EmbeddingModel is responsible for creating embedding layers for different types of input data.
Initialize an EmbeddingModel by constructing embedding layers based on input processors.
Perform a forward pass by computing embeddings (or passing through) for each field.
This method takes inputs, a named list of torch_tensor objects, with names matching dataset$input_processors.
Return a concise string representation of the EmbeddingModel, listing its embedding layers.
Value
An EmbeddingModel object that inherits from BaseModel.
None (initializes fields inside the object).
A named list of torch_tensor objects after embedding (or passthrough).
A character string representation.
Details
Inherits from BaseModel. For each entry in dataset$input_processors, if the processor is a
SequenceProcessor, an embedding layer (nn_embedding) is created; if it is a TimeseriesProcessor,
a linear layer (nn_linear) is created. During the forward pass, each input tensor is moved to the
model’s device; if an embedding layer exists for that field, it is applied to the tensor. Otherwise,
the tensor is passed through unchanged.
Examples
if (FALSE) { # \dontrun{
# Assume `my_dataset` is a SampleDataset with input_processors that includes
# SequenceProcessor and/or TimeseriesProcessor instances.
model <- EmbeddingModel(dataset = my_dataset, embedding_dim = 128)
# Suppose `inputs` is a named list of torch tensors, e.g.:
inputs <- list(
sequence_field = torch_tensor(matrix(sample(1:100, 16), nrow = 4, ncol = 4)),
timeseries_field = torch_tensor(matrix(rnorm(20), nrow = 5, ncol = 4))
)
outputs <- model(inputs)
# `outputs` is a named list of embedded tensors:
# - For “sequence_field”, a tensor of shape (batch_size, seq_len, embedding_dim)
# - For “timeseries_field”, a tensor of shape (batch_size, embedding_dim)
} # }