Codebook#

pydantic model vision_architectures.layers.codebook.CodebookConfig[source]#

Bases: CustomBaseModel

Show JSON schema
{
   "title": "CodebookConfig",
   "type": "object",
   "properties": {
      "num_vectors": {
         "title": "Num Vectors",
         "type": "integer"
      },
      "dim": {
         "title": "Dim",
         "type": "integer"
      },
      "revive_dead_vectors_after_n_steps": {
         "default": 100,
         "title": "Revive Dead Vectors After N Steps",
         "type": "integer"
      },
      "ema_decay": {
         "anyOf": [
            {
               "type": "number"
            },
            {
               "type": "null"
            }
         ],
         "default": 0.99,
         "title": "Ema Decay"
      }
   },
   "required": [
      "num_vectors",
      "dim"
   ]
}

Config:
  • arbitrary_types_allowed: bool = True

  • extra: str = ignore

  • validate_default: bool = True

  • validate_assignment: bool = True

  • validate_return: bool = True

Fields:
Validators:

field num_vectors: int [Required]#
Validated by:
field dim: int [Required]#
Validated by:
field revive_dead_vectors_after_n_steps: int = 100#
Validated by:
field ema_decay: float | None = 0.99#
Validated by:
class vision_architectures.layers.codebook.Codebook(config={}, use_ema=True, **kwargs)[source]#

Bases: Module, PyTorchModelHubMixin

__init__(config={}, use_ema=True, **kwargs)[source]#

Initialize internal Module state, shared by both nn.Module and ScriptModule.

calculate_perplexity(indices)#
calculate_losses(x, z)[source]#
quantize(x)[source]#
revive_dead_vectors()[source]#
forward(x, channels_first=None)[source]#

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.