jtyoui.neuralNetwork.paddle.ernie package¶

Submodules¶

jtyoui.neuralNetwork.paddle.ernie.run_msra.data_generator(input_str)[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.run_msra.evaluate(exe, program, reader, graph_vars)[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.run_msra.init_checkpoint(exe, init_checkpoint_path, main_program)[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.run_msra.match(words, init_st: list)[源代码]¶

抽取实体函数

参数:	words – 需要抽取的文字 init_st – 初始化参数。st()
返回:	数字列表，这些数字是在label_map_config中配置的

jtyoui.neuralNetwork.paddle.ernie.run_msra.pad_batch_data(inst, pad_idx=0, input_mask=False)[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.run_msra.pre_process_layer(out, process_cmd, dropout_rate=0.0, name='')¶

jtyoui.neuralNetwork.paddle.ernie.run_msra.prepare_batch_data(example)[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.run_msra.st(new_model_path=None, new_config=None, new_label_map_config=None) → list[源代码]¶

初始化模型，只需要加载一次即可

参数:	new_model_path – 模型路径 new_config – 模型配置参数 new_label_map_config – 模型实体映射

jtyoui.neuralNetwork.paddle.ernie.transformer_encoder.encoder(n_layer, **kwargs)[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.transformer_encoder.encoder_layer(enc_input, attn_bias, n_head, d_key, d_value, d_model, d_inner_hid, prepostprocess_dropout, attention_dropout, relu_dropout, hidden_act, preprocess_cmd='n', postprocess_cmd='da', param_initializer=None, name='')[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.transformer_encoder.multi_head_attention(queries, keys, values, attn_bias, d_key, d_value, d_model, n_head=1, dropout_rate=0.0, cache=None, param_initializer=None, name='multi_head_att')[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.transformer_encoder.positionwise_feed_forward(x, d_inner_hid, d_hid, dropout_rate, hidden_act, param_initializer=None, name='ffn')[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.transformer_encoder.post_process_layer(prev_out, out, process_cmd, dropout_rate=0.0, name='')¶

jtyoui.neuralNetwork.paddle.ernie.transformer_encoder.pre_post_process_layer(prev_out, out, process_cmd, dropout_rate=0.0, name='')[源代码]¶

jtyoui.neuralNetwork.paddle.ernie.transformer_encoder.pre_process_layer(out, process_cmd, dropout_rate=0.0, name='')¶