ISubGVQA.utils.misc

Misc functions, including distributed helpers.

Mostly copy-paste from torchvision references.

Classes

`SmoothedValue`	Track a series of values and provide access to smoothed values over a
`MetricLogger`
`NestedTensor`

Functions

`all_gather`(data)	Run all_gather on arbitrary picklable data (not necessarily tensors)
`reduce_dict`(input_dict[, average])	Args:
`get_sha`()
`collate_fn`(batch)
`_max_by_axis`(→ List[int])
`nested_tensor_from_tensor_list`(tensor_list)
`setup_for_distributed`(is_master)	This function disables printing when not in master process
`is_dist_avail_and_initialized`()
`get_world_size`()
`get_rank`()
`is_main_process`()
`save_on_master`(args, *kwargs)
`init_distributed_mode`(args)
`accuracy`(output, target[, topk])	Computes the precision@k for the specified values of k
`interpolate`(→ torch.Tensor)	Equivalent to nn.functional.interpolate, but with support for empty batch sizes.

Module Contents

class ISubGVQA.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.

deque

total = 0.0

count = 0

fmt = None

update(value, n=1)

synchronize_between_processes(): Warning: does not synchronize the deque!

property median

property avg

property global_avg

property max

property value

__str__()

ISubGVQA.utils.misc.all_gather(data)

Run all_gather on arbitrary picklable data (not necessarily tensors) Args:

data: any picklable object

Returns:: list[data]: list of data gathered from each rank

ISubGVQA.utils.misc.reduce_dict(input_dict, average=True)

Args:: input_dict (dict): all the values will be reduced average (bool): whether to do average or sum

Reduce the values in the dictionary from all processes so that all processes have the averaged results. Returns a dict with the same fields as input_dict, after reduction.

class ISubGVQA.utils.misc.MetricLogger(delimiter='\t')

Bases: object

meters

delimiter = '\t'

update(**kwargs)

__getattr__(attr)

__str__()

synchronize_between_processes()

add_meter(name, meter)

log_every(iterable, print_freq, header=None)

ISubGVQA.utils.misc.get_sha()

ISubGVQA.utils.misc.collate_fn(batch)

ISubGVQA.utils.misc._max_by_axis(the_list: List[List[int]]) → List[int]

ISubGVQA.utils.misc.nested_tensor_from_tensor_list(tensor_list: List[torch.Tensor])

class ISubGVQA.utils.misc.NestedTensor(tensors, mask: torch.Tensor | None)

Bases: object

tensors

mask

to(device: Device) → NestedTensor

decompose()

__repr__()

ISubGVQA.utils.misc.setup_for_distributed(is_master): This function disables printing when not in master process

ISubGVQA.utils.misc.is_dist_avail_and_initialized()

ISubGVQA.utils.misc.get_world_size()

ISubGVQA.utils.misc.get_rank()

ISubGVQA.utils.misc.is_main_process()

ISubGVQA.utils.misc.save_on_master(*args, **kwargs)

ISubGVQA.utils.misc.init_distributed_mode(args)

ISubGVQA.utils.misc.accuracy(output, target, topk=(1,)): Computes the precision@k for the specified values of k

ISubGVQA.utils.misc.interpolate(input: torch.Tensor, size: List[int] | None = None, scale_factor: float | None = None, mode: str = 'nearest', align_corners: bool | None = None) → torch.Tensor: Equivalent to nn.functional.interpolate, but with support for empty batch sizes. This will eventually be supported natively by PyTorch, and this class can go away.