'loss': ['compound_mssim'],
'multicoil': [True],
'refine_smaps': [True],
}
params = [
dict(dcomp=[True], normalize_image=[False], **base_params),
# dict(dcomp=[False], normalize_image=[True], **base_params),
]
run_ids = [
'ncpdnet_sense___rfs_radial_compound_mssim_dcomp_1611913984',
'ncpdnet_sense___rfs_spiral_compound_mssim_dcomp_1611913984',
]
eval_results = eval_grid(
'nc_pdnet_mc',
# train_fun,
eval_fun,
params,
run_ids=run_ids,
# n_gpus_train=1,
# timeout_train=25,
# n_gpus_eval=1,
# n_samples_eval=100,
# checkpoints_train=3,
n_samples=100,
n_gpus=3,
)
print(eval_results)
dict(use_mixed_precision=[False], model=['unet'], **base_params),
]
run_ids = [
'ncpdnet_3d___i6_radial_stacks_mse_dcomp_1612291359',
'ncpdnet_3d___i6_spiral_stacks_mse_dcomp_1612291359',
'vnet_3d___radial_stacks_mse_dcomp_1612291357',
'vnet_3d___spiral_stacks_mse_dcomp_1612291357',
]
# eval_results = train_eval_grid(
eval_results = eval_grid(
'3d_nc_eval',
# train_ncnet_multinet,
evaluate_nc_multinet,
params,
run_ids=run_ids,
# n_gpus_train=1,
# timeout_train=40,
# n_gpus_eval=1,
# n_samples_eval=100,
timeout=20,
n_gpus=1,
params_to_ignore=['use_mixed_precision', 'scale_factor'],
# checkpoints_train=7,
# resume_checkpoint=4,
# resume_run_run_ids=run_ids,
project='fastmri4',
)
print(eval_results)
from grappa.deep.model import DeepKSpaceFiller
from grappa.evaluate.deep_evaluation import test_model
from jean_zay.submitit.general_submissions import train_eval_grid, eval_grid
job_name = 'd_grappa_eval'
metrics = dict()
param_grid = {
'model_fun': [DeepKSpaceFiller],
'model_kwargs': [{
'n_dense': 2
}, {
'n_dense': 3
}],
'distance_from_center_feat': [True, False],
'n_epochs': [1000],
'lr': [1e-3],
'instance_normalisation': [True, False],
'kernel_learning': [True, False],
}
eval_grid(
job_name,
test_model,
param_grid,
n_samples=2,
timeout=2,
n_gpus=1,
)
base_params = {
'n_epochs': [100],
'af': [4],
'acq_type': ['radial', 'spiral'],
'loss': ['compound_mssim'],
'multicoil': [True],
'refine_smaps': [True],
'brain': [True],
}
params = [
dict(dcomp=[True], normalize_image=[False], **base_params),
]
run_ids = [
'ncpdnet_sense___rfs_radial_compound_mssim_dcomp_1611913984',
'ncpdnet_sense___rfs_spiral_compound_mssim_dcomp_1611913984',
]
eval_results = eval_grid(
'nc_pdnet_mc_brain',
eval_fun,
params,
run_ids=run_ids,
n_samples=250,
n_gpus=3,
timeout=20,
project='fastmri4',
)
print(eval_results)
from jean_zay.submitit.general_submissions import train_eval_grid, eval_grid
base_params = {
'n_epochs': [100],
'af': [4],
'acq_type': ['radial', 'spiral'],
'loss': ['compound_mssim'],
'multicoil': [True],
'refine_smaps': [True],
}
params = [
dict(dcomp=[True], normalize_image=[False], **base_params),
]
run_ids = [
'ncpdnet_sense___rfs_spiral_compound_mssim_dcomp_1611913984',
'ncpdnet_sense___rfs_radial_compound_mssim_dcomp_1611913984',
]
eval_results = eval_grid(
'nc_pdnet_mc_rev',
eval_fun,
params,
run_ids=run_ids,
n_samples=100,
n_gpus=3,
project='fastmri4',
)
print(eval_results)
job_name = 'shine_classifier_cifar_large_contract'
n_gpus = 1
base_params = dict(
model_size='LARGE',
dataset='cifar',
n_gpus=n_gpus,
check_contract=True,
n_iter=500,
seed=0,
)
parameters = []
parameters += [
base_params,
dict(shine=True, **base_params),
dict(fpn=True, **base_params),
]
res_all = eval_grid(
job_name,
evaluate_classifier,
parameters,
to_grid=False,
timeout=20,
n_gpus=n_gpus,
project='shine',
params_to_ignore=['n_epochs'],
torch=True,
no_force_32=True,
)
'n_epochs': [100],
'af': [4],
'acq_type': ['radial', 'spiral'],
'loss': ['compound_mssim'],
}
params = [
dict(dcomp=[True], normalize_image=[False], **base_params),
# dict(dcomp=[False], normalize_image=[True], **base_params),
]
run_ids = [
'ncpdnet_singlecoil___radial_compound_mssim_dcomp_1610872636',
'ncpdnet_singlecoil___spiral_compound_mssim_dcomp_1610911070',
]
eval_results = eval_grid(
'nc_pdnet',
# train_fun,
eval_fun,
params,
run_ids=run_ids,
# n_gpus_train=1,
# timeout_train=100,
# n_gpus_eval=1,
# n_samples_eval=100,
n_samples=100,
n_gpus=1,
)
print(eval_results)
from fastmri_recon.evaluate.scripts.nc_dip_eval import evaluate_dip_nc as eval_fun
from jean_zay.submitit.general_submissions import eval_grid
base_params = {
'af': [4],
'acq_type': ['radial', 'spiral'],
'model_kwargs': [{}],
}
eval_results = eval_grid(
'nc_pdnet',
eval_fun,
base_params,
run_ids=None,
n_samples=100,
n_gpus=1,
)
print(eval_results)
'MWCNN_medium_1603197894',
'MWCNN_small_1603197894',
'U-net_big_1603197894',
'U-net_medium_1603197894',
'U-net_medium-ca_1603197894',
'U-net_small_1603197894',
]
eval_results = eval_grid(
run_ids,
'denoise',
# train_denoiser,
evaluate_xpdnet_denoising,
parameter_grid,
# n_samples_eval=500,
# timeout_train=20,
# n_gpus_train=1,
# timeout_eval=4,
# n_gpus_eval=1,
n_samples=200,
timeout=10,
n_gpus=1,
to_grid=False,
noise_std=1, # just for eval
)
df_results = pd.DataFrame(columns='model_name model_size psnr ssim'.split())
for (name, model_size, _, _, _, _, _), eval_res in zip(model_specs,
eval_results):
df_results = df_results.append(dict(
model_name=name,
noise_config_eval = dict(
noise_input=True,
fixed_noise=True,
noise_power_spec=noise_level / 255,
)
eval_results = eval_grid(
job_name,
# train,
evaluate,
parameter_grid,
run_ids=run_ids,
# n_samples_eval=20,
# timeout_train=20,
# n_gpus_train=1,
# timeout_eval=4,
# n_gpus_eval=1,
n_samples=n_samples_eval,
timeout=4,
n_gpus=1,
to_grid=False,
patch_size=patch_size_eval, # just for eval
batch_size=batch_size_eval, # just for eval
noise_config=noise_config_eval, # just for eval
mode='bsd68', # just for eval
project='soft_thresholding',
)
eval_res.append(eval_results)
eval_res.append(eval_results_50)
data_for_df = []
for eval_results, noise_level in zip(eval_res, noise_levels):