def run():
model_state_path = "/home/ryan/code/nanopore_assembly/output/training_2018-8-30-14-31-58-3-242/model_checkpoint_4"
# directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_chr1_full/test" # spoa 2 pass variants excluded
directory = "/home/ryan/code/nanopore_assembly/output/chr1_800k_2500_windows/test" # spoa 2 pass arbitray region 2500 windows
file_paths = FileManager.get_all_file_paths_by_type(
parent_directory_path=directory, file_extension=".npz", sort=False)
# Architecture parameters
hidden_size = 16
input_channels = 1 # 1-dimensional signal
output_size = 5 # '-','A','C','T','G' one hot vector
n_layers = 3
# Hyperparameters
dropout_rate = 0
# Training parameters
batch_size_train = 1
checkpoint_interval = 300
data_loader = DataLoader(file_paths=file_paths,
batch_size=batch_size_train)
model = Decoder(hidden_size=hidden_size,
input_size=input_channels,
output_size=output_size,
n_layers=n_layers,
dropout_rate=dropout_rate)
model.load_state_dict(torch.load(model_state_path))
# consensus_caller = ConsensusCaller(sequence_to_index, sequence_to_float)
test(model=model, data_loader=data_loader)
def run():
model_state_path = "/home/ryan/code/nanopore_assembly/output/training_2018-9-14-14-17-3-4-257/model_checkpoint_5"
directory = "/home/ryan/code/nanopore_assembly/output/chr1_800k-1200k_standard_20width/chr1/test" # no variants, chr1:1200k 200 window test
file_paths = FileManager.get_all_file_paths_by_type(parent_directory_path=directory, file_extension=".npz", sort=False)
# Architecture parameters
hidden_size = 128
input_channels = 1 # 1-dimensional signal
output_size = 5 # '-','A','C','T','G' one hot vector
n_layers = 1
# Hyperparameters
dropout_rate = 0.1
# Training parameters
batch_size_train = 1
n_batches = 8000
checkpoint_interval = 1000
data_loader = DataLoader(file_paths=file_paths, batch_size=batch_size_train)
model = EncoderDecoder(hidden_size=hidden_size, input_size=input_channels, output_size=output_size, n_layers=n_layers, dropout_rate=dropout_rate)
model.load_state_dict(torch.load(model_state_path))
test(model=model, data_loader=data_loader)
consensus_caller = ConsensusCaller(sequence_to_float=sequence_to_float, sequence_to_index=sequence_to_index)
test_consensus(consensus_caller=consensus_caller, data_loader=data_loader)
def run():
# model_state_path = "/home/ryan/code/nanopore_assembly/output/training_2018-8-28-17-13-26-1-240/model_checkpoint_15"
# model_state_path = "/home/ryan/code/nanopore_assembly/output/training_2018-8-29-12-11-15-2-241/model_checkpoint_21"
# model_state_path = "/home/ryan/code/nanopore_assembly/output/training_2018-8-30-11-49-32-3-242/model_checkpoint_43"
# directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_chr1_full/test" # spoa 2 pass variants excluded
# directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_2018-9-4-17-30-38-1-247" # arbitrary 2500 window test region
directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_celegans_chr1_1mbp_NONRUNLENGTH_2018-9-19" # c elegans
# directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_human_chr1_1mbp_NONRUNLENGTH_2018-9-18" # human
file_paths = FileManager.get_all_file_paths_by_type(
parent_directory_path=directory, file_extension=".npz", sort=False)
# Architecture parameters
hidden_size = 16
input_channels = 1 # 1-dimensional signal
output_size = 5 # '-','A','C','T','G' one hot vector
n_layers = 3
# Hyperparameters
dropout_rate = 0
# Training parameters
batch_size_train = 1
n_batches = 5000
checkpoint_interval = 300
data_loader = DataLoader(file_paths=file_paths,
batch_size=batch_size_train)
# model = EncoderDecoder(hidden_size=hidden_size, input_size=input_channels, output_size=output_size, n_layers=n_layers, dropout_rate=dropout_rate)
# model.load_state_dict(torch.load(model_state_path))
consensus_caller = ConsensusCaller(sequence_to_index, sequence_to_float)
# test(model=model, data_loader=data_loader)
print(n_batches, len(data_loader))
test_consensus(consensus_caller=consensus_caller,
data_loader=data_loader,
n_batches=n_batches)
def run():
model_state_path = "/home/ryan/code/nanopore_assembly/output/simple_rnn_2_layer_32_hidden_celegans/model_checkpoint_10"
# directory = "/home/ryan/code/nanopore_assembly/output/chr1_800k-1200k_standard_20width/chr1/test" # no variants, chr1:1200k 200 window test
directory = "/home/ryan/code/nanopore_assembly/output/celegans_250_window_test_1m" # celegans 50 windows
