ludwig-ai / ludwig Goto Github PK

Low-code framework for building custom LLMs, neural networks, and other AI models

License: Apache License 2.0

Python 99.88% Dockerfile 0.12%

deep-learning deeplearning deep learning machine-learning machinelearning natural-language-processing natural-language computer-vision data-centric

ludwig's People

Contributors

Stargazers

Watchers

Forkers

shyamalschandra codeaudit chunjee eycab derekdjia oswaldoludwig codinguser 7472741 yibit zlszhonglongshen julianmartinezmansell parety roy-engineering kustomzone 0xflotus harryprince ricklentz lichao312214129 ebgaaron ideaplexus stoensin wguo123 luyifanlu hiekay luyx2412 hzitoun ebazarov emushtaq shafiahmed prakhar64 amdhacks barathwaja sangelastro sunn-e sunhawk2100 tricki zcan tspannhw yuchengwang syeef yyht shujian2015 cuulee tarsbase abiraja2004 autogyro vladimiralencar francishero mgc26 53x cnjelita jurjsorinliviu itssujeeth neuroradiology vamsi256 alzayats ieee820 freedeaths raijinspecial dpflann vash-zhou xuezhisd jdetras jaynoel mbrukman awesome-archive bharatr21 esmaeilinia awesomemachinelearning reynolddoss chaochuandea shooter2062424 paulclas jyf2100 b-xiang thecooltechguy lamsking lazyxd mehrdad-shokri gargshantanu14 francescorossi1987 carina28 iamsibasish ssameerr angelitahappy91 winnerineast dalestone danlg tarrysingh niteeshkanungo ntulip dahburj volkerh guangmo bbrodsky luism78 sundarrajan-mugunthan dvsseed marcelopalacioabreu stjordanis

ludwig's Issues

ludwig fails to work with uppercase extension names such as CSV and JSON

On my Mac I'm trying to use a file with a .CSV extension such as input.CSV. Unfortunately, ludwig fails to load such files in preprocessing since string.replace in python is case sensitive.

Steps to reproduce (provided you already have a trained model):

Create your prediction data csv file and name it prediction.CSV
Execute command ludwig predict --data_csv prediction.CSV --model_path results/experiment_run_0/model

Expected behavior:
ludwig can read the file

Actual:
An exception is thrown:

| |_  _ __| |_ __ _(_)__ _
| | || / _` \ V  V / / _` |
|_|\_,_\__,_|\_/\_/|_\__, |
                     |___/
ludwig v0.1.0 - Predict

Dataset type: generic
Dataset path: prediction.CSV
Model path: results/experiment_run_0/model
Output path: results_0

Found hdf5 with the same filename of the csv, using it instead
Loading metadata from: results/experiment_run_0/model/train_set_metadata.json
Loading data from: prediction.CSV
Traceback (most recent call last):
  File "/Users/vood/.pyenv/versions/3.6.3/bin/ludwig", line 11, in <module>
    sys.exit(main())
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/ludwig/cli.py", line 86, in main
    CLI()
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/ludwig/cli.py", line 64, in __init__
    getattr(self, args.command)()
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/ludwig/cli.py", line 73, in predict
    predict.cli(sys.argv[2:])
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/ludwig/predict.py", line 379, in cli
    full_predict(**vars(args))
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/ludwig/predict.py", line 86, in full_predict
    only_predictions
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/ludwig/data/preprocessing.py", line 660, in preprocess_for_prediction
    shuffle_training=False
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/ludwig/data/preprocessing.py", line 239, in load_data
    hdf5_data = h5py.File(hdf5_file_path, 'r')
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/h5py/_hl/files.py", line 394, in __init__
    swmr=swmr)
  File "/Users/vood/.pyenv/versions/3.6.3/lib/python3.6/site-packages/h5py/_hl/files.py", line 170, in make_fid
    fid = h5f.open(name, flags, fapl=fapl)
  File "h5py/_objects.pyx", line 54, in h5py._objects.with_phil.wrapper
  File "h5py/_objects.pyx", line 55, in h5py._objects.with_phil.wrapper
  File "h5py/h5f.pyx", line 85, in h5py.h5f.open
OSError: Unable to open file (file signature not found)```

Request: Example Docs showing training configuration

It might save others a bit of time to see an example of specifying training parameters in a YAML file. Took me a bit to figure out the role of dashes in YAML format, and that the training section of the model definition file expects a dictionary, not a list of dictionaries.

