1.. role:: hidden 2 :class: hidden-section 3 4torch.nn.functional 5=================== 6 7.. currentmodule:: torch.nn.functional 8 9Convolution functions 10---------------------------------- 11 12.. autosummary:: 13 :toctree: generated 14 :nosignatures: 15 16 conv1d 17 conv2d 18 conv3d 19 conv_transpose1d 20 conv_transpose2d 21 conv_transpose3d 22 unfold 23 fold 24 25Pooling functions 26---------------------------------- 27 28.. autosummary:: 29 :toctree: generated 30 :nosignatures: 31 32 avg_pool1d 33 avg_pool2d 34 avg_pool3d 35 max_pool1d 36 max_pool2d 37 max_pool3d 38 max_unpool1d 39 max_unpool2d 40 max_unpool3d 41 lp_pool1d 42 lp_pool2d 43 lp_pool3d 44 adaptive_max_pool1d 45 adaptive_max_pool2d 46 adaptive_max_pool3d 47 adaptive_avg_pool1d 48 adaptive_avg_pool2d 49 adaptive_avg_pool3d 50 fractional_max_pool2d 51 fractional_max_pool3d 52 53Attention Mechanisms 54------------------------------- 55 56The :mod:`torch.nn.attention.bias` module contains attention_biases that are designed to be used with 57scaled_dot_product_attention. 58 59.. autosummary:: 60 :toctree: generated 61 :nosignatures: 62 63 scaled_dot_product_attention 64 65Non-linear activation functions 66------------------------------- 67 68.. autosummary:: 69 :toctree: generated 70 :nosignatures: 71 72 threshold 73 threshold_ 74 relu 75 relu_ 76 hardtanh 77 hardtanh_ 78 hardswish 79 relu6 80 elu 81 elu_ 82 selu 83 celu 84 leaky_relu 85 leaky_relu_ 86 prelu 87 rrelu 88 rrelu_ 89 glu 90 gelu 91 logsigmoid 92 hardshrink 93 tanhshrink 94 softsign 95 softplus 96 softmin 97 softmax 98 softshrink 99 gumbel_softmax 100 log_softmax 101 tanh 102 sigmoid 103 hardsigmoid 104 silu 105 mish 106 batch_norm 107 group_norm 108 instance_norm 109 layer_norm 110 local_response_norm 111 rms_norm 112 normalize 113 114.. _Link 1: https://arxiv.org/abs/1611.00712 115.. _Link 2: https://arxiv.org/abs/1611.01144 116 117Linear functions 118---------------- 119 120.. autosummary:: 121 :toctree: generated 122 :nosignatures: 123 124 linear 125 bilinear 126 127Dropout functions 128----------------- 129 130.. autosummary:: 131 :toctree: generated 132 :nosignatures: 133 134 dropout 135 alpha_dropout 136 feature_alpha_dropout 137 dropout1d 138 dropout2d 139 dropout3d 140 141Sparse functions 142---------------------------------- 143 144.. autosummary:: 145 :toctree: generated 146 :nosignatures: 147 148 embedding 149 embedding_bag 150 one_hot 151 152Distance functions 153---------------------------------- 154 155.. autosummary:: 156 :toctree: generated 157 :nosignatures: 158 159 pairwise_distance 160 cosine_similarity 161 pdist 162 163 164Loss functions 165-------------- 166 167.. autosummary:: 168 :toctree: generated 169 :nosignatures: 170 171 binary_cross_entropy 172 binary_cross_entropy_with_logits 173 poisson_nll_loss 174 cosine_embedding_loss 175 cross_entropy 176 ctc_loss 177 gaussian_nll_loss 178 hinge_embedding_loss 179 kl_div 180 l1_loss 181 mse_loss 182 margin_ranking_loss 183 multilabel_margin_loss 184 multilabel_soft_margin_loss 185 multi_margin_loss 186 nll_loss 187 huber_loss 188 smooth_l1_loss 189 soft_margin_loss 190 triplet_margin_loss 191 triplet_margin_with_distance_loss 192 193Vision functions 194---------------- 195 196.. autosummary:: 197 :toctree: generated 198 :nosignatures: 199 200 pixel_shuffle 201 pixel_unshuffle 202 pad 203 interpolate 204 upsample 205 upsample_nearest 206 upsample_bilinear 207 grid_sample 208 affine_grid 209 210DataParallel functions (multi-GPU, distributed) 211----------------------------------------------- 212 213:hidden:`data_parallel` 214~~~~~~~~~~~~~~~~~~~~~~~ 215 216.. autosummary:: 217 :toctree: generated 218 :nosignatures: 219 220 torch.nn.parallel.data_parallel 221