Domino Valdano created MADLIB-1455:
--------------------------------------
Summary: compile params doesn't accept double quoted strings
Key: MADLIB-1455
URL: https://issues.apache.org/jira/browse/MADLIB-1455
Project: Apache MADlib
Issue Type: Bug
Components: Deep Learning
Reporter: Domino Valdano
Fix For: v1.18.0
When I try to train with this mst table:
mnist=# select * from mst_table_mnist;
┌─────────┬──────────┬─────────────────────────────────────────────────────────────────────────────────────┬──────────────────────────┐
│ mst_key │ model_id │ compile_params │ fit_params │
├─────────┼──────────┼─────────────────────────────────────────────────────────────────────────────────────┼──────────────────────────┤
│ 4 │ 1 │ loss="categorical_crossentropy", optimizer="SGD",
metrics=["accuracy"] │ epochs=1, batch_size=200 │
│ 2 │ 1 │ loss="categorical_crossentropy", optimizer="Adam(lr=0.00001)",
metrics=["accuracy"] │ epochs=1, batch_size=100 │
│ 3 │ 1 │ loss="categorical_crossentropy", optimizer="Adam(lr=0.001)",
metrics=["accuracy"] │ epochs=1, batch_size=200 │
│ 1 │ 1 │ loss="categorical_crossentropy", optimizer="Adam(lr=0.0001)",
metrics=["accuracy"] │ epochs=1, batch_size=50 │
└─────────┴──────────┴─────────────────────────────────────────────────────────────────────────────────────┴──────────────────────────┘
I get this error message:
ERROR: XX000: spiexceptions.InternalError: plpy.Error: model_keras error:
invalid optimizer name: "Adam (plpy_elog.c:121) (seg2 127.0.0.1:6004 pid=33258)
(plpy_elog.c:121)
CONTEXT: Traceback (most recent call last):
PL/Python function "madlib_keras_fit_multiple_model", line 24, in <module>
fit_obj.fit_multiple_model()
PL/Python function "madlib_keras_fit_multiple_model", line 233, in
fit_multiple_model
PL/Python function "madlib_keras_fit_multiple_model", line 265, in
train_multiple_model
PL/Python function "madlib_keras_fit_multiple_model", line 912, in run_training
PL/Python function "madlib_keras_fit_multiple_model"
It works if I replace the double quotes with single quotes, even though these
are supposed to mean the same thing in a python expression. ie, keras's fit()
function accepts it that way but MADlib won't
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
