Misc

biapy.utils.misc.setup_for_distributed(is_master)

This function disables printing when not in master process

biapy.utils.misc.is_dist_avail_and_initialized()

biapy.utils.misc.get_world_size()

biapy.utils.misc.get_rank()

biapy.utils.misc.is_main_process()

biapy.utils.misc.init_devices(args, cfg)

biapy.utils.misc.set_seed(seed=42)

Sets the seed on multiple python modules to obtain results as reproducible as possible.

Parameters:

seed (int, optional) – Seed value.

biapy.utils.misc.get_grad_norm_(parameters, norm_type: float = 2.0) → Tensor

biapy.utils.misc.save_model(cfg, jobname, epoch, model, model_without_ddp, optimizer, loss_scaler)

biapy.utils.misc.save_on_master(*args, **kwargs)

biapy.utils.misc.get_checkpoint_path(cfg, jobname)

biapy.utils.misc.load_model_checkpoint(cfg, jobname, model_without_ddp, device, optimizer=None, loss_scaler=None)

biapy.utils.misc.all_reduce_mean(x)

biapy.utils.misc.to_pytorch_format(x, axis_order, device, dtype=torch.float32)

biapy.utils.misc.to_numpy_format(x, axis_order_back)

biapy.utils.misc.time_text(t)

class biapy.utils.misc.NativeScalerWithGradNormCount

Bases: object

state_dict_key = 'amp_scaler'

state_dict()

load_state_dict(state_dict)

class biapy.utils.misc.TensorboardLogger(log_dir)

Bases: object

set_step(step=None)

update(head='scalar', step=None, **kwargs)

flush()

class biapy.utils.misc.SmoothedValue(window_size=20, fmt=None)

Bases: object

Track a series of values and provide access to smoothed values over a window or the global series average.

update(value, n=1)

synchronize_between_processes()

Warning: does not synchronize the deque!

property median

property avg

property global_avg

property max

property value

class biapy.utils.misc.MetricLogger(delimiter='\t', verbose=False)

Bases: object

update(**kwargs)

synchronize_between_processes()

add_meter(name, meter)

log_every(iterable, print_freq, header=None)