As in (note the lack of dashes under the 'training' section):

input_features:
  -
    name: image_path
    type: image
    encoder: stacked_cnn
  -
    name: some_number
    type: numerical

output_features:
  -
    name: label
    type: category

training:
    batch_size: 64

KeyError: 'item_id_char'

Hello,

My small_dataset.csv looks like this

user_id,item_id,dow,hod,category_id,cusine_id,restaurant_id,item_count
-6088859261566846925,601d4dc5-2ad,7,3,9c64431e-79f,fb8b5b7f-013,706247b3-86e,1
-4662982070704311223,af48a23f-f80,7,2,a989cb37-db5,8631ea96-a78,ea90883c-a2f,1
-4849981472186386714,bab8cfae-930,1,1,4f31b9ea-059,5aa1a680-942,d9ab8a13-380,1
-5400975689533971605,3ae9d553-282,4,0,37d510f8-745,c9aa3993-86f,0b8d021a-c5c,1
4810905585891548302,3d673534-8f8,3,1,4264559d-470,5aa1a680-942,5772716f-1f3,1
8523163246689801258,f5cab6a9-8f5,2,3,60db56cd-21e,2d470ebd-e46,4e4c60a4-bca,1
6141444640923780397,e1d46baf-89c,6,17,b7226a64-569,5aa1a680-942,910396af-1dd,1
7782246672655206426,617f0d50-e63,2,23,68f8231d-6e1,8631ea96-a78,344071ad-459,1
-3808793554476278979,56903221-9d1,5,5,79803e34-c64,e153fa17-ac6,5746c349-5e1,1

My model_definition looks like this (I defined it programatically)

INPUT_COLUMN_NAMES = "user_id, item_id, dow, hod, category_id, cusine_id, restaurant_id, item_count".split(", ")
NUMERIC_COLUMN_NAMES = ["dow", "hod", "item_count"]
OUTPUT_COLUMN_NAMES = "item_id, category_id, restaurant_id".split(", ")

model_definition = {
        "input_features": [
            {
                "name": col_name,
                "type": "numerical" if col_name in NUMERIC_COLUMN_NAMES else "text",
                "encoder": "parallel_cnn"
            } for col_name in INPUT_COLUMN_NAMES
        ],
        "output_features": [
            {
                "name": col_name,
                "type": "numerical" if col_name in NUMERIC_COLUMN_NAMES else "text",
            } for col_name in OUTPUT_COLUMN_NAMES
        ],
        "training": {
            "epochs": 10
        }
    }

When I run the train() function I get the error KeyError: 'item_id_char'

Log

INFO:root:Model name: run
INFO:root:Output path: ../ludwig_model\_run_5
INFO:root:

