Prompt Provider

class

ruprompts.prompt_provider.TensorPromptProvider

(init: typing_extensions.Literal['random', 'vocab'] = 'vocab')

Directly stores prompt embeddings as a tensor.

Parameters:

Name	Type	Description	Default
init	typing_extensions.Literal['random', 'vocab']	Initialization mode. Initializes embbeddings from random embeddings from vocabulary when set to vocab, randomly otherwise.	'vocab'

class

ruprompts.prompt_provider.LSTMPromptProvider

(hidden_dim: int = -1, input_dim: int = -1, num_lstm_layers: int = 2)

Generates prompt embeddings from LSTM and MLP.

Parameters:

Name	Type	Description	Default
hidden_dim	int	Hidden dim of LSTM. Defaults to embedding dim of backbone when set to -1.	-1
input_dim	int	Input dim of LSTM. Defaults to embedding dim of backbone when set to -1.	-1
num_lstm_layers	int	Number of LSTM layers.	2

class

ruprompts.prompt_provider.BasePromptProvider

Base class for all prompt providers.