MCPcopy Create free account

hub / github.com/HazyResearch/butterfly / functions

Functions1,242 in github.com/HazyResearch/butterfly

⨍ Functions 1,242◇ Types & classes 260
↓ 3 callersFunctionbutterfly_multiply_untied_eval
butterfly/factor_multiply/factor_multiply.cpp:678
↓ 3 callersFunctionbutterfly_multiply_untied_forward_fast
butterfly/factor_multiply_fast/butterfly_multiply.cpp:125
↓ 3 callersFunctioncirculant_as_strided
(c, x)
cnn/models/circulant1x1conv.py:107
↓ 3 callersFunctioncirculant_fft
(c, x)
cnn/models/circulant1x1conv.py:86
↓ 3 callersFunctioncirculant_indexing
(c, x)
cnn/models/circulant1x1conv.py:93
↓ 3 callersFunctioncomplex_matmul_torch
(X, Y)
torch_butterfly/complex_utils.py:41
↓ 3 callersFunctioncomplex_reshape
(x, *shape)
torch_butterfly/complex_utils.py:146
↓ 3 callersFunctionconv1x1
1x1 convolution
cnn/models/presnet.py:37
↓ 3 callersMethodconv1x1
1x1 convolution with padding
cnn/imagenet/resnet.py:46
↓ 3 callersFunctionconv3x3
3x3 convolution with padding
cnn/models/presnet.py:32
↓ 3 callersMethodconv3x3
3x3 convolution with padding
cnn/imagenet/resnet.py:41
↓ 3 callersFunctioncp2torch
(tensor)
torch_butterfly/complex_utils.py:36
↓ 3 callersFunctiondct
Construct an nn.Module based on Butterfly that exactly performs the DCT. Parameters: n: size of the DCT. Must be a power of 2. ty
torch_butterfly/special.py:164
↓ 3 callersFunctionexpand
(num_classes, dtype, tensor)
cnn/imagenet/dataloaders.py:185
↓ 3 callersMethodforward
Parameters: input: (..., size) if real or (..., size, 2) if complex Return: output: (..., size) if real or (.
learning_transforms/butterfly_old.py:86
↓ 3 callersFunctionget_loaders
(traindir, valdir, use_val_sampler=True, min_scale=0.08)
cnn/imagenet_experiment.py:99
↓ 3 callersMethodget_permutations
(self, perm=None)
cnn/models/presnet.py:275
↓ 3 callersMethodlog_iteration_metric
(self, name, val)
cnn/imagenet/logger.py:207
↓ 3 callersFunctionmake_random_batch
(batch_size, n_numbers, prob_inc, samples_per_num)
gumbel-sinkhorn/my_sorting_train.py:18
↓ 3 callersFunctionmatrix_to_butterfly_factor
Converts a matrix to a butterfly factor B_k. Assumes that it indeed has the correct sparsity pattern.
torch_butterfly/permutation.py:151
↓ 3 callersMethodmetrics
(outs, y, len_batch=None)
convolution/tasks.py:8
↓ 3 callersFunctionops_transpose_mult_br
Fast algorithm to multiply P^T v where P is the matrix of coefficients of OPs, specified by the coefficients a, b, c, and the starting polynomials
learning_transforms/ops.py:90
↓ 3 callersMethodpolish
(self, nmaxsteps=50, patience=5, threshold=1e-10, save_to_self_model=False)
learning_transforms/training.py:74
↓ 3 callersMethodpre_process
(self, input)
torch_butterfly/butterfly.py:152
↓ 3 callersMethodrecord
(self, val, n = 1)
cnn/imagenet/logger.py:30
↓ 3 callersMethodreset
(self)
cnn/imagenet/logger.py:27
↓ 3 callersFunctionshould_backup_checkpoint
(args)
cnn/imagenet/utils.py:8
↓ 3 callersFunctiontwiddle_base2_to_base4
(twiddle, increasing_stride=True)
torch_butterfly/multiply_base4.py:44
↓ 2 callersFunctionBlock2x2DiagProduct_to_ABCDs
Convert a model of the type Block2x2DiagProduct into list of ABCDs factors, ready for butterfly_mul_np.