INFO:root:ludwig_version: '0.1.0'
INFO:root:command: 'D:/playground/ml-playground/ludwig-test/cnn.py'
INFO:root:dataset_type: 'generic'
INFO:root:random_seed: 42
INFO:root:model_definition: {   'combiner': {'type': 'concat'},
    'input_features': [   {   'encoder': 'parallel_cnn',
                              'level': 'word',
                              'name': 'user_id',
                              'tied_weights': None,
                              'type': 'text'},
                          {   'encoder': 'parallel_cnn',
                              'level': 'word',
                              'name': 'item_id',
                              'tied_weights': None,
                              'type': 'text'},
                          {   'encoder': 'parallel_cnn',
                              'name': 'dow',
                              'tied_weights': None,
                              'type': 'numerical'},
                          {   'encoder': 'parallel_cnn',
                              'name': 'hod',
                              'tied_weights': None,
                              'type': 'numerical'},
                          {   'encoder': 'parallel_cnn',
                              'level': 'word',
                              'name': 'category_id',
                              'tied_weights': None,
                              'type': 'text'},
                          {   'encoder': 'parallel_cnn',
                              'level': 'word',
                              'name': 'cusine_id',
                              'tied_weights': None,
                              'type': 'text'},
                          {   'encoder': 'parallel_cnn',
                              'level': 'word',
                              'name': 'restaurant_id',
                              'tied_weights': None,
                              'type': 'text'},
                          {   'encoder': 'parallel_cnn',
                              'name': 'item_count',
                              'tied_weights': None,
                              'type': 'numerical'}],
    'output_features': [   {   'decoder': 'generator',
                               'dependencies': [],
                               'level': 'char',
                               'loss': {   'class_distance_temperature': 0,
                                           'class_weights': 1,
                                           'type': 'softmax_cross_entropy',
                                           'weight': 1},
                               'name': 'item_id',
                               'reduce_dependencies': 'sum',
                               'reduce_input': 'sum',
                               'type': 'text',
                               'weight': 1},
                           {   'decoder': 'generator',
                               'dependencies': [],
                               'level': 'char',
                               'loss': {   'class_distance_temperature': 0,
                                           'class_weights': 1,
                                           'type': 'softmax_cross_entropy',
                                           'weight': 1},
                               'name': 'category_id',
                               'reduce_dependencies': 'sum',
                               'reduce_input': 'sum',
                               'type': 'text',
                               'weight': 1},
                           {   'decoder': 'generator',
                               'dependencies': [],
                               'level': 'char',
                               'loss': {   'class_distance_temperature': 0,
                                           'class_weights': 1,
                                           'type': 'softmax_cross_entropy',
                                           'weight': 1},
                               'name': 'restaurant_id',
                               'reduce_dependencies': 'sum',
                               'reduce_input': 'sum',
                               'type': 'text',
                               'weight': 1}],
    'preprocessing': {   'bag': {   'fill_value': '',
                                    'format': 'space',
                                    'lowercase': 10000,
                                    'missing_value_strategy': 'fill_with_const',
                                    'most_common': False},
                         'binary': {   'fill_value': 0,
                                       'missing_value_strategy': 'fill_with_const'},
                         'category': {   'fill_value': '<UNK>',
                                         'lowercase': False,
                                         'missing_value_strategy': 'fill_with_const',
                                         'most_common': 10000},
                         'force_split': False,
                         'image': {'missing_value_strategy': 'backfill'},
                         'numerical': {   'fill_value': 0,
                                          'missing_value_strategy': 'fill_with_const'},
                         'sequence': {   'fill_value': '',
                                         'format': 'space',
                                         'lowercase': False,
                                         'missing_value_strategy': 'fill_with_const',
                                         'most_common': 20000,
                                         'padding': 'right',
                                         'padding_symbol': '<PAD>',
                                         'sequence_length_limit': 256,
                                         'unknown_symbol': '<UNK>'},
                         'set': {   'fill_value': '',
                                    'format': 'space',
                                    'lowercase': False,
                                    'missing_value_strategy': 'fill_with_const',
                                    'most_common': 10000},
                         'split_probabilities': (0.7, 0.1, 0.2),
                         'stratify': None,
                         'text': {   'char_format': 'characters',
                                     'char_most_common': 70,
                                     'char_sequence_length_limit': 1024,
                                     'fill_value': '',
                                     'lowercase': True,
                                     'missing_value_strategy': 'fill_with_const',
                                     'padding': 'right',
                                     'padding_symbol': '<PAD>',
                                     'unknown_symbol': '<UNK>',
                                     'word_format': 'space_punct',
                                     'word_most_common': 20000,
                                     'word_sequence_length_limit': 256},
                         'timeseries': {   'fill_value': '',
                                           'format': 'space',
                                           'missing_value_strategy': 'fill_with_const',
                                           'padding': 'right',
                                           'padding_value': 0,
                                           'timeseries_length_limit': 256}},
    'training': {   'batch_size': 128,
                    'bucketing_field': None,
                    'decay': False,
                    'decay_rate': 0.96,
                    'decay_steps': 10000,
                    'dropout_rate': 0.0,
                    'early_stop': 3,
                    'epochs': 50,
                    'gradient_clipping': None,
                    'increase_batch_size_on_plateau': 0,
                    'increase_batch_size_on_plateau_max': 512,
                    'increase_batch_size_on_plateau_patience': 5,
                    'increase_batch_size_on_plateau_rate': 2,
                    'learning_rate': 0.001,
                    'learning_rate_warmup_epochs': 5,
                    'optimizer': {   'beta1': 0.9,
                                     'beta2': 0.999,
                                     'epsilon': 1e-08,
                                     'type': 'adam'},
                    'reduce_learning_rate_on_plateau': 0,
                    'reduce_learning_rate_on_plateau_patience': 5,
                    'reduce_learning_rate_on_plateau_rate': 0.5,
                    'regularization_lambda': 0,
                    'regularizer': 'l2',
                    'staircase': False,
                    'validation_field': 'combined',
                    'validation_measure': 'loss'}}
INFO:root:

INFO:root:Using full dataframe
INFO:root:Building dataset (it may take a while)
INFO:root:Loading NLP pipeline
D:\playground\ml-playground\ludwig-test\venv\lib\site-packages\ludwig\features\numerical_feature.py:63: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.
  np.float32).as_matrix()
INFO:root:Writing train set metadata with vocabulary
Traceback (most recent call last):
  File "D:/playground/ml-playground/ludwig-test/cnn.py", line 27, in <module>
    main()
  File "D:/playground/ml-playground/ludwig-test/cnn.py", line 23, in main
    train(model)
  File "D:/playground/ml-playground/ludwig-test/cnn.py", line 13, in train
    model.train()
  File "D:\playground\ml-playground\ludwig-test\model.py", line 63, in train
    self.model.train(data_df=self.data_frame, logging_level=logging.INFO, output_directory="../ludwig_model")
  File "D:\playground\ml-playground\ludwig-test\venv\lib\site-packages\ludwig\api.py", line 448, in train
    random_seed=random_seed)
  File "D:\playground\ml-playground\ludwig-test\venv\lib\site-packages\ludwig\data\preprocessing.py", line 561, in preprocess_for_training
    [training_set, validation_set, test_set]
  File "D:\playground\ml-playground\ludwig-test\venv\lib\site-packages\ludwig\data\preprocessing.py", line 769, in replace_text_feature_level
    feature['level']
KeyError: 'item_id_char'

error installing on win 7 - which component of MS Visual C++ is needed ?

Hello ,
when trying to install Ludwig on win 7
Python 3.6.8 used
I started with this command

pip install ludwig

it goes well for some time then it show the below error , it says requires Microsoft Visual C++ 14,so which component of it does it need?

copying cytoolz\tests\test_tlz.py -> build\lib.win-amd64-3.6\cytoolz\tests
copying cytoolz\tests\test_utils.py -> build\lib.win-amd64-3.6\cytoolz\tests
running build_ext
building 'cytoolz.dicttoolz' extension
error: Microsoft Visual C++ 14.0 is required. Get it with "Microsoft Visual C++ Build Tools": https://visualstudio.microsoft.com/downloads/

thanks
Ahmad

gpu training allocates too much memory

Maybe I missed something, but trying to get the image classification example going, i always end up with the following error just after the training starts:

Allocator (GPU_0_bfc) ran out of memory trying to allocate 10.99GiB. Current allocation summary follows.

