def run_model_test(trainer_options,
model,
on_gpu: bool = True,
version=None,
with_hpc: bool = True):
reset_seed()
save_dir = trainer_options['default_root_dir']
# logger file to get meta
logger = get_default_logger(save_dir, version=version)
trainer_options.update(logger=logger)
if 'checkpoint_callback' not in trainer_options:
# logger file to get weights
checkpoint = init_checkpoint_callback(logger)
trainer_options.update(checkpoint_callback=checkpoint)
# fit model
trainer = Trainer(**trainer_options)
result = trainer.fit(model)
# correct result and ok accuracy
assert result == 1, 'amp + ddp model failed to complete'
# test model loading
pretrained_model = load_model_from_checkpoint(
logger, trainer.checkpoint_callback.dirpath)
# test new model accuracy
test_loaders = model.test_dataloader()
if not isinstance(test_loaders, list):
test_loaders = [test_loaders]
[
run_prediction(dataloader, pretrained_model)
for dataloader in test_loaders
]
if with_hpc:
if trainer.use_ddp or trainer.use_ddp2:
# on hpc this would work fine... but need to hack it for the purpose of the test
trainer.model = pretrained_model
trainer.optimizers, trainer.lr_schedulers, trainer.optimizer_frequencies = \
trainer.init_optimizers(pretrained_model)
# test HPC loading / saving
trainer.hpc_save(save_dir, logger)
trainer.hpc_load(save_dir, on_gpu=on_gpu)
def run_test_from_config(trainer_options):
"""Trains the default model with the given config."""
set_random_master_port()
reset_seed()
ckpt_path = trainer_options['weights_save_path']
trainer_options.update(checkpoint_callback=ModelCheckpoint(ckpt_path))
model = EvalModelTemplate()
trainer = Trainer(**trainer_options)
result = trainer.fit(model)
assert result == 1
# Horovod should be initialized following training. If not, this will raise an exception.
assert hvd.size() == 2
if trainer.global_rank > 0:
# on higher ranks the checkpoint location is unknown
# we want to test checkpointing on rank 0 only
assert not hasattr(trainer, 'ckpt_path')
assert not trainer.checkpoint_callback.best_model_path
return
# test model loading
pretrained_model = EvalModelTemplate.load_from_checkpoint(
trainer.checkpoint_callback.best_model_path)
# test new model accuracy
test_loaders = model.test_dataloader()
if not isinstance(test_loaders, list):
test_loaders = [test_loaders]
for dataloader in test_loaders:
run_prediction(dataloader, pretrained_model)
# test HPC loading / saving
trainer.hpc_save(ckpt_path, trainer.logger)
trainer.hpc_load(ckpt_path, on_gpu=args.on_gpu)
if args.on_gpu:
trainer = Trainer(gpus=1, distributed_backend='horovod', max_epochs=1)
# Test the root_gpu property
assert trainer.root_gpu == hvd.local_rank()
def run_model_test(trainer_options, model, on_gpu=True):
save_dir = trainer_options['default_save_path']
# logger file to get meta
logger = get_test_tube_logger(save_dir, False)
# logger file to get weights
checkpoint = init_checkpoint_callback(logger)
# add these to the trainer options
trainer_options['checkpoint_callback'] = checkpoint
trainer_options['logger'] = logger
# fit model
trainer = Trainer(**trainer_options)
result = trainer.fit(model)
# correct result and ok accuracy
assert result == 1, 'amp + ddp model failed to complete'
# test model loading
pretrained_model = load_model(logger, trainer.checkpoint_callback.filepath)
# test new model accuracy
test_loaders = model.test_dataloader()
if not isinstance(test_loaders, list):
test_loaders = [test_loaders]
[
run_prediction(dataloader, pretrained_model)
for dataloader in test_loaders
]
if trainer.use_ddp or trainer.use_ddp2:
# on hpc this would work fine... but need to hack it for the purpose of the test
trainer.model = pretrained_model
trainer.optimizers, trainer.lr_schedulers = pretrained_model.configure_optimizers(
)
# test HPC loading / saving
trainer.hpc_save(save_dir, logger)
trainer.hpc_load(save_dir, on_gpu=on_gpu)
def run_gpu_model_test(trainer_options, model, hparams, on_gpu=True):
save_dir = init_save_dir()
# logger file to get meta
logger = get_test_tube_logger(False)
logger.log_hyperparams(hparams)
logger.save()
# logger file to get weights
checkpoint = ModelCheckpoint(save_dir)
# add these to the trainer options
trainer_options['checkpoint_callback'] = checkpoint
trainer_options['logger'] = logger
# fit model
trainer = Trainer(**trainer_options)
result = trainer.fit(model)
# correct result and ok accuracy
assert result == 1, 'amp + ddp model failed to complete'
# test model loading
pretrained_model = load_model(logger.experiment, save_dir)
# test new model accuracy
[
run_prediction(dataloader, pretrained_model)
