Call self as a function.
Call self as a function.
Call self as a function.
Build right-edges of bins (in DAYS) for the discrete-time hazard model.
Returns: cuts_days (np.ndarray): shape [J], right edges in days.
Discretize a tensor of observed times (in days) into bin indices for discrete-time survival.
Args: time_days (Tensor): observed times in days (event or censoring). bin_years (float): width of bins in years (e.g., 0.25 = quarter-year). min_year (float): lower bound of the window in years. max_year (float): upper bound of the window in years.
Returns: bin_idx (Tensor): indices in [0, n_bins-1] for each input time
# Yearly bins from 0–10y
cuts_days = make_cuts_days(0, 10, 1)
# edges_years = [1,2,3,...,10]
# bins = (0,1], (1,2], ..., (9,10]
# Times
t = torch.tensor([100., 365., 3700.]) # ~0.27y, 1y, ~10.1y
idx = discretize_time(t, bin_years=1, min_year=0, max_year=10)
# -> [0, 0, 9] (last one clipped at 10y)
print(cuts_days), print(idx)[ 365.25 730.5 1095.75 1461. 1826.25 2191.5 2556.75 2922. 3287.25
3652.5 ]
tensor([0, 0, 9])(None, None)
def PatchTFTSurvivalDemo(
learning_rate, # desired learning rate, initial learning rate in if one_cycle_scheduler
train_size, # the training data size (for one_cycle_scheduler=True)
batch_size, # the batch size (for one_cycle_scheduler=True)
n_gpus, # number of gpus to use
linear_probing_head, # model head to linear probe/train
preloaded_model, # loaded pretrained model to use for linear probing
evaluate_risk_years:list=[1, 5, 10],
discretize_time:bool=True, # indicator to discretize time for discrete-time survival
discrete_time_bins:float=0.25, # function to discretize time for discrete-time survival
time_range_years:tuple=(0, 15), # events after this time are censored
fine_tune:bool=False, # indicator to fine tune encoder or freeze encoder weights and perform linear probing
epochs:int=100, # number of epochs for one_cycle_scheduler
scheduler_type:str='OneCycle', optimizer_type:str='AdamW',
weight_decay:float=0.001, # weight decay for Adam optimizer
final_weight_decay:float=1e-06, # weight decay for final layer
use_weight_decay_scheduler:bool=False,
demographic_embeddings:dict={'age_bin_idx': 9, 'bmi_bin_idx': 7, 'gender': 2}, age_mlp_hidden_size:int=32,
demographics_only:bool=False, scheduler_kwargs:dict={}, transforms:NoneType=None
):
Hooks to be used in LightningModule.
def SurvivalDemo(
learning_rate, # desired learning rate, initial learning rate in if one_cycle_scheduler
train_size, # the training data size (for one_cycle_scheduler=True)
batch_size, # the batch size (for one_cycle_scheduler=True)
n_gpus, # number of gpus to use
evaluate_risk_years:list=[1, 5, 10],
discretize_time:bool=True, # indicator to discretize time for discrete-time survival
discrete_time_bins:float=0.25, # function to discretize time for discrete-time survival
time_range_years:tuple=(0, 15), # events after this time are censored
fine_tune:bool=False, # indicator to fine tune encoder or freeze encoder weights and perform linear probing
epochs:int=100, # number of epochs for one_cycle_scheduler
scheduler_type:str='OneCycle', optimizer_type:str='AdamW',
weight_decay:float=0.001, # weight decay for Adam optimizer
final_weight_decay:float=1e-06, # weight decay for final layer
use_weight_decay_scheduler:bool=False,
demographic_embeddings:dict={'age_bin_idx': 9, 'bmi_bin_idx': 7, 'gender': 2}, age_mlp_hidden_size:int=32,
scheduler_kwargs:dict={}, transforms:NoneType=None
):
Hooks to be used in LightningModule.
def PatchTFTSurvivalMomentum(
learning_rate, # desired learning rate, initial learning rate in if one_cycle_scheduler
train_size, # the training data size (for one_cycle_scheduler=True)
batch_size, # the batch size (for one_cycle_scheduler=True)
n_gpus, # number of gpus to use
linear_probing_head, # model head to linear probe/train
preloaded_model, # loaded pretrained model to use for linear probing
loss_func, evaluate_risk_years:list=[1, 5, 10],
discretize_time:bool=False, # indicator to discretize time for discrete-time survival
discrete_time_bins:float=0.25, # function to discretize time for discrete-time survival
time_range_years:tuple=(0, 15), # events after this time are censored
fine_tune:bool=False, # indicator to fine tune encoder or freeze encoder weights and perform linear probing
epochs:int=100, # number of epochs for one_cycle_scheduler
scheduler_type:str='OneCycle', momentum_steps:int=4, momentum_rate:float=0.999, optimizer_type:str='AdamW',
weight_decay:float=0.001, # weight decay for Adam optimizer
final_weight_decay:float=1e-06, # weight decay for final layer
use_weight_decay_scheduler:bool=False, demographic_predictor:bool=False, n_demographics:int=1,
demographic_predictor_hidden:int=64, demographic_adversarial_weight:float=1.0, scheduler_kwargs:dict={}
):
Hooks to be used in LightningModule.