No matter what images I use, or if I do it with resize_image in the input_features or bigger or tinier width/height, it always wants to allocate those 10.99GiB (which my GTX1080 hasnt :))

Even when using --gpu_fraction it wants those 10.99GiB (tried values from 0.1 to 0.5)

I'm a tensorfloor noob, so maybe I just got something completely wrong.

Definition:
_input_features:
-
name: image_path
type: image
encoder: stacked_cnn
resize_image: true
width: 28
height: 28

output_features:
-
name: class
type: category_

Finetune from existing model?

Its possible that I missed this in the Documentation, but is there a way to initialize from an existing model? For example: initializing an image encoder with a ResNet trained for the ImageNet task.

Thanks,
csg

Feature Request: Export models as TensorFlow models

Would it be possible to export the models as TensorFlow models. In that way we wouldn'nt need to depend on ludwig and its dependencies to just run the inference part, and additionally we could convert the models to tflite so that we run them on mobile.

May be related to #55.

API documentation shows wrong keyword argument for predict

Seems like the example here for the predict API is wrong. It says the argument is dataset_csv and dataset_df when it should be data_csv and data_df. It's correct in the later part of the docs though.

Example Docs show wrong CLI argument for specifying model file

The doc page here shows the command line argument to use for specifying a model definition file as "--model_definition", but I just learned that it should be "--model_definition_file". Please update ASAP as new users are likely to get tripped up by this (as I did). Thanks!

I installed ludwig on my python 3.7 instance but when I type ludwig as a program as shown in the example, it says "command not found"

Do I need to do some setting with the PYTHONPATH in the ~/.bash_profile and source ~/.bash_profile?

What is going on? Why can I not run it after running the commands in the getting started document for ludwig?

Thanks in advance!

TypeError: list indices must be integers or slices, not str

Hi there, thanks for launching this exciting tool! I am trying to modify the training parameters in the yaml a model_definition file and I am getting an error that the dictionary is a list and not a dictionary.

YAML File

# ludwig yaml test
input_features:
  - 
    name: image_path
    type: image
    encoder: stacked_cnn

output_features:
  - 
    name: class
    type: category
    
training:
  -
    epochs: 01

Stack Trace:

  File "/home/mendeza/Documents/uber-ludwig/py3env/lib/python3.6/site-packages/ludwig/cli.py", line 86, in main
    CLI()
  File "/home/mendeza/Documents/uber-ludwig/py3env/lib/python3.6/site-packages/ludwig/cli.py", line 64, in __init__
    getattr(self, args.command)()
  File "/home/mendeza/Documents/uber-ludwig/py3env/lib/python3.6/site-packages/ludwig/cli.py", line 67, in experiment
    experiment.cli(sys.argv[2:])
  File "/home/mendeza/Documents/uber-ludwig/py3env/lib/python3.6/site-packages/ludwig/experiment.py", line 548, in cli
    experiment(**vars(args))
  File "/home/mendeza/Documents/uber-ludwig/py3env/lib/python3.6/site-packages/ludwig/experiment.py", line 159, in experiment
    model_definition = merge_with_defaults(yaml.load(def_file))
  File "/home/mendeza/Documents/uber-ludwig/py3env/lib/python3.6/site-packages/ludwig/utils/defaults.py", line 148, in merge_with_defaults
    default_training_params[param])
  File "/home/mendeza/Documents/uber-ludwig/py3env/lib/python3.6/site-packages/ludwig/utils/misc.py", line 136, in set_default_value
    dictionary[key] = value
TypeError: list indices must be integers or slices, not str

Feature Request: Expose the models

It would a nice touch to expose the models once they are created. Perhaps Tensor Serving or Seldon could play a role in here.

install error on windows

HI using py 3.6 under windows 10 with tensorflow or tensorflow gpu got this message
i installed it yesterday on another windows machine no problem but failed on my personal one
failed with the version of this evening #74 i think
btw great job!

same thing if i try in import ludwg under jupyter notebook

D:>ludwig
Traceback (most recent call last):
File "d:\anaconda3\lib\runpy.py", line 193, in run_module_as_main
"main", mod_spec)
File "d:\anaconda3\lib\runpy.py", line 85, in run_code
exec(code, run_globals)
File "d:\Anaconda3\Scripts\ludwig.exe_main.py", line 5, in
File "d:\anaconda3\lib\site-packages\ludwig_init.py", line 16, in
from ludwig.api import LudwigModel
File "d:\anaconda3\lib\site-packages\ludwig\api.py", line 38, in
from ludwig.data.dataset import Dataset
File "d:\anaconda3\lib\site-packages\ludwig\data\dataset.py", line 17, in
import h5py
File "d:\anaconda3\lib\site-packages\h5py_init_.py", line 26, in
from . import _errors
ImportError: cannot import name '_errors'

pip install fails with tensorflow 1.13

Hello!

