Bug: ERROR: FluidMLPRegressor - Ouput tap too large - when reading a written json training

[tremblap]

Please tell us what you were doing! You can include code and files by drag and dropping them into the text area.

Attempting to load a trainned network is broken. it reads fine, but then predicting gives us:
ERROR: FluidMLPRegressor - Ouput tap too large

to reproduce: 'write' any state of MLPRegressor. instantiate a new network and read the state. try querying.

in the latest public, it works. in the current nightlies, it doesn't.

What was the expected result?

to continue to load as before

What was the actual result?

ERROR: FluidMLPRegressor - Ouput tap too large

What operating system were you using?

Mac

Operating system version

mac

FluCoMa Version

tip of nighlties

[tremblap]

oh, it seems to be only in sc... I'll confirm with pd and move accordingly

[tremblap]

with the attached file, it works in max and pd but not in SC...

[tremblap]

this code works:

----------begin_max5_patcher----------
422.3ocuTtsbBCBDF95jmBFtN0QhGpoW02iNc5fIqVbRfTfX0wQe1KrjziQq
sis2.y9yut7s6R1EGQmq1.FJ4FxcjnncwQQnjWHpMNhVw2jWxMnMZEXL7k.M
IblE1XQ8ZMTHxs0JgzRdpAzaIpFamMQAZRMe0UW2oIapbNJAK9GyZUCR1s0P
3VQoj6aOplayeTHW9fFxsgSSmMYvvDRVlekkN1uklNX3q+FWRDxtbj501GG6
WRNSdkvyta8WvcdyhEf9fmQxsFdUsgvF1CsS5m1ziQ6hRE2U0bIfKWdRzGcM
hNKaFtM5aXmc4YOzkaoOsG3G+mC+zv13gWd3O1fddIv0IDMvK5A4QWxoaVF1
TyR8qAT+WFtWT1HJFTU5dSuT6JCJ8gdPk8C6t9F6owsCwvDMic9MUzAsTH+7
Gyvz30+XUvnZz4cWr1uJQdKOEfwJjbqPIemGVvSu04yMOi9U4IfGutdMnMsl
wT3ZmqTZe3rDLTHCgSwPMrVz4eBpv0tpt0Uxaz30htYZ3UJsRU.ZYinsz5fy
kRbTQxcuEp4ANvIp38wu.PlX7kA
-----------end_max5_patcher-----------

but this doesn't in the tip of nightlies in SC but works in 1.0.9

x = FluidMLPRegressor(s).read("/Users/pa/Desktop/test.json");
b = Buffer.alloc(s,2);
c = Buffer.alloc(s,10);
x.predictPoint(b,c);

// when it works this allows to draw the predicted array
c.getn(0,10,{|x|{x.plot}.defer});

all with this file:

test.json