for dataloader in model.test_dataloader()
]
if trainer.use_ddp:
# on hpc this would work fine... but need to hack it for the purpose of the test
trainer.model = pretrained_model
trainer.optimizers, trainer.lr_schedulers = pretrained_model.configure_optimizers(
)
# test HPC loading / saving
trainer.hpc_save(save_dir, logger)
trainer.hpc_load(save_dir, on_gpu=on_gpu)
clear_save_dir()
def run_gpu_model_test(trainer_options, model, hparams, on_gpu=True):
save_dir = init_save_dir()
# exp file to get meta
exp = get_exp(False)
exp.argparse(hparams)
exp.save()
# exp file to get weights
checkpoint = ModelCheckpoint(save_dir)
# add these to the trainer options
trainer_options['checkpoint_callback'] = checkpoint
trainer_options['experiment'] = exp
# fit model
trainer = Trainer(**trainer_options)
result = trainer.fit(model)
# correct result and ok accuracy
assert result == 1, 'amp + ddp model failed to complete'
# test model loading
pretrained_model = load_model(exp, save_dir, on_gpu)
# test model preds
run_prediction(model.test_dataloader, pretrained_model)
if trainer.use_ddp:
# on hpc this would work fine... but need to hack it for the purpose of the test
trainer.model = pretrained_model
trainer.optimizers, trainer.lr_schedulers = pretrained_model.configure_optimizers(
)
# test HPC loading / saving
trainer.hpc_save(save_dir, exp)
trainer.hpc_load(save_dir, on_gpu=on_gpu)
clear_save_dir()
def test_dp_resume(tmpdir):
"""Make sure DP continues training correctly."""
hparams = EvalModelTemplate.get_default_hparams()
model = EvalModelTemplate(**hparams)
trainer_options = dict(
max_epochs=1,
gpus=2,
distributed_backend='dp',
)
# get logger
logger = tutils.get_default_logger(tmpdir)
# exp file to get weights
# logger file to get weights
checkpoint = tutils.init_checkpoint_callback(logger)
# add these to the trainer options
trainer_options['logger'] = logger
trainer_options['checkpoint_callback'] = checkpoint
# fit model
trainer = Trainer(**trainer_options)
trainer.is_slurm_managing_tasks = True
result = trainer.fit(model)
# track epoch before saving. Increment since we finished the current epoch, don't want to rerun
real_global_epoch = trainer.current_epoch + 1
# correct result and ok accuracy
assert result == 1, 'amp + dp model failed to complete'
# ---------------------------
# HPC LOAD/SAVE
# ---------------------------
# save
trainer.hpc_save(tmpdir, logger)
# init new trainer
new_logger = tutils.get_default_logger(tmpdir, version=logger.version)
trainer_options['logger'] = new_logger
trainer_options['checkpoint_callback'] = ModelCheckpoint(tmpdir)
trainer_options['limit_train_batches'] = 0.5
trainer_options['limit_val_batches'] = 0.2
trainer_options['max_epochs'] = 1
new_trainer = Trainer(**trainer_options)
# set the epoch start hook so we can predict before the model does the full training
def assert_good_acc():
assert new_trainer.current_epoch == real_global_epoch and new_trainer.current_epoch > 0
# if model and state loaded correctly, predictions will be good even though we
# haven't trained with the new loaded model
dp_model = new_trainer.model
dp_model.eval()
dataloader = trainer.train_dataloader
tutils.run_prediction(dataloader, dp_model, dp=True)
# new model
model = EvalModelTemplate(**hparams)
model.on_train_start = assert_good_acc
# fit new model which should load hpc weights
new_trainer.fit(model)
# test freeze on gpu
model.freeze()
model.unfreeze()
def test_amp_gpu_ddp_slurm_managed():
"""
Make sure DDP + AMP work
:return:
"""
if not can_run_gpu_test():
return
# simulate setting slurm flags
os.environ['MASTER_PORT'] = str(np.random.randint(12000, 19000, 1)[0])
os.environ['SLURM_LOCALID'] = str(0)
hparams = get_hparams()
model = LightningTestModel(hparams)
trainer_options = dict(show_progress_bar=True,
max_nb_epochs=1,
gpus=[0],
distributed_backend='ddp',
use_amp=True)
save_dir = init_save_dir()
# exp file to get meta
exp = get_exp(False)
exp.argparse(hparams)
exp.save()
# exp file to get weights
checkpoint = ModelCheckpoint(save_dir)
# add these to the trainer options
trainer_options['checkpoint_callback'] = checkpoint
trainer_options['experiment'] = exp
# fit model
trainer = Trainer(**trainer_options)
trainer.is_slurm_managing_tasks = True
result = trainer.fit(model)
# correct result and ok accuracy
assert result == 1, 'amp + ddp model failed to complete'
# test root model address
assert trainer.resolve_root_node_address('abc') == 'abc'
assert trainer.resolve_root_node_address('abc[23]') == 'abc23'
assert trainer.resolve_root_node_address('abc[23-24]') == 'abc23'
assert trainer.resolve_root_node_address(