def PatchTFTHypnogramSurvivalMomentum(
learning_rate, # desired learning rate, initial learning rate in if one_cycle_scheduler
train_size, # the training data size (for one_cycle_scheduler=True)
batch_size, # the batch size (for one_cycle_scheduler=True)
n_gpus, linear_probing_head, # model head to linear probe/train
preloaded_model, # loaded pretrained model to use for linear probing
loss_func, d_model, patch_len, hypnogram_index:int=7, hypnogram_only:bool=False, signal_only:bool=False,
zero_hypnogram:bool=False, auc_years:list=[1, 5, 10],
fine_tune:bool=False, # indicator to fine tune encoder or freeze encoder weights and perform linear probing
epochs:int=100, # number of epochs for one_cycle_scheduler
scheduler_type:str='OneCycle', momentum_steps:int=4, momentum_rate:float=0.999, optimizer_type:str='AdamW',
weight_decay:float=0.001, # weight decay for Adam optimizer
final_weight_decay:float=1e-06, # weight decay for final layer
use_weight_decay_scheduler:bool=False,
create_zero_channel_mask:bool=False, # create a zero channel mask for the encoder when the a data channel is all zeros
scheduler_kwargs:dict={}
):
Hooks to be used in LightningModule.
Base class for all neural network modules.
Your models should also subclass this class.
Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes::
import torch.nn as nn
import torch.nn.functional as F
class Model(nn.Module):
def __init__(self) -> None:
super().__init__()
self.conv1 = nn.Conv2d(1, 20, 5)
self.conv2 = nn.Conv2d(20, 20, 5)
def forward(self, x):
x = F.relu(self.conv1(x))
return F.relu(self.conv2(x))Submodules assigned in this way will be registered, and will also have their parameters converted when you call :meth:to, etc.
.. note:: As per the example above, an __init__() call to the parent class must be made before assignment on the child.
:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool
hyp = torch.randint(0, 5, (5, 101))
#print(hyp.max(), hyp.min(), hyp.shape)
#hyp[2, 10:30] = 5
hypnogram_encoder = HypnogramEncoder(patch_len=5, num_categories=6, d_model=512, y_padding_mask=5)#, y_padding_mask=5)
hypnogram_enc = hypnogram_encoder(hyp)
#hypnogram_enc2 = hypnogram_encoder(hyp)
#print(hypnogram_enc.shape, hypnogram_enc2.shape)
#torch.cat([hypnogram_enc, hypnogram_enc2], dim=1).shapetorch.Size([5, 1, 101])
torch.Size([5, 21, 1, 5])
torch.Size([5, 21, 1, 5, 128])
torch.Size([5, 21, 1, 128])
torch.Size([5, 21, 1, 512])
torch.Size([5, 1, 512, 21])
def PatchTFTSurvivalWeibull(
learning_rate, # desired learning rate, initial learning rate in if one_cycle_scheduler
train_size, # the training data size (for one_cycle_scheduler=True)
batch_size, # the batch size (for one_cycle_scheduler=True)
n_gpus, # number of gpus
linear_probing_head, # model head to linear probe/train
preloaded_model, # loaded pretrained model to use for linear probing
loss_func, auc_years:list=[1, 5, 10],
fine_tune:bool=False, # indicator to fine tune encoder or freeze encoder weights and perform linear probing
max_lr:float=0.01, # maximum learning rate for one_cycle_scheduler
epochs:int=100, # number of epochs for one_cycle_scheduler
scheduler_type:str='OneCycle', optimizer_type:str='Adam',
weight_decay:float=0.0, # weight decay for Adam optimizer
torch_model_name:str='model', # name of the pytorch model within the lightning model module, this is to remove layers (for example lightning_model.pytorch_model.head = nn.Identity())
remove_pretrain_layers:list=['head'], # layers within the lightning model or lightning model.pytorch_model to remove
create_zero_channel_mask:bool=False, # create a zero channel mask for the encoder when the a data channel is all zeros
):
Hooks to be used in LightningModule.