About metrics about disent HOT 3 CLOSED

This is my code
from disent.metrics import metric_dci, metric_mig
import torch
from torch.utils.data import DataLoader
from disent.metrics import metric_dci, metric_mig, metric_factor_vae
import torch
from torch.utils.data import DataLoader
from disent.dataset import DisentDataset
from model.β_vae import BetaVAE_H
from disent.dataset.data import DSpritesData
from disent.dataset.transform import ToImgTensorF32

def train():
device = torch.device("cuda:0" if torch.cuda.is_available() else 'cpu')
data = DSpritesData()
dataset = DisentDataset(data, transform=ToImgTensorF32(), augment=None)
dataloader = DataLoader(dataset=dataset, batch_size=32, shuffle=True)

checkpoint = torch.load('./checkpoints/beta_vae50.pt', map_location=device)
model = BetaVAE_H().to(device)
model.load_state_dict(checkpoint['model'])
model.eval()

# we cannot guarantee which device the representation is on
get_repr = lambda x: model.mlp_encoder(x.to(device))

# evaluate
return {
    **metric_dci(dataset, get_repr, num_train=20, boost_mode='sklearn'),
    **metric_mig(dataset, get_repr, num_train=20),
    **metric_factor_vae(dataset, get_repr, num_train=20),
}

a_results = train()
print('beta=4: ', a_results)

from disent.

Hi there, and thank you!

Unfortunately docs for this are sparse. I understand this is not the most ideal, would gladly accept PRs to fix this.

However, for context, the mig, dci and factor vae scores are largely based on those from https://github.com/google-research/disentanglement_lib (Default values should be similar) From what I remember without looking at the code num_train and batch_size affect the sample size of underlying data that is used to compute the metrics. Too little data and the metrics will be inaccurate, too much and processing time will be too much. Often for metrics during training I would lower these values and then do a final larger compute at the end with the default values.

from disent.

hydra config experiments metrics:
https://github.com/nmichlo/disent/tree/8f061a87076adeae8d6e5b0fa984b660cd40e026/experiment/config/metrics

actual code that selects these:

metric wrapper:

• see compute and compute_fast
  

  disent/disent/metrics/utils.py

  Lines 52 to 109 in 8f061a8

  	class Metric(Generic[T]):
  	def __init__(
  	self,
  	name: str,
  	metric_fn: T, # Callable[[...], Dict[str, Number]]
  	default_kwargs: Optional[Dict[str, Any]] = None,
  	fast_kwargs: Optional[Dict[str, Any]] = None,
  	):
  	self._name = name
  	self._orig_fn = metric_fn
  	self._metric_fn_default = wrapped_partial(self._orig_fn, **(default_kwargs if default_kwargs else {}))
  	self._metric_fn_fast = wrapped_partial(self._orig_fn, **(fast_kwargs if fast_kwargs else {}))
  	# How do we get a type hint for `__call__` so that its signature matches `T`?
  	def __call__(self, *args, **kwargs) -> Dict[str, Number]:
  	return self._metric_fn_default(*args, **kwargs)
  	@property
  	def compute(self) -> T:
  	return self._metric_fn_default
  	@property
  	def compute_fast(self) -> T:
  	return self._metric_fn_fast
  	@property
  	def unwrap(self) -> T:
  	return self._orig_fn
  	@property
  	def name(self) -> str:
  	return self._name
  	def __str__(self):
  	return f"metric-{self.name}"
  	def make_metric(
  	name: str,
  	default_kwargs: Optional[Dict[str, Any]] = None,
  	fast_kwargs: Optional[Dict[str, Any]] = None,
  	) -> Callable[[T], Union[Metric[T], T]]:
  	"""
  	Metrics should be decorated using this function to set defaults!
  	Two versions of the metric should exist.
  	1. Recommended settings
  	- This should give reliable results, but may be very slow, multiple minutes to half an
  	hour or more for some metrics depending on the underlying model, data and ground-truth factors.
  	2. Faster settings
  	- This should give a decent results, but should be decently fast, a few seconds/minutes at most.
  	This is not used for testing
  	"""
  	# `Union[Metric[T], T]` is hack to get type hint on `__call__`
  	def _wrap_fn_as_metric(metric_fn: T) -> Union[Metric[T], T]:
  	return Metric(name=name, metric_fn=metric_fn, default_kwargs=default_kwargs, fast_kwargs=fast_kwargs)
  	return _wrap_fn_as_metric

fast version kwargs:

• disent/disent/metrics/_dci.py

  Line 48 in 8f061a8

  @make_metric("dci", fast_kwargs=dict(num_train=1000, num_test=500))

• disent/disent/metrics/_mig.py

  Line 45 in 8f061a8

  @make_metric("mig", fast_kwargs=dict(num_train=2000))

• disent/disent/metrics/_factor_vae.py

  Line 47 in 8f061a8

  "factor_vae", fast_kwargs=dict(num_train=700, num_eval=350, num_variance_estimate=1000)

NOTE: kwargs for fast versions were arbitrarily chosen. The standard versions should follow kwargs from disentanglement_lib.

NOTE: batch_size is like batch size from dataset loaders, the model is often used within these metrics and is run on the GPU if possible.

from disent.