def prune_params(layer: nn.PReLU, idxs: list) -> nn.Module:
if layer.num_parameters == 1:
return layer
keep_idxs = list(set(range(layer.num_parameters)) - set(idxs))
layer.num_parameters = layer.num_parameters - len(idxs)
layer.weight = torch.nn.Parameter(layer.weight.data.clone()[keep_idxs])
return layer
def prune_prelu(layer: nn.PReLU,
idxs: list,
inplace: bool = True,
dry_run: bool = False):
"""Prune PReLU layers, e.g. [128] => [64] or [1] => [1] (no pruning if prelu has only 1 parameter)
Args:
layer: a PReLU layer.
idxs: pruning index.
"""
num_pruned = 0 if layer.num_parameters == 1 else len(idxs)
if dry_run:
return layer, num_pruned
if not inplace:
layer = deepcopy(layer)
if layer.num_parameters == 1: return layer, num_pruned
keep_idxs = [i for i in range(layer.num_parameters) if i not in idxs]
layer.num_parameters = layer.num_parameters - len(idxs)
layer.weight = torch.nn.Parameter(layer.weight.data.clone()[keep_idxs])
return layer, num_pruned
def prune_prelu(layer: nn.PReLU,
idxs: list,
inplace: bool = True,
dry_run: bool = False): #line:147
"" #line:153
OOO00O000O0OO0O0O = 0 if layer.num_parameters == 1 else len(
idxs) #line:154
if dry_run: #line:155
return layer, OOO00O000O0OO0O0O #line:156
if not inplace: #line:157
layer = deepcopy(layer) #line:158
if layer.num_parameters == 1: return layer, OOO00O000O0OO0O0O #line:159
OO0OOOOO0OO00O0O0 = [
OOOO00OOOOOOO00O0 for OOOO00OOOOOOO00O0 in range(layer.num_parameters)
if OOOO00OOOOOOO00O0 not in idxs
] #line:160
layer.num_parameters = layer.num_parameters - len(idxs) #line:161
layer.weight = torch.nn.Parameter(
layer.weight.data.clone()[OO0OOOOO0OO00O0O0]) #line:162
return layer, OOO00O000O0OO0O0O #line:163
def prune_prelu(OO0O00OO0O0O000O0: nn.PReLU,
OOOOO0O000O00O0OO: list,
inplace: bool = True,
dry_run: bool = False): #line:147
"" #line:153
OOOO00OOOOO000OO0 = 0 if OO0O00OO0O0O000O0.num_parameters == 1 else len(
OOOOO0O000O00O0OO) #line:154
if dry_run: #line:155
return OO0O00OO0O0O000O0, OOOO00OOOOO000OO0 #line:156
if not inplace: #line:157
OO0O00OO0O0O000O0 = deepcopy(OO0O00OO0O0O000O0) #line:158
if OO0O00OO0O0O000O0.num_parameters == 1:
return OO0O00OO0O0O000O0, OOOO00OOOOO000OO0 #line:159
OO0000O0O0OO00000 = [
OO0OOO000OOOOO000
for OO0OOO000OOOOO000 in range(OO0O00OO0O0O000O0.num_parameters)
if OO0OOO000OOOOO000 not in OOOOO0O000O00O0OO
] #line:160
OO0O00OO0O0O000O0.num_parameters = OO0O00OO0O0O000O0.num_parameters - len(
OOOOO0O000O00O0OO) #line:161
OO0O00OO0O0O000O0.weight = torch.nn.Parameter(
OO0O00OO0O0O000O0.weight.data.clone()[OO0000O0O0OO00000]) #line:162
return OO0O00OO0O0O000O0, OOOO00OOOOO000OO0 #line:163
