New error in my get mlops working via dagster saga ```dagste

Question

New error in my get mlops working via dagster saga ```dagster core errors DagsterInvalidDefinitionError Input df train of op target extractor op has no way of being resolved Must provide a resolution to this input via another op graph or via a direct input value mapped from the top level graph To learn more see the docs for unconnected inputs <https docs dagster io concepts io management unconnected inputs unconnected inputs> ```

Rahul Dave · Answer

Code is here

Rahul Dave · Answer

```target extractor op = define dagstermill op name= target extractor op notebook path=file relative path file notebooks target extractor ipynb output notebook name= output target extractor outs= target Out pd DataFrame io manager key= lake io manager ins= df train In pd DataFrame input manager key= raw data input manager graph out = target GraphOut def target extractor graph df train = read train data target = target extractor op df train target = target extractor op return target local target extractor job = target extractor graph to job name= target extractor job resource defs= output notebook io manager local output notebook io manager training data current training data lake io manager local pandas parquet io manager raw data input manager local pandas csv io manager ```

Rahul Dave · Answer

So indeed I have connected an input manager to it This input manager is actually an IO manager i used this pattern succesfully in another op

Rahul Dave · Answer

Here is an example of a successful file input via an inputmanager pointing to an iomanager for the `encoders` input in this op graph combo ```transformer op = define dagstermill op name= transformer op notebook path=file relative path file notebooks transform ipynb output notebook name= output transform outs= transformed data Out pd DataFrame outs= transformed data Out pd DataFrame io manager key= lake io manager ins= df In pd DataFrame encoders In dict datatype In str ins= df In pd DataFrame encoders In dict input manager key= model input manager datatype In str ins= df In pd DataFrame input manager key= lake input manager encoders In dict input manager key= model input manager datatype In str graph out = transformed data GraphOut def transformer graph df = read data file datatype = read data type edict = read encoder file transformed data = transformer op datatype = datatype df = df encoders = edict transformed data = transformer op datatype = datatype df=df return transformed data local train transformer job = transformer graph to job name= train transformer job resource defs= output notebook io manager local output notebook io manager data file current training data data type train type encoder file encoder file model input manager local model fixedpath io manager lake io manager local pandas parquet io manager lake input manager local pandas parquet io manager2 raw data io manager local pandas csv io manager ```

Rahul Dave · Answer

Over here `encoders` also has no problem being resolved which makes the above error perplexing Meanwhile if i try and resolve df using an inputmanager rather than passing it into an argument I get the same error

claire · Answer

Thanks for flagging this Rahul I think this error is happening because our op resolution logic misses the case where a custom input type is used with an input manager key should fix this cc < chris>