Functions — github.com/ajbrock/SMASH

CodeHub MCPcopy Create free account

hub / github.com/ajbrock/SMASH / functions

Functions68 in github.com/ajbrock/SMASH

⨍ Functions 68 ◇ Types & classes 23

↓ 9 callersMethod__init__

↓ 5 callersMethodsample_weights

(self, incoming, outgoing, G, ops, gate, dilation, activation, b

↓ 4 callersMethodsample_architecture

(self, budgets=None, bank_sizes=None)

↓ 3 callersFunctionfactors

↓ 2 callersFunctioncount_flops

(arch, which_dataset='cifar')

↓ 2 callersFunctionget_data_loader

(which_dataset,augment=True,validate=True,batch_size=50,fold='all',num_workers=3)

↓ 2 callersMethodlog

Assumption: no newlines in the input.

↓ 2 callersFunctionwn2d

↓ 1 callersMethod__init__

(self, depth=12, width=4, N=8, N_max=64, nClasses=100, bottleneck=4, max_bottleneck=4, depth_c

↓ 1 callersMethod__init__

(self, root, train=True, transform=None, target_transform=None, download=Fal

↓ 1 callersFunctioncount_params

↓ 1 callersFunctioneval_parser

↓ 1 callersFunctionevaluate

(SMASH, which_dataset, batch_size, seed, validate, num_random, num_perturb, num_markov,

↓ 1 callersMethodinit_from_SMASH

(self, w1x1, W, conv1, trans1, trans2, bn1, fc)

↓ 1 callersFunctionmain

↓ 1 callersFunctionmain

evaluate.py:139

↓ 1 callersFunctionprogress

Returns a generator over `items`, printing the number and percentage of items processed and the estimated remaining processing time before yi

↓ 1 callersMethodsample_budgets

↓ 1 callersFunctiontest_fn

↓ 1 callersFunctiontrain_fn

↓ 1 callersFunctiontrain_parser

↓ 1 callersFunctiontrain_test

(depth, width, N, N_max, which_dataset, bottleneck, max_bottleneck, depth_compression,

↓ 1 callersMethodupdate_lr

Method__getitem__

Args: index (int): Index Returns: tuple: (image, target) where target is index of the target class.

Method__getitem__

Args: index (int): Index Returns: tuple: (image, target) where target is index of the target class.

(self, n_in, n_out, dilation=1, batchnorm=False, preactivation=True, kernel_size=7)

(self, n_in, n_out, batchnorm=False, preactivation=True, gate_style='add_split', kernel_size=7)

(self, n_in, n_out, kernel_size=(3,3), dilation=(1,1), preactivation=True, batchnorm=False, groups=1, activati

(self, n_in, n_bottle, n_out, ops, gate, dilation=[(1,1)]*4, activation=[F.relu]*4, kernel_size=[(3,3)]*4, gro

(self, nChannels, nOutChannels)

(self, nChannels, nOutChannels)

(self, n_in,n_out, dilation)

(self, arch, nClasses=100, var_ks=True, var_op=False, big_op=False, op_bn=Fa

(self, root, train=True, transform=None, target_transform=None, download=Fal

(self, root, train=True, transform=None, target_transform=None, download=Fal

(self, root, train=True, transform=None, target_transform=None, download=Fal

(self, fname, reinitialize=False)

Functionarch_grads

(net, arch, arch_g=None, c=None)

perturb_arch.py:246

Functionconstruct_arch_grads

perturb_arch.py:200

(self, input, n_out, dilation, ks = (3,3), groups=1)

(self, x, n_out, f=F.relu, dilation=(1,1), ks=[3,3], groups=1)

(self, x, n_out, op, gate, dilation, nl, ks=[[3,3]]*4, groups=[1]*4)

(self, x, f=F.relu)

(self, x, n_out)

(self, x, w1x1=None, incoming=None, outgoing=None, G=None, ops=None, gate=None, dilation=None,

Functionjitter_chunk

Functionperturb_SGD

(net, arch, arch_g, perturb_probability=0)

perturb_arch.py:408

Functionperturb_architecture

(net, arch, perturb_probability=0.05)

perturb_arch.py:5

Functionread_records

convenience for reading back.

Functionsoftmax

Functionsoftmax

Methodupdate_lr