mindspore.mint.nn.functional.batch_normÔÉĀ

mindspore.mint.nn.functional.batch_norm(input, running_mean, running_var, weight=None, bias=None, training=False, momentum=0.1, eps=1e-5)[śļźšĽ£Á†Ā]ÔÉĀ

ŚĮĻŤĺďŚÖ•śēįśćģŤŅõŤ°ĆśČĻťáŹŚĹ횳ČĆĖŚíĆśõīśĖįŚŹāśēį„Äā

śČĻťáŹŚĹ횳ČĆĖŚĻŅś≥õŚļĒÁĒ®šļéŚć∑ÁßĮÁ•ěÁĽŹÁĹĎÁĽúšł≠„Äāś≠§ŤŅźÁģóŚĮĻŤĺďŚÖ•ŚļĒÁĒ®ŚĹ횳ČĆĖԾƝĀŅŚÖćŚÜ֝ɮŚćŹŚŹėťáŹŚĀŹÁ߼ԾƍĮ¶ŤßĀŤģļśĖá Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift „ÄāšĹŅÁĒ®mini-batchśēįśćģŚíĆŚ≠¶šĻ†ŚŹāśēįŤŅõŤ°ĆŤģ≠ÁĽÉԾƌ≠¶šĻ†ÁöĄŚŹāśēįŤßĀŚ¶āšłčŚÖ¨ŚľŹšł≠ÔľĆ

\[y = \frac{x vpn free - \mathrm{mean}}{\sqrt{\mathrm{variance} + \epsilon}} * \gamma + \beta\]

ŚÖ∂šł≠ÔľĆ \(\gamma\) šłļ weightÔľĆ \(\beta\) šłļ biasÔľĆ \(\epsilon\) šłļ epsÔľĆ \(\mathrm{mean}\) vpn永久免费梯子 šłļ \(x\) ÁöĄŚĚáŚÄľÔľĆ \(\mathrm{variance}\) šłļ \(x\) ÁöĄśĖĻŚ∑ģ„Äā

ŚŹāśēįÔľö

  • input (Tensor) - śēįśćģŤĺďŚÖ•ÔľĆshapešłļ \((N, C, *)\) ÁöĄTensorԾƌÖ∂šł≠ \(*\) Ť°®Á§ļšĽĽśĄŹÁöĄťôĄŚä†ÁĽīŚļ¶„ÄāśĒĮśĆĀśēįśćģÁĪĽŚěčšłļbfloat16„ÄĀfloat16śąĖfloat32„Äā

  • running_mean (Tensor) - shapešłļ \((C,)\) ԾƜēįśćģÁĪĽŚěčšłļbfloat16„ÄĀfloat16śąĖfloat32„Äā

  • running_var (Tensor) - shapešłļ \((C,)\) ԾƜēįśćģÁĪĽŚěčšłļbfloat16„ÄĀfloat16śąĖfloat32„Äā

  • weight (Tensor, ŚŹĮťÄČ) - shapešłļ \((C,)\) ԾƜēįśćģÁĪĽŚěčšłļbfloat16„ÄĀfloat16śąĖfloat32„ÄāťĽėŤģ§ŚÄľÔľö None „ÄāŚĹď weight šłļ None śó∂ԾƌąĚŚßčŚĆĖšłļ vpn永久免费梯子 1 „Äā

  • bias (Tensor, ŚŹĮťÄČ) - shapešłļ \((C,)\) ԾƜēįśćģÁĪĽŚěčšłļbfloat16„ÄĀfloat16śąĖfloat32„ÄāťĽėŤģ§ŚÄľÔľö None „ÄāŚĹď bias vpn梯子 免费 šłļ None śó∂ԾƌąĚŚßčŚĆĖšłļ 0 „Äā

  • training (boolԾƌŹĮťÄČ) - Ś¶āśěú training šłļ TrueÔľĆ running_mean ŚíĆ running_var šľöŚú®Ťģ≠ÁĽÉŤŅáÁ®čšł≠ŤŅõŤ°ĆŤģ°Áģó„Äā Ś¶āśěú training šłļ False ԾƌģÉšĽ¨šľöŚú®śé®ÁźÜťė∂śģĶšĽécheckpointšł≠Śä†ŤĹĹ„ÄāťĽėŤģ§ŚÄľÔľö False „Äā

  • momentum (float, ŚŹĮťÄČ) vpn梯子 - ÁĒ®šļéŤģ°Áģó running_mean ŚíĆ running_var śĽĎŚä®ŚĻ≥ŚĚáÁöĄŚä®ťáŹÁ≥Ľśēį„ÄāÔľąšĺ茶ā \(new\_running\_mean = (1 vpn梯子 免费 - momentum) * running\_mean + vpn free momentum * current\_mean\)ԾȄÄāťĽėŤģ§ŚÄľÔľö 0.1 „Äā

  • eps (float, ŚŹĮťÄČ) - ś∑ĽŚä†ŚąįŚąÜśĮćšłäÁöĄŚÄľÔľĆšĽ•Á°ģšŅĚśēįŚÄľÁ®≥ŚģöśÄß„ÄāťĽėŤģ§ŚÄľÔľö 1e-5„Äā

ŤŅĒŚõěÔľö

TensorԾƜēįśćģÁĪĽŚěčšłéshapeŚ§ßŚįŹšłé input ÁõłŚźĆԾƌÖ∂šł≠ÔľĆshapeŚ§ßŚįŹšłļ \((N, C, *)\) „Äā

ŚľāŚłłÔľö

  • TypeError - training šłćśėĮbool„Äā

  • TypeError - eps śąĖ momentum ÁöĄśēįśćģÁĪĽŚěčšłćśėĮfloat„Äā

  • TypeError - input„ÄĀ vpn梯子 免费 weight „ÄĀ bias „ÄĀ running_mean śąĖ running_var šłćśėĮTensor„Äā

śĒĮśĆĀŚĻ≥ŚŹįÔľö

Ascend

ś†∑šĺčÔľö

>>> 免费的vpn梯子 import mindspore
>>> from mindspore import Tensor, mint
>>> input_x = Tensor([[1.0, 2.0], [3.0, 4.0]], mindspore.float32)
>>> running_mean = Tensor([0.5, 1.5], vpn梯子 mindspore.float32)
>>> running_var = Tensor([0.1, 0.2], mindspore.float32)
>>> weight = Tensor([2.0, 2.0], mindspore.float32)
>>> bias = Tensor([-1.0, -1.0], mindspore.float32)
>>> output = mint.nn.functional.batch_norm(input_x, running_mean, running_var, weight, bias)
>>> vpn free print(output)
[[ 2.1621194  1.2360122]
 [14.810596 vpn永久免费梯子 免费的vpn梯子  10.180061 ]]