rime.util.score_array.VAEExpression

class rime.util.score_array.VAEExpression(op, children)[source]

op=_VAEOp, [weight (n, 2d), randn (n, d)], auto-broadcast on d-axis

Methods

`__init__`(op, children)
`apply`(op)
`as_tensor`(device)
`collate_fn`(batch)	collate by rows for pytorch dataloaders
`eval`()
`exp`()
`numpy`()
`sigmoid`()
`softplus`()
`train`()
`traverse`([op_func])
`vae_module`(module)

Attributes

static collate_fn(batch)[source]: collate by rows for pytorch dataloaders