learning_transforms/inference.py:174
↓ 2 callersMethod__init__
(self, in_size, out_size, bias=True, complex=False, increasing_stride=True, init='randn', nbl
torch_butterfly/butterfly.py:34
↓ 2 callersMethod__init__
(self, size, share_logit=False, increasing_stride=False)
butterfly/permutation.py:19
↓ 2 callersMethod__init__
(self, dalipipeline, num_classes, one_hot)
cnn/imagenet/dataloaders.py:113
↓ 2 callersMethod__init__
(self, builder, block, layers, num_classes=1000)
cnn/imagenet/resnet.py:155
↓ 2 callersMethod__init__
(self, method='linear', **kwargs)
cnn/models/lenet.py:53
↓ 2 callersMethod__init__
(self, f)
cnn/models/layers.py:15
↓ 2 callersMethod__init__
(self, block, num_blocks, num_classes=10)
cnn/models/senet.py:80
↓ 2 callersMethod__init__
(self, block, nblocks, growth_rate=12, reduction=0.5, num_classes=10)
cnn/models/densenet.py:37
↓ 2 callersMethod__init__
(self, lambd)
cnn/models/resnet_original.py:46
↓ 2 callersMethod__init__
(self, in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, fused_unfold=
cnn/models/butterfly_conv.py:42
↓ 2 callersMethod__init__
(self, in_planes, out_planes, stride, groups)
cnn/models/shufflenet.py:23
↓ 2 callersMethod__init__
(self, in_planes, planes, stride=1)
cnn/models/preact_resnet.py:16
↓ 2 callersMethod__init__
(self, block, num_blocks, num_classes=10, num_structured_layers=0, structure_type='B', **kwargs)
cnn/models/resnet.py:100
↓ 2 callersMethod_downsample
(self, planes)
cnn/models/pnasnet.py:95
↓ 2 callersFunctionaccuracy
Computes the acccuracy@k for the specified values of k
cnn/imagenet_experiment.py:426
↓ 2 callersFunctionaccuracy
Computes the precision@k for the specified values of k
cnn/imagenet_amp.py:455
↓ 2 callersMethodbackward
(ctx, grad)
butterfly/complex_utils.py:71
↓ 2 callersFunctionbbt_mult_conv2d
(twiddle, input, kernel_size, padding)
butterfly/butterfly_multiply.py:626
↓ 2 callersFunctionbbt_mult_untied_torch
(twiddle, input)
butterfly/butterfly_multiply.py:377
↓ 2 callersFunctionbbt_ortho_mult_tied
(twiddle, input)
butterfly/butterfly_multiply.py:389
↓ 2 callersFunctionbenchmark_fw_bw
(fn, inputs, nrepeats=7)
torch_butterfly/benchmark_utils.py:20
↓ 2 callersFunctionbitreversal_permutation
Return the bit reversal permutation used in FFT. Parameter: n: integer, must be a power of 2. Return: perm: bit reversal permu
cnn/pdataset_utils.py:12
↓ 2 callersFunctionbn1
(planes)
cnn/models/resnet_imagenet.py:36
↓ 2 callersFunctionbutterfly3x3
(in_planes, planes, stride=1, structure_type='B', nblocks=1, param='regular')
cnn/models/resnet_imagenet.py:15
↓ 2 callersFunctionbutterfly_bbs_multiply_untied_forward_backward_fast
butterfly/factor_multiply_fast/butterfly_multiply.cpp:328
↓ 2 callersFunctionbutterfly_bbs_multiply_untied_forward_fast
butterfly/factor_multiply_fast/butterfly_multiply.cpp:295
↓ 2 callersFunctionbutterfly_factor_multiply
butterfly/factor_multiply/factor_multiply.cpp:89
↓ 2 callersFunctionbutterfly_factor_multiply_backward
butterfly/factor_multiply/factor_multiply.cpp:154
↓ 2 callersFunctionbutterfly_kronecker
Combine two butterflies of size n1 and n2 into their Kronecker product of size n1 * n2. They must both have increasing_stride=True or increasing_s
torch_butterfly/combine.py:101
↓ 2 callersFunctionbutterfly_mult_conv2d_torch
Parameters: twiddle: (nstack, log n, n/2, 2, 2) where n = c_in input: (b_in, c_in, h_in, w_in) kernel_size: int, size of
butterfly/butterfly_multiply.py:491
↓ 2 callersFunctionbutterfly_multiply
csrc/butterfly.cpp:120
↓ 2 callersFunctionbutterfly_multiply_base4_torch
(twiddle4, twiddle2, input, increasing_stride=True, output_size=None)
torch_butterfly/multiply_base4.py:7
↓ 2 callersFunctionbutterfly_multiply_untied_forward_backward_fast
butterfly/factor_multiply_fast/butterfly_multiply.cpp:202
↓ 2 callersFunctionbutterfly_multiply_untied_forward_backward_max5_fast
butterfly/factor_multiply_fast/butterfly_multiply.cpp:246
↓ 2 callersFunctionbutterfly_multiply_untied_forward_max5_fast
butterfly/factor_multiply_fast/butterfly_multiply.cpp:164
↓ 2 callersFunctionbutterfly_multiply_untied_vector_twiddle
butterfly/factor_multiply/factor_multiply.cpp:563
↓ 2 callersFunctionconv3x3