Can't install with pip on Ubuntu 18.10. The installation fails with the following error:

Collecting tensorflow>=1.12 (from ludwig)
  Could not find a version that satisfies the requirement tensorflow>=1.12 (from ludwig) (from versions: 1.13.0rc1, 1.13.0rc2)
No matching distribution found for tensorflow>=1.12 (from ludwig)

Could you please look into it? Tensorflow version is 1.13 GPU.

Incorrect datatype name in the Getting started guide

The datatype "numerical" is incorrectly specified as "numeric" in the getting started guide.

Subprocess does not find git on Windows

Upon first running, I get the error below. I traced this back to a problem with the way in which Windows resolves executable extensions, per the links below. I am able to resolve this for now by adding the shell=True argument to line 41 of ludwig\utils\misc.py, but this solution is not recommended. Any advice?

I am running Anaconda on Windows 10. Github Desktop and Git for Windows installed.

the Windows API function CreateProcess, used by subprocess under the hood, doesn't auto-resolve other executable extensions than .exe. On Windows, the 'git' command is really installed as git.cmd. Therefore, you should modify your example to explicitly invoke git.cmd

Traceback (most recent call last):
File "c:\anaconda3\lib\runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "c:\anaconda3\lib\runpy.py", line 85, in run_code
exec(code, run_globals)
File "C:\Anaconda3\Scripts\ludwig.exe_main.py", line 9, in
File "c:\anaconda3\lib\site-packages\ludwig\cli.py", line 86, in main
CLI()
File "c:\anaconda3\lib\site-packages\ludwig\cli.py", line 64, in init
getattr(self, args.command)()
File "c:\anaconda3\lib\site-packages\ludwig\cli.py", line 67, in experiment
experiment.cli(sys.argv[2:])
File "c:\anaconda3\lib\site-packages\ludwig\experiment.py", line 548, in cli
experiment(**vars(args))
File "c:\anaconda3\lib\site-packages\ludwig\experiment.py", line 201, in experiment
random_seed
File "c:\anaconda3\lib\site-packages\ludwig\utils\misc.py", line 43, in get_experiment_description
stdout=open(os.devnull, 'w')) == 0
File "c:\anaconda3\lib\subprocess.py", line 267, in call
with Popen(*popenargs, **kwargs) as p:
File "c:\anaconda3\lib\subprocess.py", line 709, in init
restore_signals, start_new_session)
File "c:\anaconda3\lib\subprocess.py", line 997, in _execute_child
startupinfo)
FileNotFoundError: [WinError 2] The system cannot find the file specified

Request: Provide Example CSVs

I'm not sure if I missed it somewhere, but having the csvs used in the examples would be really helpful

Feature Request: Deep Feature Synthesis

Please add support for automated feature engineering, ideally build on top of the already wonderful feature tool library.

This is especially helpful when one could just declare the target feature in the YAML file and then just declare "autofeatures=True" to let feature tool find and rank the best features for classification.

https://github.com/Featuretools/featuretools
https://blog.featurelabs.com/deep-feature-synthesis/

Similarly, for time series, please add automated feature extraction based on the tsfrsh library.
This would be beyond incredible because it would save so much time on feature engineering.

https://github.com/blue-yonder/tsfresh

Installation error on ubuntu

I tried to install it on ubuntu, details:

Ubuntu 18.04.1 LTS under Linux Subsystem / Win 10
Python 3.6.7
pip latest version (just installed)
Installed all dependencies via pip manually

I get this error when trying to install ludwig:
Collecting ludwig
Could not find a version that satisfies the requirement ludwig (from versions: )
No matching distribution found for ludwig

utf-8 encoding error on text prediction

Hi, I'm getting the following stack trace when trying to do a simple prediction that should yield a text output containing utf-8 text, in this case the symbol ♋ . I believe this comes down to csv.writer (in data_utils.py function save_csv) not playing nice with utf-8 out of the box. I'm not super familiar with using python's native csv writing though.

