Currently, PipelineInference creates an EntryPoint JSON graph beginning with:
{'Nodes' : [{
"Name": "TextAnalytics.TextTransform",
"Inputs": {
"Column": {
"Name": "text_tf",
"Source": [
"text"
]
},
}}]}The initial 'Nodes' should have double-quotes like "Nodes". JSON spec requires double quotes on keys.
As is, this fails a JSON.parse(), which makes the created JSON difficult to work with in NodeJS.
Source of the issue:
|
return new PipelineResultRow($"{{'Nodes' : [{graphDef.Graph.ToJsonString()}]}}", |
Currently, PipelineInference creates an EntryPoint JSON graph beginning with:
{'Nodes' : [{ "Name": "TextAnalytics.TextTransform", "Inputs": { "Column": { "Name": "text_tf", "Source": [ "text" ] }, }}]}The initial
'Nodes'should have double-quotes like"Nodes". JSON spec requires double quotes on keys.As is, this fails a JSON.parse(), which makes the created JSON difficult to work with in NodeJS.
Source of the issue:
machinelearning/src/Microsoft.ML.PipelineInference/PipelinePattern.cs
Line 276 in 9d19d0e