Model is an abstract class representing an AllenNLP model.

class allennlp.models.model.Model(vocab:, regularizer: allennlp.nn.regularizers.regularizer_applicator.RegularizerApplicator = None) → None[source]

Bases: torch.nn.modules.module.Module, allennlp.common.registrable.Registrable

This abstract class represents a model to be trained. Rather than relying completely on the Pytorch Module, we modify the output spec of forward to be a dictionary.

Models built using this API are still compatible with other pytorch models and can be used naturally as modules within other models - outputs are dictionaries, which can be unpacked and passed into other layers. One caveat to this is that if you wish to use an AllenNLP model inside a Container (such as nn.Sequential), you must interleave the models with a wrapper module which unpacks the dictionary into a list of tensors.

In order for your model to be trained using the Trainer api, the output dictionary of your Model must include a “loss” key, which will be optimised during the training process.

Finally, you can optionally implement Model.get_metrics() in order to make use of early stopping and best-model serialization based on a validation metric in Trainer.

decode(output_dict: typing.Dict[str, torch.Tensor]) → typing.Dict[str, torch.Tensor][source]

Takes the result of forward() and runs inference / decoding / whatever post-processing you need to do your model. The intent is that model.forward() should produce potentials or probabilities, and then model.decode() can take those results and run some kind of beam search or constrained inference or whatever is necessary. This does not handle all possible decoding use cases, but it at least handles simple kinds of decoding.

This method modifies the input dictionary, and also returns the same dictionary.

By default in the base class we do nothing. If your model has some special decoding step, override this method.

forward(*inputs) → typing.Dict[str, torch.Tensor][source]

Defines the forward pass of the model. In addition, to facilitate easy training, this method is designed to compute a loss function defined by a user.

The input is comprised of everything required to perform a training update, including labels - you define the signature here! It is down to the user to ensure that inference can be performed without the presence of these labels. Hence, any inputs not available at inference time should only be used inside a conditional block.

The intended sketch of this method is as follows:

def forward(self, input1, input2, targets=None):
    output1 = self.layer1(input1)
    output2 = self.layer2(input2)
    output_dict = {"output1": output1, "output2": output2}
    if targets is not None:
        # Function returning a scalar torch.Tensor, defined by the user.
        loss = self._compute_loss(output1, output2, targets)
        output_dict["loss"] = loss
    return output_dict

Tensors comprising everything needed to perform a training update, including labels, which should be optional (i.e have a default value of None). At inference time, simply pass the relevant inputs, not including the labels.

output_dict: ``Dict[str, torch.Tensor]``

The outputs from the model. In order to train a model using the Trainer api, you must provide a “loss” key pointing to a scalar torch.Tensor representing the loss to be optimized.

forward_on_instance(instance: → typing.Dict[str, numpy.ndarray][source]

Takes an Instance, which typically has raw text in it, converts that text into arrays using this model’s Vocabulary, passes those arrays through self.forward() and self.decode() (which by default does nothing) and returns the result. Before returning the result, we convert any torch.Tensors into numpy arrays and remove the batch dimension.

forward_on_instances(instances: typing.List[]) → typing.List[typing.Dict[str, numpy.ndarray]][source]

Takes a list of Instance`s, converts that text into arrays using this model's :class:`Vocabulary, passes those arrays through self.forward() and self.decode() (which by default does nothing) and returns the result. Before returning the result, we convert any torch.Tensors into numpy arrays and separate the batched output into a list of individual dicts per instance. Note that typically this will be faster on a GPU (and conditionally, on a CPU) than repeated calls to forward_on_instance().

instances : List[Instance], required

The instances to run the model on.

cuda_device : int, required

The GPU device to use. -1 means use the CPU.

A list of the models output for each instance.
classmethod from_params(vocab:, params: allennlp.common.params.Params) → allennlp.models.model.Model[source]
get_metrics(reset: bool = False) → typing.Dict[str, float][source]

Returns a dictionary of metrics. This method will be called by in order to compute and use model metrics for early stopping and model serialization. We return an empty dictionary here rather than raising as it is not required to implement metrics for a new model. A boolean reset parameter is passed, as frequently a metric accumulator will have some state which should be reset between epochs. This is also compatible with Metrics should be populated during the call to ``forward`, with the Metric handling the accumulation of the metric until this method is called.

get_parameters_for_histogram_tensorboard_logging() → typing.List[str][source]

Returns the name of model parameters used for logging histograms to tensorboard.

get_regularization_penalty() → typing.Union[float, torch.Tensor][source]

Computes the regularization penalty for the model. Returns 0 if the model was not configured to use regularization.

classmethod load(config: allennlp.common.params.Params, serialization_dir: str, weights_file: str = None, cuda_device: int = -1) → allennlp.models.model.Model[source]

Instantiates an already-trained model, based on the experiment configuration and some optional overrides.

config: Params

The configuration that was used to train the model. It should definitely have a model section, and should probably have a trainer section as well.

serialization_dir: str = None

The directory containing the serialized weights, parameters, and vocabulary of the model.

weights_file: str = None

By default we load the weights from in the serialization directory, but you can override that value here.

cuda_device: int = -1

By default we load the model on the CPU, but if you want to load it for GPU usage you can specify the id of your GPU here

model: Model

The model specified in the configuration, loaded with the serialized vocabulary and the trained weights.

allennlp.models.model.remove_pretrained_embedding_params(params: allennlp.common.params.Params)[source]