AutoCVWrap

class lightautoml.transformers.image.AutoCVWrap(model='efficientnet-b0', weights_path=None, cache_dir='./cache_CV', subs=None, device=torch.device, n_jobs=4, random_state=42, is_advprop=True, batch_size=128, verbose=True)[source]

Bases: lightautoml.transformers.base.LAMLTransformer

Calculate image embeddings.

property features

Features list.

Return type: List[str]
Returns: List of features names.

__init__(model='efficientnet-b0', weights_path=None, cache_dir='./cache_CV', subs=None, device=torch.device, n_jobs=4, random_state=42, is_advprop=True, batch_size=128, verbose=True)[source]

Parameters

model – Name of effnet model.
weights_path – Path to saved weights.
cache_dir – Path to cache directory or None.
subs – Subsample to fit transformer. If None - full data.
device – Torch device.
n_jobs – Number of threads for dataloader.
random_state – Random state to take subsample and set torch seed.
is_advprop – Use adversarial training.
batch_size – Batch size for embedding model.
verbose – Verbose data processing.

fit(dataset)[source]

Fit chosen transformer and create feature names.

Parameters: dataset (Union[NumpyDataset, PandasDataset]) – Pandas or Numpy dataset of text features.

transform(dataset)[source]

Transform dataset to image embeddings.

Parameters: dataset (Union[NumpyDataset, PandasDataset]) – Pandas or Numpy dataset of image paths.
Return type: NumpyDataset
Returns: Numpy dataset with image embeddings.