Traceback (most recent call last):
  File "C:\Users\xxx\AppData\Local\Programs\Python\Python36\Scripts\ludwig-script.py", line 11, in <module>
    load_entry_point('ludwig==0.1.0', 'console_scripts', 'ludwig')()
  File "c:\users\xxx\appdata\local\programs\python\python36\lib\site-packages\ludwig\cli.py", line 86, in main
    CLI()
  File "c:\users\xxxi\appdata\local\programs\python\python36\lib\site-packages\ludwig\cli.py", line 64, in __init__
    getattr(self, args.command)()
  File "c:\users\xxx\appdata\local\programs\python\python36\lib\site-packages\ludwig\cli.py", line 73, in predict
    predict.cli(sys.argv[2:])
  File "c:\users\xxx\appdata\local\programs\python\python36\lib\site-packages\ludwig\predict.py", line 379, in cli
    full_predict(**vars(args))
  File "c:\users\xxx\appdata\local\programs\python\python36\lib\site-packages\ludwig\predict.py", line 120, in full_predict
    save_prediction_outputs(postprocessed_output, experiment_dir_name)
  File "c:\users\xxx\appdata\local\programs\python\python36\lib\site-packages\ludwig\predict.py", line 210, in save_prediction_outputs
    save_csv(csv_filename.format(output_field, output_type), values)
  File "c:\users\xxx\appdata\local\programs\python\python36\lib\site-packages\ludwig\utils\data_utils.py", line 60, in save_csv
    writer.writerow(row)
  File "c:\users\xxx\appdata\local\programs\python\python36\lib\encodings\cp1252.py", line 19, in encode
    return codecs.charmap_encode(input,self.errors,encoding_table)[0]
UnicodeEncodeError: 'charmap' codec can't encode character '\u264b' in position 0: character maps to <undefined>

KeyError: 'text'

Hello,
my train csv file looks like

mbploreto:script loretoparisi$ head -n2 /root/spam_dataset.csv 
label	text
HAM	waiting waiting waiting waiting solitude stands by the window as someone said i tried hard to find you i found fake promises instead the thought behind to join the thought before i thought i was blind sometimes i feel i feel the way to live i thought i had strength to overcome these walls i thought i was wonderful memories keep together things now would you like to know how it feels to be always stuck in the past without any rest the thought behind to join the thought before i thought i was blind
SPAM	please every body click cross

so I have my configuration as string

"{input_features: [{name: text, type: text}], output_features: [{name: label, type: category}]}"

and I start training then:

ludwig train --data_csv /root/spam_dataset.csv --model_definition "{input_features: [{name: text, type: text}], output_features: [{name: label, type: category}]}"

Suddenly I get that error about the text field:

 _         _        _      
| |_  _ __| |_ __ _(_)__ _ 
| | || / _` \ V  V / / _` |
|_|\_,_\__,_|\_/\_/|_\__, |
                     |___/ 
ludwig v0.1.0 - Train

Experiment name: experiment
Model name: run
Output path: results/experiment_run_1


ludwig_version: '0.1.0'
command: ('ludwig train '
 '--data_csv /root/spam_dataset.csv --model_definition {input_features: '
 '[{name: text, type: text}], output_features: [{name: label, type: '
 'category}]}')
commit_hash: '98b82b3f56c0'
dataset_type: '/root/spam_dataset.csv'
model_definition: {   'combiner': {'type': 'concat'},
    'input_features': [   {   'encoder': 'parallel_cnn',
                              'level': 'word',
                              'name': 'text',
                              'tied_weights': None,
                              'type': 'text'}],
    'output_features': [   {   'dependencies': [],
                               'loss': {   'class_distance_temperature': 0,
                                           'class_weights': 1,
                                           'confidence_penalty': 0,
                                           'distortion': 1,
                                           'labels_smoothing': 0,
                                           'negative_samples': 0,
                                           'robust_lambda': 0,
                                           'sampler': None,
                                           'type': 'softmax_cross_entropy',
                                           'unique': False,
                                           'weight': 1},
                               'name': 'label',
                               'reduce_dependencies': 'sum',
                               'reduce_input': 'sum',
                               'top_k': 3,
                               'type': 'category'}],
    'preprocessing': {   'bag': {   'fill_value': '',
                                    'format': 'space',
                                    'lowercase': 10000,
                                    'missing_value_strategy': 'fill_with_const',
                                    'most_common': False},
                         'binary': {   'fill_value': 0,
                                       'missing_value_strategy': 'fill_with_const'},
                         'category': {   'fill_value': '<UNK>',
                                         'lowercase': False,
                                         'missing_value_strategy': 'fill_with_const',
                                         'most_common': 10000},
                         'force_split': False,
                         'image': {'missing_value_strategy': 'backfill'},
                         'numerical': {   'fill_value': 0,
                                          'missing_value_strategy': 'fill_with_const'},
                         'sequence': {   'fill_value': '',
                                         'format': 'space',
                                         'lowercase': False,
                                         'missing_value_strategy': 'fill_with_const',
                                         'most_common': 20000,
                                         'padding': 'right',
                                         'padding_symbol': '<PAD>',
                                         'sequence_length_limit': 256,
                                         'unknown_symbol': '<UNK>'},
                         'set': {   'fill_value': '',
                                    'format': 'space',
                                    'lowercase': False,
                                    'missing_value_strategy': 'fill_with_const',
                                    'most_common': 10000},
                         'split_probabilities': (0.7, 0.1, 0.2),
                         'stratify': None,
                         'text': {   'char_format': 'characters',
                                     'char_most_common': 70,
                                     'char_sequence_length_limit': 1024,
                                     'fill_value': '',
                                     'lowercase': True,
                                     'missing_value_strategy': 'fill_with_const',
                                     'padding': 'right',
                                     'padding_symbol': '<PAD>',
                                     'unknown_symbol': '<UNK>',
                                     'word_format': 'space_punct',
                                     'word_most_common': 20000,
                                     'word_sequence_length_limit': 256},
                         'timeseries': {   'fill_value': '',
                                           'format': 'space',
                                           'missing_value_strategy': 'fill_with_const',
                                           'padding': 'right',
                                           'padding_value': 0,
                                           'timeseries_length_limit': 256}},
    'training': {   'batch_size': 128,
                    'bucketing_field': None,
                    'decay': False,
                    'decay_rate': 0.96,
                    'decay_steps': 10000,
                    'dropout_rate': 0.0,
                    'early_stop': 3,
                    'epochs': 200,
                    'gradient_clipping': None,
                    'increase_batch_size_on_plateau': 0,
                    'increase_batch_size_on_plateau_max': 512,
                    'increase_batch_size_on_plateau_patience': 5,
                    'increase_batch_size_on_plateau_rate': 2,
                    'learning_rate': 0.001,
                    'learning_rate_warmup_epochs': 5,
                    'optimizer': {   'beta1': 0.9,
                                     'beta2': 0.999,
                                     'epsilon': 1e-08,
                                     'type': 'adam'},
                    'reduce_learning_rate_on_plateau': 0,
                    'reduce_learning_rate_on_plateau_patience': 5,
                    'reduce_learning_rate_on_plateau_rate': 0.5,
                    'regularization_lambda': 0,
                    'regularizer': 'l2',
                    'staircase': False,
                    'validation_field': 'combined',
                    'validation_measure': 'loss'}}


