According to self._device_name, generating one Cpu/Gpu/Arm Server, and setting the model config amd startup params. Args:

read sequencefile header

Find and reserve a port for all subprocesses to use.

Run postprocess stage. Args: parsed_data_dict: data returned in preprocess stage midped_data_dict: data retu

Run preprocess stage Args: parsed_data_dict: data to be pre-processed op_info_prefix: input op info

Run process stage Args: preped_data_dict: feed the data to be predicted by the model. op_info_prefix: prefi

Splits punctuation on a piece of text.

Splits punctuation on a piece of text.

Strips accents from a piece of text.

Strips accents from a piece of text.

Set in_channel of DAG Args: in_channel: input channel of DAG Returns: None

Set out_channel of DAG Args: out_channel: output channel of DAG Returns: None

Adds whitespace around any CJK character.

Adds whitespace around any CJK character.

Topological sort of DAG, creates inverted multi-layers views. Args: used_ops: op used in DAG response_op: re

This method returns the HashMap which is unpackaged from Response. @param resp Response(the grpc protobuf Response). @return HashMap<String,String> (t

Write sequence file header to the underlying file

Write sync marker to this sequence file

Add the front op of current vritual op. Args: op: one op object, may be a virtual op or not. Returns:

all_args_in_doc checks if arguments are mentioned in doc Args: node (astroid.node): the node is visiting. doc (Docstri

synchronized operation because Upper level callers of this function have already locked. 能进到此函数的task都是同类task，在该函数之前已保证了这点。

This method gives an example of asynchronous prediction whose input type is string.

Computes the area under the receiver-operater characteristic (AUC) This function computes the AUC error metric for binary classification.

Args: results: request a dict, should include: `bbox`, `im_id`, if is_bbox_normalized=True, also need `im_sh

benchmark logger

_bitmap: single map with shape (1, H, W), whose values are binarized as {0, 1}

DAGExcutor enterance function. There are 5 steps: 1._get_next_data_id: Generate an incremental ID 2._pack_channeldata: pack t

return true if m_fileLens is prefix of lens

Check whether to send data to jump ops.WhileOp needs to rewrite this interface. this function returns False default. A

Through OP time consuming and op_config_yaml to calculate the theoretical QPS, as well as the number of concurrency required

This method returns the sign of connect status. @param target String type(The server ipv4 and port) such as "192.168.10.10:8891". @return boolean (the

convert

This method convert INDArray to specified String type. @param npdata INDArray type(The input data). @return String (specified String type for python N

------------------ Conf Interface -------------------