[
https://issues.apache.org/jira/browse/MADLIB-1364?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Frank McQuillan closed MADLIB-1364.
-----------------------------------
Resolution: Fixed
https://github.com/apache/madlib/pull/416
> Misc message for 1.16 release
> -----------------------------
>
> Key: MADLIB-1364
> URL: https://issues.apache.org/jira/browse/MADLIB-1364
> Project: Apache MADlib
> Issue Type: Improvement
> Components: Deep Learning
> Reporter: Frank McQuillan
> Assignee: Nikhil
> Priority: Minor
> Fix For: v1.16
>
>
> (1)
> confusing error message if forgot to preprocess source table
> {code}
> SELECT madlib.madlib_keras_fit('train_lt5', -- source table (NOT
> PREPROCESSED)
> 'mnist_model', -- model output table
> 'model_arch_library', -- model arch table
> 1, -- model arch id
> $$ loss='categorical_crossentropy',
> optimizer='adadelta', metrics=['accuracy']$$, -- compile_params
> $$ batch_size=batch_size, epochs=1 $$, --
> fit_params
> 5, -- num_iterations
> 0, -- gpus_per_host
> 'test_lt5_packed', -- validation
> table
> 1 --
> metrics_compute_frequency
> );
> InternalError: (psycopg2.InternalError) plpy.Error: madlib_keras_fit error:
> Input table 'train_lt5_summary' does not exist (plpython.c:5038)
> {code}
> A better message would be:
> {code}
> InternalError: (psycopg2.InternalError) plpy.Error: madlib_keras_fit error:
> Input table 'train_lt5_summary' does not exist. Please ensure that the
> source table you specify has been preprocessed by the image preprocessor.
> (plpython.c:5038)
> {code}
> (2)
> confusing error message if forgot to preprocess validation table
> {code}
> SELECT madlib.madlib_keras_fit('train_lt5_packed', -- source table
> (YES PREPROCESSED)
> 'mnist_model', -- model output table
> 'model_arch_library', -- model arch table
> 1, -- model arch id
> $$ loss='categorical_crossentropy',
> optimizer='adadelta', metrics=['accuracy']$$, -- compile_params
> $$ batch_size=batch_size, epochs=1 $$, --
> fit_params
> 5, -- num_iterations
> 0, -- gpus_per_host
> 'test_lt5', -- validation table
> (NOT PREPROCESSED)
> 1 --
> metrics_compute_frequency
> );
> InternalError: (psycopg2.InternalError) plpy.Error: madlib_keras_fit: invalid
> independent_varname ('independent_var') for table (test_lt5).
> (plpython.c:5038)
> CONTEXT: Traceback (most recent call last):
> PL/Python function "madlib_keras_fit", line 21, in <module>
> madlib_keras.fit(**globals())
> PL/Python function "madlib_keras_fit", line 42, in wrapper
> PL/Python function "madlib_keras_fit", line 71, in fit
> PL/Python function "madlib_keras_fit", line 233, in __init__
> PL/Python function "madlib_keras_fit", line 274, in _validate_input_args
> PL/Python function "madlib_keras_fit", line 288, in
> _validate_validation_table
> PL/Python function "madlib_keras_fit", line 242, in _validate_input_table
> PL/Python function "madlib_keras_fit", line 96, in _assert
> PL/Python function "madlib_keras_fit"
> [SQL: "SELECT madlib.madlib_keras_fit('train_lt5_packed', --
> source table\n 'mnist_model', -- model
> output table\n 'model_arch_library', -- model
> arch table\n 1, -- model
> arch id\n $$ loss='categorical_crossentropy',
> optimizer='adadelta', metrics=['accuracy']$$, -- compile_params\n
> $$ batch_size=batch_size, epochs=1 $$, -- fit_params\n
> 5, -- num_iterations\n
> 0, -- gpus_per_host\n
> 'test_lt5', -- validation table\n
> 1 -- metrics_compute_frequency\n
> );"]
> {code}
> A better message would be:
> {code}
> InternalError: (psycopg2.InternalError) plpy.Error: madlib_keras_fit: invalid
> independent_varname ('independent_var') for table (test_lt5). Please ensure
> that this table has been preprocessed by the image preprocessor.
> (plpython.c:5038)
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