Using full raw csv, no hdf5 and json file with the same name have been found
Building dataset (it may take a while)
Traceback (most recent call last):
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/pandas/core/indexes/base.py", line 2656, in get_loc
    return self._engine.get_loc(key)
  File "pandas/_libs/index.pyx", line 108, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/index.pyx", line 132, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/hashtable_class_helper.pxi", line 1601, in pandas._libs.hashtable.PyObjectHashTable.get_item
  File "pandas/_libs/hashtable_class_helper.pxi", line 1608, in pandas._libs.hashtable.PyObjectHashTable.get_item
KeyError: 'text'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "ludwig", line 11, in <module>
    load_entry_point('ludwig==0.1.0', 'console_scripts', 'ludwig')()
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/cli.py", line 86, in main
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/cli.py", line 64, in __init__
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/cli.py", line 70, in train
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/train.py", line 663, in cli
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/train.py", line 224, in full_train
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/data/preprocessing.py", line 457, in preprocess_for_training
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/data/preprocessing.py", line 62, in build_dataset
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/data/preprocessing.py", line 83, in build_dataset_df
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/ludwig-0.1.0-py3.6.egg/ludwig/data/preprocessing.py", line 123, in build_metadata
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/pandas/core/frame.py", line 2927, in __getitem__
    indexer = self.columns.get_loc(key)
  File "/Users/loretoparisi/Documents/Projects/AI/ludwig/venv/lib/python3.6/site-packages/pandas/core/indexes/base.py", line 2658, in get_loc
    return self._engine.get_loc(self._maybe_cast_indexer(key))
  File "pandas/_libs/index.pyx", line 108, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/index.pyx", line 132, in pandas._libs.index.IndexEngine.get_loc
  File "pandas/_libs/hashtable_class_helper.pxi", line 1601, in pandas._libs.hashtable.PyObjectHashTable.get_item
  File "pandas/_libs/hashtable_class_helper.pxi", line 1608, in pandas._libs.hashtable.PyObjectHashTable.get_item
KeyError: 'text'

How to get the visualization result?

After running ludwig visualize --visualization learning_curves -ts results/experiment_run_4/training_statistics.json, it returns empty. Where to get the visualization results?Is it saved as a image in somewhere?

--model_load_path not working ?

I'm trying to use an existing model to improve it using train feature.
However, using --model_load_path results/experiment_run_0/model/ is creating a new experiment_run_1 with starting results as the experiment_run_0. Am I doing something wrong?

KeyError when using text features

When using text features, the following error is raised:

Traceback (most recent call last):
  File "/Users/user/.virtualenvs/ml/bin/ludwig", line 10, in <module>
    sys.exit(main())
  File "/Users/user/.virtualenvs/ml/lib/python3.6/site-packages/ludwig/cli.py", line 86, in main
    CLI()
  File "/Users/user/.virtualenvs/ml/lib/python3.6/site-packages/ludwig/cli.py", line 64, in __init__
    getattr(self, args.command)()
  File "/Users/user/.virtualenvs/ml/lib/python3.6/site-packages/ludwig/cli.py", line 70, in train
    train.cli(sys.argv[2:])
  File "/Users/user/.virtualenvs/ml/lib/python3.6/site-packages/ludwig/train.py", line 663, in cli
    full_train(**vars(args))
  File "/Users/user/.virtualenvs/ml/lib/python3.6/site-packages/ludwig/train.py", line 224, in full_train
    random_seed=random_seed
  File "/Users/user/.virtualenvs/ml/lib/python3.6/site-packages/ludwig/data/preprocessing.py", line 562, in preprocess_for_training
    [training_set, validation_set, test_set]
  File "/Users/user/.virtualenvs/ml/lib/python3.6/site-packages/ludwig/data/preprocessing.py", line 777, in replace_text_feature_level
    level)
KeyError: 'name_word'

This is due to the following lines https://github.com/uber/ludwig/blob/9de6ee32f0e2e6cc6157f0772aa8c28a8d662fe8/ludwig/data/preprocessing.py#L774-L778 as it removes the columns regarding of whether they exist or not. Conditionally removing the columns fixes the issue and the model trains successfully.

