Do not call this function from user code. It is called by __call__(). Args: xs, Tensor instance(s) Returns:

:param cand: the candidate to evaluate :param epoch_per_model: how many resource it can use, epoch number :return:

:param cand: the candidate to evaluate :param epoch_per_model: how many resource it can use, epoch number :return:

Args: seq: [batch_size, tgt_len] Returns:

Args: seq: [batch_size, tgt_len] Returns:

get inputs and outputs from a given operator Args: op: a given operator Returns: inputs and outputs

Randomly sample some data, some class may not be sampled :param train_dataloader: torch dataLoader :param batch_size: batch_size of the d

Sample some data while guarantee example class has equal number of samples. :param train_dataloader: torch dataLoader :param num_classes:

Checks whether CP is the codepoint of a CJK character.

Checks whether `chars` is a control character.

Checks whether `chars` is a punctuation character.

Checks whether `chars` is a whitespace character.

parse onnx constatn node to numpy array Args: onnx_node (OnnxNode): a given onnx node opset_version (int): th

get all intermediate params, operators, and input info from onnx model Args: graph (Graph): the loaded ONNX graph

This involves simulate get alls core,

This involves get rank, and get jacflow

parse the inits, outputs from onnx graph Args: graph (Graph): a given onnx graph device (string): # CPU or CU

parse the parameters from onnx graph Args: graph (Graph): a given onnx graph device (string): CPU or CUDA

run a single singa operator from singa operator Args: operator (Operator): the Operator instance inputs (Tens

Splits punctuation on a piece of text.

Strips accents from a piece of text.

Adds whitespace around any CJK character.

:param model_id: :param alg_score: {alg_name1: score1, alg_name2: score2}

Returns the TestResult for the ad hoc test.

Aggregate collected weights to update server weight. Args: weights (Dict[str, List[tensor.Tensor]]): The collected weights.

Performs a single optimization step. Args: param_name(String): the name of the param param_value(Tensor): par

Performs a single optimization step. Args: param_name(String): the name of the param param_value(Tensor): par

Performs a single optimization step. Args: param_name(String): the name of the param param_value(Tensor): par

Performs a single optimization step. Args: param_name(String): the name of the param param_value(Tensor): par

Performs a single optimization step. Args: param_name(String): the name of the param param_value(Tensor): par

Performs a single optimization step. Args: param_name(String): the name of the param param_value(Tensor): par

Performs a single optimization step. Args: param_name(String): the name of the param param_value(Tensor): para

check which axes the x has been broadcasted Args: y_shape: the shape of result x_shape: the shape of x Return: a

Run the backward propagation starting at y. Args: y: a Tensor instance, usually the loss dy: a number or a Tensor instance, f

Backward propagation. Args: dys: input args consisting of only CTensors. Returns: CTensor instance(s)

Sample 0/1 for each element according to the given probability. Args: p (float): with probability p, each element is sample to 1.

Insert the new_item to rank_list, then get the rank of it. :param rank_list: 0: id, 1: score :param new_item: :return:

Close the server.

Add a random value from [-offset, offset] to each channel