yangjun023
Results
2
issues of
yangjun023
1.the model use ReshapePreProcessor ``` graphConfBuilder.addVertex(reshapedForConvName, new PreprocessorVertex(new ReshapePreProcessor( new int[] {batchSize * docLength, embeddingsDim}, new int[]{batchSize, 1, docLength, embeddingsDim})), embeddedName); ``` 2. when predict,there is only one sample ,there...
question
in the code of bert model, there is no placeholder tensor, how do you known the tensor name of input_ids is "Placeholder:0"?