'abc[23-24, 45-40, 40]') == 'abc23'
# test model loading with a map_location
map_location = 'cuda:1'
pretrained_model = load_model(exp, save_dir, True, map_location)
# test model preds
run_prediction(model.test_dataloader, pretrained_model)
if trainer.use_ddp:
# on hpc this would work fine... but need to hack it for the purpose of the test
trainer.model = pretrained_model
trainer.optimizers, trainer.lr_schedulers = pretrained_model.configure_optimizers(
)
# test HPC loading / saving
trainer.hpc_save(save_dir, exp)
trainer.hpc_load(save_dir, on_gpu=True)
# test freeze on gpu
model.freeze()
model.unfreeze()
clear_save_dir()
def test_cpu_slurm_save_load():
"""
Verify model save/load/checkpoint on CPU
:return:
"""
hparams = get_hparams()
model = LightningTestModel(hparams)
save_dir = init_save_dir()
# exp file to get meta
exp = get_exp(False)
exp.argparse(hparams)
exp.save()
version = exp.version
trainer_options = dict(max_nb_epochs=1,
experiment=exp,
checkpoint_callback=ModelCheckpoint(save_dir))
# fit model
trainer = Trainer(**trainer_options)
result = trainer.fit(model)
real_global_step = trainer.global_step
# traning complete
assert result == 1, 'amp + ddp model failed to complete'
# predict with trained model before saving
# make a prediction
for batch in model.test_dataloader:
break
x, y = batch
x = x.view(x.size(0), -1)
model.eval()
pred_before_saving = model(x)
# test HPC saving
# simulate snapshot on slurm
saved_filepath = trainer.hpc_save(save_dir, exp)
assert os.path.exists(saved_filepath)
# new exp file to get meta
exp = get_exp(False, version=version)
exp.argparse(hparams)
exp.save()
trainer_options = dict(
max_nb_epochs=1,
experiment=exp,
checkpoint_callback=ModelCheckpoint(save_dir),
)
trainer = Trainer(**trainer_options)
model = LightningTestModel(hparams)
# set the epoch start hook so we can predict before the model does the full training
def assert_pred_same():
assert trainer.global_step == real_global_step and trainer.global_step > 0
# predict with loaded model to make sure answers are the same
trainer.model.eval()
new_pred = trainer.model(x)
assert torch.all(torch.eq(pred_before_saving, new_pred)).item() == 1
model.on_epoch_start = assert_pred_same
# by calling fit again, we trigger training, loading weights from the cluster
# and our hook to predict using current model before any more weight updates
trainer.fit(model)
clear_save_dir()
def test_cpu_slurm_save_load():
"""
Verify model save/load/checkpoint on CPU
:return:
"""
hparams = get_hparams()
model = LightningTestModel(hparams)
save_dir = init_save_dir()
# exp file to get meta
exp = get_exp(False)
exp.argparse(hparams)
exp.save()
cluster_a = SlurmCluster()
trainer_options = dict(
max_nb_epochs=1,
cluster=cluster_a,
experiment=exp,
checkpoint_callback=ModelCheckpoint(save_dir)
)
# fit model
trainer = Trainer(**trainer_options)
result = trainer.fit(model)
real_global_step = trainer.global_step
# traning complete
assert result == 1, 'amp + ddp model failed to complete'
# predict with trained model before saving
# make a prediction
for batch in model.test_dataloader:
break
x, y = batch
x = x.view(x.size(0), -1)
model.eval()
pred_before_saving = model(x)
# test registering a save function
trainer.enable_auto_hpc_walltime_manager()
# test HPC saving
# simulate snapshot on slurm
saved_filepath = trainer.hpc_save(save_dir, exp)
assert os.path.exists(saved_filepath)
# wipe-out trainer and model
# retrain with not much data... this simulates picking training back up after slurm
# we want to see if the weights come back correctly
continue_tng_hparams = get_hparams(continue_training=True,
hpc_exp_number=cluster_a.hpc_exp_number)
trainer_options = dict(
max_nb_epochs=1,
cluster=SlurmCluster(continue_tng_hparams),
experiment=exp,
checkpoint_callback=ModelCheckpoint(save_dir),
)
trainer = Trainer(**trainer_options)
model = LightningTestModel(hparams)
# set the epoch start hook so we can predict before the model does the full training
def assert_pred_same():
assert trainer.global_step == real_global_step and trainer.global_step > 0
# predict with loaded model to make sure answers are the same
trainer.model.eval()
new_pred = trainer.model(x)
assert torch.all(torch.eq(pred_before_saving, new_pred)).item() == 1
model.on_epoch_start = assert_pred_same
# by calling fit again, we trigger training, loading weights from the cluster
# and our hook to predict using current model before any more weight updates
trainer.fit(model)
clear_save_dir()
def test_cpu_slurm_save_load(tmpdir):
"""Verify model save/load/checkpoint on CPU."""
