMethod · 0.45

Tested by 15

testIsDefaultParamInitMethod · 0.64
testMultiHeadedCrossAttentionDotProductMethod · 0.64
testMultiHeadedAttentionDotProductSegmentMaskMethod · 0.64
testContextOutSegmentMaskMethod · 0.64
testFPropCrossAttentionMethod · 0.64
testExtendStepAsyncTimeStepSelfAttentionMethod · 0.64
testMultipleExtendStepAsyncTimeStepSelfAttentionMethod · 0.64
testExtendStepMethod · 0.64
testFPropCrossAttentionMethod · 0.64
testFPropSelfAttentionMethod · 0.64
testExtendStepAsyncTimeStepSelfAttentionMethod · 0.64
testExtendStepSelfAttentionMethod · 0.64