3x3 convolution with padding
cnn/models/resnet_imagenet.py:10
↓ 2 callersFunctioncupy2torch
(tensor)
butterfly/complex_utils.py:36
↓ 2 callersFunctiondst
Construct an nn.Module based on Butterfly that exactly performs the DST. Parameters: n: size of the DST. Must be a power of 2. ty
torch_butterfly/special.py:210
↓ 2 callersMethodend
(self)
cnn/imagenet/logger.py:155
↓ 2 callersFunctionfft_unitary
Construct an nn.Module based on ButterflyUnitary that exactly performs the FFT. Since it's unitary, it corresponds to normalized=True. Parame
torch_butterfly/special.py:52
↓ 2 callersFunctionflat_dist_call
(tensors, call, extra_args=None)
cnn/distributed.py:18
↓ 2 callersFunctionget_optimizer
(parameters, fp16, lr, momentum, structured_momentum, weight_decay, nesterov=False,
cnn/imagenet/training.py:85
↓ 2 callersFunctionget_size
(input)
cnn/teacher.py:60
↓ 2 callersFunctionhalf_balance
Return the permutation vector that makes the permutation vector v n//2-balanced. Directly follows the proof of Lemma G.2. Parameters:
torch_butterfly/permutation.py:176
↓ 2 callersFunctionifft_unitary
Construct an nn.Module based on ButterflyUnitary that exactly performs the iFFT. Since it's unitary, it corresponds to normalized=True. Param
torch_butterfly/special.py:125
↓ 2 callersMethoditeration_generator_wrapper
(self, gen, val = False)
cnn/imagenet/logger.py:159
↓ 2 callersFunctionkrylov_construct
(A, v, m)
learning_transforms/target_matrix.py:64
↓ 2 callersFunctionlistperm2matperm
Converts permutation list to matrix form. Args: listperm: (..., n) - tensor of list permutations of the set [n]. Return: matperm: (.
cnn/permutation_utils.py:9
↓ 2 callersMethodlog_epoch_metric
(self, name, val)
cnn/imagenet/logger.py:192
↓ 2 callersMethodlog_run_tag
(self, name, val)
cnn/imagenet/logger.py:101
↓ 2 callersMethodloss
(self)
learning_transforms/training.py:41
↓ 2 callersMethodloss
(self)
learning_transforms/training.py:64
↓ 2 callersFunctionlr_cosine_policy
(base_lr, warmup_length, epochs, epoch_length, logger=None)
cnn/imagenet/training.py:168
↓ 2 callersFunctionlr_linear_policy
(base_lr, warmup_length, epochs, epoch_length, logger=None)
cnn/imagenet/training.py:154
↓ 2 callersFunctionlr_step_policy
(base_lr, steps, decay_factor, warmup_length, epoch_length, logger=None)
cnn/imagenet/training.py:139
↓ 2 callersFunctionmake_mmap_file
(path, input_size)
cnn/teacher.py:91
↓ 2 callersMethodmixed_model_state_dict
(self, full_model_path, distilled_param_path)
cnn/mobilenet_imagenet.py:146
↓ 2 callersFunctionops_transpose_mult
Fast algorithm to multiply P^T v where P is the matrix of coefficients of OPs, specified by the coefficients a, b, c, and the starting polynomials
learning_transforms/ops.py:40
↓ 2 callersFunctionpermutation_factor_even_odd_multiply
butterfly/factor_multiply/factor_multiply.cpp:1360
↓ 2 callersFunctionpermutation_factor_even_odd_multiply_backward
butterfly/factor_multiply/factor_multiply.cpp:1423
↓ 2 callersFunctionpermutation_factor_reverse_multiply
butterfly/factor_multiply/factor_multiply.cpp:1515
↓ 2 callersFunctionpermutation_factor_reverse_multiply_backward
butterfly/factor_multiply/factor_multiply.cpp:1588
↓ 2 callersFunctionpermutation_kronecker
Combine two permutations of size n1 and n2 into their Kronecker product of size n1 * n2.
torch_butterfly/combine.py:133
↓ 2 callersFunctionpermutation_mult_single_factor_torch
Multiply by a single permutation factor. Parameters: prob: (3, ), where prob[0] is the probability of separating the even and odd indices,
butterfly/permutation_multiply.py:172
↓ 2 callersFunctionpermutation_mult_torch
Multiply by permutation factors, parameterized by the probabilities. Parameters: prob: (nsteps, 3), where prob[:, 0] is the probability of
butterfly/permutation_multiply.py:17
↓ 2 callersFunctionpl_train
(cfg, pl_module_cls, **kwargs)
convolution/pl_runner.py:9
↓ 2 callersMethodpost_process
(self, input, output)
butterfly/butterfly.py:273
↓ 2 callersMethodpreload
(self)
cnn/imagenet_experiment.py:246
↓ 2 callersMethodpreload
(self)
cnn/train_utils.py:46
↓ 2 callersFunctionremove_postfix
(text, postfix)
convolution/utils.py:7
← previousnext →101–200 of 1,242, ranked by callers