Create LaTeX tables from the results as csv files, and save them in a txt file.

Generate the given number of samples from the forecasted distribution. Parameters: ----------- num_samples: int

Computes sample paths by unrolling the RNN starting with a initial input and state. Parameters ---------- pa

Computes sample paths by unrolling the RNN starting with a initial input and state. Parameters ---------- pa

Computes sample paths by unrolling the RNN starting with a initial input and state. Parameters ---------- pa

Computes sample paths by unrolling the RNN starting with a initial input and state. Parameters ---------- pa

Computes sample paths by unrolling the RNN starting with a initial input and state. Parameters ---------- pa

Format lowest and second lowest values in DataFrame

Function to split the dataset into train and validation datasets. Args: dataset_name (str): name of the dataset grouped_train (Li

Trains the tMCL neural network with specific trajectories (Brownian Motion, Brownian Bridge, AR(p)). :param model: Instance of the tMCL m

Validates a Forecast object by checking it for `NaN` values. The supplied quantiles and mean (if available) are checked. Parameters

Compute accuracy metrics for multivariate forecasts. Parameters ---------- ts_iterator iterator over tar

Initializes a multi-GPU-friendly python command line logger that logs on all processes with their rank prefixed in the log message. :

Args: input_size -- scalar; dim of inputs hidden_size -- scalar; dim of hidden layers n_hidden -- scalar;

Parameters ---------- quantiles list of strings of the form 'p10' or floats in [0, 1] with the quant

forecast_steps: number of future time steps to predict. context_length: number of observations used as context (optional, used for so

context_length: number of observations used as context. forecast_steps: number of future time steps to predict.

Parameters: ----------- num_series: int Number of series of the data which will be sent to the model. mod

A PytorchTS wrapper for TACTiS Parameters: ----------- model_parameters: Dict[str, Any] The parameters t

Parameters: ----------- input_dim: int Dimension of the encoded representation. attention_heads: int

A Deep Sigmoid Flow network, made of multiple Sigmoig Flow layers. This model a flexible transformation from any real values into the

Parameters: ----------- embedding_dim: int The dimension of the input and output embeddings for this encoding.

Parameters: ----------- hist_value: torch.Tensor [batch, series, time steps] Historical data which can be used in

Parameters: ----------- hist_value: torch.Tensor [batch, series, time steps] Historical data which can be used in

Parameters: ----------- attention_layers: int How many successive attention pairs of layers this will use.

Parameters: ----------- context_dim: int Size of the context (embedding created by the encoder) that will be sent

Only keep a small number of series for each of the input tensors. Which series will be kept is randomly selected for each batch. The

No filtering applied. Both `nan` and `inf` possible in aggregate metrics.

Filter all `nan` but keep `inf`. `nan` is only possible in the aggregate metric if all timeseries for a metric resulted in `nan`.

Filter all `nan` & `inf` values from `metric_per_ts`. If all metrics in a column of `metric_per_ts` are `nan` or `inf` the result will b