hparams = EvalModelTemplate.get_default_hparams()
model = EvalModelTemplate(hparams)
# logger file to get meta
logger = tutils.get_default_logger(tmpdir)
version = logger.version
# fit model
trainer = Trainer(max_epochs=1,
logger=logger,
checkpoint_callback=ModelCheckpoint(tmpdir))
result = trainer.fit(model)
real_global_step = trainer.global_step
# traning complete
assert result == 1, 'cpu model failed to complete'
# predict with trained model before saving
# make a prediction
dataloaders = model.test_dataloader()
if not isinstance(dataloaders, list):
dataloaders = [dataloaders]
for dataloader in dataloaders:
for batch in dataloader:
break
x, y = batch
x = x.view(x.size(0), -1)
model.eval()
pred_before_saving = model(x)
# test HPC saving
# simulate snapshot on slurm
saved_filepath = trainer.hpc_save(tmpdir, logger)
assert os.path.exists(saved_filepath)
# new logger file to get meta
logger = tutils.get_default_logger(tmpdir, version=version)
trainer = Trainer(
max_epochs=1,
logger=logger,
checkpoint_callback=ModelCheckpoint(tmpdir),
)
model = EvalModelTemplate(hparams)
# set the epoch start hook so we can predict before the model does the full training
def assert_pred_same():
assert trainer.global_step == real_global_step and trainer.global_step > 0
# predict with loaded model to make sure answers are the same
trainer.model.eval()
new_pred = trainer.model(x)
assert torch.all(torch.eq(pred_before_saving, new_pred)).item() == 1
model.on_epoch_start = assert_pred_same
# by calling fit again, we trigger training, loading weights from the cluster
# and our hook to predict using current model before any more weight updates
trainer.fit(model)
def test_dp_resume():
"""
Make sure DP continues training correctly
:return:
"""
if not tutils.can_run_gpu_test():
return
tutils.reset_seed()
hparams = tutils.get_hparams()
model = LightningTestModel(hparams)
trainer_options = dict(
show_progress_bar=True,
max_nb_epochs=2,
gpus=2,
distributed_backend='dp',
)
save_dir = tutils.init_save_dir()
# get logger
logger = tutils.get_test_tube_logger(debug=False)
# exp file to get weights
# logger file to get weights
checkpoint = tutils.init_checkpoint_callback(logger)
# add these to the trainer options
trainer_options['logger'] = logger
trainer_options['checkpoint_callback'] = checkpoint
# fit model
trainer = Trainer(**trainer_options)
trainer.is_slurm_managing_tasks = True
result = trainer.fit(model)
# track epoch before saving
real_global_epoch = trainer.current_epoch
# correct result and ok accuracy
assert result == 1, 'amp + dp model failed to complete'
# ---------------------------
# HPC LOAD/SAVE
# ---------------------------
# save
trainer.hpc_save(save_dir, logger)
# init new trainer
new_logger = tutils.get_test_tube_logger(version=logger.version)
trainer_options['logger'] = new_logger
trainer_options['checkpoint_callback'] = ModelCheckpoint(save_dir)
trainer_options['train_percent_check'] = 0.2
trainer_options['val_percent_check'] = 0.2
trainer_options['max_nb_epochs'] = 1
new_trainer = Trainer(**trainer_options)
# set the epoch start hook so we can predict before the model does the full training
def assert_good_acc():
assert new_trainer.current_epoch == real_global_epoch and new_trainer.current_epoch > 0
# if model and state loaded correctly, predictions will be good even though we
# haven't trained with the new loaded model
dp_model = new_trainer.model
dp_model.eval()
dataloader = trainer.get_train_dataloader()
tutils.run_prediction(dataloader, dp_model, dp=True)
# new model
model = LightningTestModel(hparams)
model.on_sanity_check_start = assert_good_acc
# fit new model which should load hpc weights
new_trainer.fit(model)
# test freeze on gpu
model.freeze()
model.unfreeze()
tutils.clear_save_dir()
