nice_svhn_gen(pretrained: bool = False, overrides: object = {})NICE generator for the SVHN setup (NLL loss + .generate()).
Same bijection as nice_svhn, wrapped with the exact
maximum-likelihood objective and the ancestral sampler.
Model Size
Parameters
pretrainedbool= FalseReserved for future weight registration; currently a no-op.
**overridesobject= {}Optional
NICEConfig field overrides forwarded into the
underlying config.Returns
NICEForImageGenerationGenerator configured with the SVHN setup and any overrides.
Notes
Reference: Dinh, Krueger, and Bengio, "NICE: Non-linear Independent Components Estimation", ICLR Workshop, 2015 (arXiv:1410.8516), Figure 3. Samples live in the ZCA-whitened space; apply the inverse ZCA transform to view them as images.
Examples
>>> from lucid.models.generative.nice import nice_svhn_gen
>>> model = nice_svhn_gen().eval()
>>> model.generate(n_samples=2).samples.shape
(2, 3072)