Stuck unable to customise model

I cannot seem to figure out why I cannot input my own data into perceptilabs. From what I understand from reading the forums you can have a maximum number of 2 inputs and 1 target. I reduced the data set of interest to 2 input columns however I still am unable to press run/customise model. Does anyone have any advice/ know what might be going wrong? Thanks in advance.

Hi @Oettamus,
Welcome to the forum!

Hmm that is very strange, would you mind running the tool as “perceptilabs -v=3”, trying to do everything again and copy all the text you see?
That way we can take a look at what may have crashed or gone wrong :slight_smile:

All the best,
Robert

Thanks for the welcome!

This is the console output when I tried the same process again:

[25/May/2021 12:53:21] "POST /mixpanel/track/?ip=1&_=1621943600864 HTTP/1.1" 200 0
[25/May/2021 12:53:22] "GET /static/img/file-picker/home.svg HTTP/1.1" 304 0
[25/May/2021 12:53:22] "GET /directories/get_folder_content?path=c:/users/Matteo/Downloads&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 200 17869
[25/May/2021 12:53:23] "GET /static/img/file-picker/folder.svg HTTP/1.1" 304 0
[25/May/2021 12:53:23] "GET /static/img/file-picker/file.svg HTTP/1.1" 304 0
[25/May/2021 12:53:23] "POST /mixpanel/engage/?ip=1&_=1621943603438 HTTP/1.1" 200 0
[25/May/2021 12:53:23] "POST /mixpanel/engage/?ip=1&_=1621943603437 HTTP/1.1" 200 0
[25/May/2021 12:53:23] "GET /mixpanel/decide/?verbose=1&version=3&lib=web&token=1480b2244fdd4d821227a29e2637f922&distinct_id=matteopeluso1922%40gmail.com&ip=1&_=1621943603436 HTTP/1.1" 200 0
('c:/users/Matteo/Downloads/Untitled spreadsheet - FPCPITOTLZGGBR.csv', 4)
[25/May/2021 12:53:32] "GET /files/get_file_content?path=c:/users/Matteo/Downloads/Untitled%20spreadsheet%20-%20FPCPITOTLZGGBR.csv&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 200 115
('c:/users/Matteo/Downloads/Untitled spreadsheet - FPCPITOTLZGGBR.csv', 4)
[25/May/2021 12:53:32] "GET /files/get_file_content?path=c:/users/Matteo/Downloads/Untitled%20spreadsheet%20-%20FPCPITOTLZGGBR.csv&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 200 115
[25/May/2021 12:53:34] "GET /directories/resolved_dir?path=C:\Users\Matteo\Documents\Perceptilabs\Default&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 200 63
[25/May/2021 12:53:34] "GET /directories/get_folder_content?path=C:\Users\Matteo\Documents\Perceptilabs\Default&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 200 145
[25/May/2021 12:53:35] "GET /static/fonts/roboto-regular-webfont.736b705.woff2 HTTP/1.1" 304 0
[25/May/2021 12:53:36] "GET /directories/resolved_dir?path=C:\Users\Matteo\Documents\Perceptilabs\Default&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 200 63
[25/May/2021 12:53:36] "GET /directories/get_folder_content?path=C:\Users\Matteo\Documents\Perceptilabs\Default&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 200 145
[25/May/2021 12:54:12] "OPTIONS /models/ HTTP/1.1" 200 0
[25/May/2021 12:54:12] "HEAD /directories?path=C:\Users\Matteo\Documents\Perceptilabs\Default/Model%208&token=HqMXzM0XTk2NSiXh05ZA5QmS0Ry0A5aNk1A5guelKCI HTTP/1.1" 204 34
[25/May/2021 12:54:12] "POST /models/ HTTP/1.1" 201 240
2021-05-25 12:54:13.078789: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-05-25 12:54:13.079614: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-05-25 12:54:13.079701: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]
2021-05-25 12:54:13.080110: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set
2021-05-25 12:54:13,487 - ERROR - server.py:293 - Error in create_response (action='getModelRecommendation') (issue origin: asyncio.events:88)
Traceback (most recent call last):
  File "perceptilabs\mainInterface.py", line 290, in perceptilabs.mainInterface.Interface.create_response
  File "perceptilabs\mainInterface.py", line 424, in perceptilabs.mainInterface.Interface._create_response
  File "perceptilabs\mainInterface.py", line 600, in perceptilabs.mainInterface.Interface._create_response_model_recommendation
  File "perceptilabs\automation\utils.py", line 16, in perceptilabs.automation.utils.get_model_recommendation
  File "perceptilabs\automation\modelrecommender\base.py", line 41, in perceptilabs.automation.modelrecommender.base.ModelRecommender.get_graph
  File "perceptilabs\automation\modelrecommender\base.py", line 87, in perceptilabs.automation.modelrecommender.base.ModelRecommender._add_decoder
NotImplementedError: No decoder found for datatype 'binary'

[25/May/2021 12:54:13] "OPTIONS /models/193/ HTTP/1.1" 200 0
[25/May/2021 12:54:13] "DELETE /models/193/ HTTP/1.1" 204 0
[25/May/2021 12:54:13] "GET /projects/ HTTP/1.1" 200 266

What do you reckon?

Many thanks in advance.

It looks like it crashes when it tries to use the data type “binary”.
We have a fix for that coming out in an update later today, but until then (or if that does not fix it) you can try to use the Categorical type instead :slight_smile:

Hope that helps!

Switching to catagorical got around this issue! I encountered a related issue soon later however when trying to train a model.

 File "perceptilabs\data\pipelines\binary.py", line 31, in 
 perceptilabs.data.pipelines.binary.build_binary_pipelines.Pipeline.call 
 ValueError: Invalid Binary inputs

It seems even by getting to the customization stage, it possibly recognised the binary nature of the target and so encountered the same problem? The model looks as follows;

Hmm that is strange, it shouldn’t mix in binary at that point.
Are you able to share your dataset (or a tiny part of it) so that I can reproduce it on my system? :slight_smile:

No problem here it is;Inflation-unemployment-government.txt (977 Bytes)
Converted to txt as csv not upload-able :slight_smile:

Thank you! :slight_smile:
I’ll test it tomorrow and come back to you.

1 Like

Hi again!
I got it working like this:


(Just an example structure, you can of course change it around as you wish)

It seems like the first column defaults to Binary in the Data Wizard, which likely is what caused this issue. If you change that one to Numerical, and the last one to Categorical, does it work for you then?
I would also recommend deleting all in-between components when you get to the workspace and placing new ones, I think the pre-placed merge component has a bit of an issue in it.

1 Like

Great I’ll try this out! Thanks. I had already assigned the inputs to numerical in the data wizard so that wasn’t the issue. I think it may have been the pre placed merge.

edit: Successfully got it working! I look forward to exploring more features built into perceptilabs! I’m really curious about the GAN functionality seen in some of the youtube videos, perceptilabs has so much potential.

2 Likes