[MADLIB-1302] MLP with minibatching fails on postgres

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: v1.16
• Component/s: Deep Learning
• Assignee: Rahul Iyer
• Reporter: Frank McQuillan
• Created: 20/Feb/19 18:26
• Updated: 21/Feb/19 01:09
• Resolved: 21/Feb/19 00:29

Description

If I run the MLP notebook
https://github.com/apache/madlib-site/blob/asf-site/community-artifacts/MLP-mnist-v3.ipynb
with MADlib 1.15.1 it fails on Postgres in "Section 2 : Train MLP model with mini-batch"

MADlib version: 1.15.1, git revision: rc/1.15.1-rc1, cmake configuration time: Wed Oct 10 04:29:25 UTC 2018, build type: Release, build system: Darwin-17.7.0, C compiler: Clang, C++ compiler: Clang

PostgreSQL 9.6.7 on x86_64-apple-darwin16.7.0, compiled by Apple LLVM version 9.0.0 (clang-900.0.39.2), 64-bit

InternalError: (psycopg2.InternalError) TypeError: must be string, not int
CONTEXT:  Traceback (most recent call last):
  PL/Python function "mlp_classification", line 33, in <module>
    grouping_col)
  PL/Python function "mlp_classification", line 42, in wrapper
  PL/Python function "mlp_classification", line 147, in mlp
  PL/Python function "mlp_classification", line 74, in quote_literal
PL/Python function "mlp_classification"
 [SQL: "SELECT madlib.mlp_classification(\n    'mnist_train_packed',        -- Packed table from preprocessor\n    'mnist_result',              -- Destination table\n    'independent_varname',       -- Independent\n    'dependent_varname',         -- Dependent\n    ARRAY[100],                    -- Hidden layer sizes\n    'learning_rate_init=0.1,\n    n_iterations=1,\n    learning_rate_policy=const,\n    lambda=0.0001,               -- Regularization\n    tolerance=0',\n    'tanh',                      -- Activation function\n    '',                          -- No weights\n    FALSE,                       -- No warmstart\n    FALSE);                       -- Verbose"]

When I ran on Greenplum it did not fail on this query