flow_matching_cifar_gen

→FlowMatchingForImageGeneration

flow_matching_cifar_gen(pretrained: bool = False, overrides: object = {})

source edit

Flow Matching generator for CIFAR-10 32 x 32.

Same field as flow_matching_cifar, wrapped with the Conditional Flow Matching objective and the sampler.

Model Size

Parameters

pretrainedbool= False

Reserved for future weight registration; currently a no-op.

**overridesobject= {}

Optional FlowMatchingConfig field overrides.

Returns

FlowMatchingForImageGeneration

Generator configured for CIFAR-10.

Notes

Reference: Lipman et al., "Flow Matching for Generative Modeling", ICLR, 2023 (arXiv:2210.02747). Reported: 6.35 FID / 2.99 bits/dim.

Examples

>>> from lucid.models.generative.flow_matching import flow_matching_cifar_gen
>>> model = flow_matching_cifar_gen(sample_size=8, base_channels=16,
...                                 channel_mult=(1, 2), num_res_blocks=1,
...                                 attention_resolutions=(),
...                                 resnet_groups=8).eval()
>>> model.generate(n_samples=2, steps=4).samples.shape
(2, 3, 8, 8)

Used by 1

lucid.models.generative.flow_matching