file_paths = FileManager.get_all_file_paths_by_type(
parent_directory_path=directory, file_extension=".npz", sort=False)
# Architecture parameters
hidden_size = 32
input_channels = 5 # 1-dimensional signal
output_size = 5 # '-','A','C','T','G' one hot vector
n_layers = 2
# Hyperparameters
dropout_rate = 0.1
# Training parameters
batch_size_train = 1
n_batches = 8000
checkpoint_interval = 1000
data_loader = DataLoader(file_paths=file_paths,
batch_size=batch_size_train,
convert_to_frequency=True)
model = Decoder(hidden_size=hidden_size,
input_size=input_channels,
output_size=output_size,
n_layers=n_layers,
dropout_rate=dropout_rate)
model.load_state_dict(torch.load(model_state_path))
test(model=model, data_loader=data_loader)
consensus_caller = ConsensusCaller(sequence_to_float=sequence_to_float,
sequence_to_index=sequence_to_index)
test_consensus(consensus_caller=consensus_caller, data_loader=data_loader)
def run(load_model=False, model_state_path=None):
directory = "/home/ryan/code/nanopore_assembly/output/chr1_800k-1200k_standard_20width/chr1/train" # spoa 2 pass arbitray region 2500 windows
file_paths = FileManager.get_all_file_paths_by_type(
parent_directory_path=directory, file_extension=".npz", sort=False)
results_handler = ResultsHandler()
# Architecture parameters
hidden_size = 128
input_channels = 1 # 1-dimensional signal
output_size = 5 # '-','A','C','T','G' one hot vector
n_layers = 1
# Hyperparameters
learning_rate = 1e-3
weight_decay = 1e-5
dropout_rate = 0.1
# Training parameters
batch_size_train = 1
n_batches = None
checkpoint_interval = 1000
data_loader = DataLoader(file_paths=file_paths,
batch_size=batch_size_train)
model = EncoderDecoder(hidden_size=hidden_size,
input_size=input_channels,
output_size=output_size,
n_layers=n_layers,
dropout_rate=dropout_rate)
# Initialize the optimizer with above parameters
optimizer = optim.Adam(model.parameters(),
lr=learning_rate,
weight_decay=weight_decay)
# Define the loss function
# loss_fn = nn.MSELoss()
loss_fn = nn.CrossEntropyLoss()
if load_model:
# get weight parameters from saved model state
model.load_state_dict(torch.load(model_state_path))
# Train and get the resulting loss per iteration
losses = train(model=model,
data_loader=data_loader,
optimizer=optimizer,
loss_fn=loss_fn,
n_batches=n_batches,
results_handler=results_handler,
checkpoint_interval=checkpoint_interval)
# test(model=model,
# data_loader=data_loader,
# n_batches=4)
results_handler.save_model(model)
results_handler.save_plot(losses)
print(model)
def run(load_model=False, model_state_path=None):
# directory = "/home/ryan/code/nanopore_assembly/output/pileup_generation_2018-8-24-12-54-20-4-236" # poapy
# directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_2018-8-27-13-51-41-0-239" # spoa
# directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_2018-8-27-16-13-23-0-239" # spoa with 2 pass alignment
directory = "/home/ryan/code/nanopore_assembly/output/spoa_pileup_generation_chr1_full/train" # spoa 2 pass variants excluded
file_paths = FileManager.get_all_file_paths_by_type(
parent_directory_path=directory, file_extension=".npz", sort=False)
results_handler = ResultsHandler()
# Architecture parameters
hidden_size = 16
input_channels = 1 # 1-dimensional signal
output_size = 5 # '-','A','C','T','G' one hot vector
n_layers = 3
# Hyperparameters
learning_rate = 1e-3
weight_decay = 0
dropout_rate = 0
# Training parameters
batch_size_train = 1
n_batches = 8000
checkpoint_interval = 1000
data_loader = DataLoader(file_paths=file_paths,
batch_size=batch_size_train)
model = EncoderDecoder(hidden_size=hidden_size,
input_size=input_channels,
output_size=output_size,
n_layers=n_layers,
dropout_rate=dropout_rate)
# Initialize the optimizer with above parameters
optimizer = optim.Adam(model.parameters(),
lr=learning_rate,
weight_decay=weight_decay)
# Define the loss function
# loss_fn = nn.MSELoss()
loss_fn = nn.CrossEntropyLoss()
if load_model:
# get weight parameters from saved model state
model.load_state_dict(torch.load(model_state_path))
# Train and get the resulting loss per iteration
losses = train(model=model,
data_loader=data_loader,
optimizer=optimizer,
loss_fn=loss_fn,
n_batches=n_batches,
results_handler=results_handler,
checkpoint_interval=checkpoint_interval)
# test(model=model,
# data_loader=data_loader,
# n_batches=4)
results_handler.save_model(model)
results_handler.save_plot(losses)
print(model)