def main():
"""
Make sure DDP + AMP continue training correctly
:return:
"""
hparams = get_hparams()
model = LightningTestModel(hparams)
trainer_options = dict(
show_progress_bar=True,
max_nb_epochs=4,
gpus=2,
distributed_backend='dp',
)
save_dir = init_save_dir()
# exp file to get meta
exp = get_exp(False)
exp.argparse(hparams)
exp.save()
# exp file to get weights
checkpoint = ModelCheckpoint(save_dir)
# add these to the trainer options
trainer_options['experiment'] = exp
trainer_options['checkpoint_callback'] = checkpoint
# fit model
trainer = Trainer(**trainer_options)
trainer.is_slurm_managing_tasks = True
result = trainer.fit(model)
# track epoch before saving
real_global_epoch = trainer.current_epoch
# correct result and ok accuracy
assert result == 1, 'amp + dp model failed to complete'
# ---------------------------
# HPC LOAD/SAVE
# ---------------------------
# save
trainer.hpc_save(save_dir, exp)
# init new trainer
new_exp = get_exp(False, version=exp.version)
trainer_options['experiment'] = new_exp
trainer_options['checkpoint_callback'] = ModelCheckpoint(save_dir)
trainer_options['train_percent_check'] = 0.2
trainer_options['val_percent_check'] = 0.2
trainer_options['max_nb_epochs'] = 1
new_trainer = Trainer(**trainer_options)
# set the epoch start hook so we can predict before the model does the full training
def assert_good_acc():
assert trainer.current_epoch == real_global_epoch and trainer.current_epoch > 0
# if model and state loaded correctly, predictions will be good even though we
# haven't trained with the new loaded model
dp_model = new_trainer.model
dp_model.eval()
_ = [run_prediction(dataloader, dp_model, dp=True) for dataloader in trainer.val_dataloader]
# new model
model = LightningTestModel(hparams)
model.on_sanity_check_start = assert_good_acc
# fit new model which should load hpc weights
new_trainer.fit(model)
# test freeze on gpu
model.freeze()
model.unfreeze()
clear_save_dir()
def test_amp_gpu_ddp_slurm_managed():
"""
Make sure DDP + AMP work
:return:
"""
if not can_run_gpu_test():
return
reset_seed()
# simulate setting slurm flags
set_random_master_port()
os.environ['SLURM_LOCALID'] = str(0)
hparams = get_hparams()
model = LightningTestModel(hparams)
trainer_options = dict(show_progress_bar=True,
max_nb_epochs=1,
gpus=[0],
distributed_backend='ddp',
use_amp=True)
save_dir = init_save_dir()
# exp file to get meta
logger = get_test_tube_logger(False)
# exp file to get weights
checkpoint = init_checkpoint_callback(logger)
# add these to the trainer options
trainer_options['checkpoint_callback'] = checkpoint
trainer_options['logger'] = logger
# fit model
trainer = Trainer(**trainer_options)
trainer.is_slurm_managing_tasks = True
result = trainer.fit(model)
# correct result and ok accuracy
assert result == 1, 'amp + ddp model failed to complete'
# test root model address
assert trainer.resolve_root_node_address('abc') == 'abc'
assert trainer.resolve_root_node_address('abc[23]') == 'abc23'
assert trainer.resolve_root_node_address('abc[23-24]') == 'abc23'
assert trainer.resolve_root_node_address(
'abc[23-24, 45-40, 40]') == 'abc23'
# test model loading with a map_location
pretrained_model = load_model(logger.experiment,
trainer.checkpoint_callback.filepath)
# test model preds
[
run_prediction(dataloader, pretrained_model)
for dataloader in trainer.get_test_dataloaders()
]
if trainer.use_ddp:
# on hpc this would work fine... but need to hack it for the purpose of the test
trainer.model = pretrained_model
trainer.optimizers, trainer.lr_schedulers = pretrained_model.configure_optimizers(
)
# test HPC loading / saving
trainer.hpc_save(save_dir, logger)
trainer.hpc_load(save_dir, on_gpu=True)
# test freeze on gpu
model.freeze()
model.unfreeze()
clear_save_dir()