The model definition I'm using is:

input_features:
    -
        name: name
        type: text
        encoder: rnn
        level: char

output_features:
    -
        name: sex
        type: category

where to get the datasets in your examples?

I want to follow the examples and where to get the datasets in the examples?

ludwig CLI: cannot 'train' or 'experiment'

Installed ludwig today (pip install ludwig) on Python 3.5.5 on Ubuntu 16.04 and on Python 3.6.8 on Windows 10 (version 1809). In both environments, ludwig seems to be failing to open the model file and issues the error "AttributeError: 'str' object has no attribute 'get'".

Here is my cmd line:
ludwig experiment --model_definition model_definition.yaml --data_csv e:\github\LabCoat\Client\data\cars.json

Here is the stack trace:

Traceback (most recent call last):
File "c:\anaconda3\envs\babyai\lib\runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "c:\anaconda3\envs\babyai\lib\runpy.py", line 85, in run_code
exec(code, run_globals)
File "c:\anaconda3\envs\babyai\Scripts\ludwig.exe_main.py", line 9, in
File "c:\anaconda3\envs\babyai\lib\site-packages\ludwig\cli.py", line 86, in main
CLI()
File "c:\anaconda3\envs\babyai\lib\site-packages\ludwig\cli.py", line 64, in init
getattr(self, args.command)()
File "c:\anaconda3\envs\babyai\lib\site-packages\ludwig\cli.py", line 67, in experiment
experiment.cli(sys.argv[2:])
File "c:\anaconda3\envs\babyai\lib\site-packages\ludwig\experiment.py", line 548, in cli
experiment(**vars(args))
File "c:\anaconda3\envs\babyai\lib\site-packages\ludwig\experiment.py", line 161, in experiment
model_definition = merge_with_defaults(model_definition)
File "c:\anaconda3\envs\babyai\lib\site-packages\ludwig\utils\defaults.py", line 126, in merge_with_defaults
model_definition.get('preprocessing', {})
AttributeError: 'str' object has no attribute 'get'

It doesn't seem to be opening the model file (same error with existing or non-existing model file specified).

Image classification: could not broadcast input array from shape into shape

Receiving this error trying to run a simple image classification example

  File "/Users/colinpetit/.pyenv/versions/3.6.5/lib/python3.6/site-packages/ludwig/features/image_feature.py", line 101, in add_feature_data
    data[feature['name']][i, :, :, :] = img
ValueError: could not broadcast input array from shape (743,620,3) into shape (383,442,4)

model_definition_file:

input_features:
  -
    name: img
    type: image
    encoder: stacked_cnn
output_features:
  -
    name: type
    type: category
training:
  epochs: 10

ModuleNotFoundError: No module named 'en_core_web_sm' while training a dataset

Getting No module found 'en_core_web_sm' while training dataset

I ran following command

ludwig train --data_csv path/to/camera_dataset.csv --model_definition "{input_
features: [{name: Model, type: text}], output_features: [{name: Price, type: numerical}]}"

Error Screenshot

Sample Data set screenshot

Predict from stdin or via the programmatic api

As soon as my brand new model is ready, I would like ludwig the enable predictions via the command line, so that one could do like

echo "call me at +3912345679 for offers" | ludwig predict --only_prediction --model_path /path/to/model -

where the - may indicate the stdin (as an example) and get the model predictions directly. This would help to integrated ludwing command in a inference pipeline.

If I have understood well the api (my model is not ready yet...) it should be possibile via the programmatic api like that:

from ludwig import LudwigModel

# load a model
model = LudwigModel.load(model_path)

# obtain predictions
myDict = { 'text': ['call me at +3912345679 for offers'] }
predictions = model.predict(data_dict=myDict,
  return_type=dict
  batch_size=128,
  gpus=None,
  gpu_fraction=1,
  logging_level=logging.DEBUG)

# close model (eventually)
model.close()

Is that correct?

Image Captioning error

Traceback (most recent call last):
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/bin/ludwig", line 11, in
sys.exit(main())
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/cli.py", line 86, in main
CLI()
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/cli.py", line 64, in init
getattr(self, args.command)()
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/cli.py", line 67, in experiment
experiment.cli(sys.argv[2:])
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/experiment.py", line 548, in cli
experiment(**vars(args))
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/experiment.py", line 258, in experiment
debug=debug
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/train.py", line 368, in train
debug=debug
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/models/model.py", line 110, in init
**kwargs
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/models/model.py", line 184, in __build
is_training=self.is_training
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/models/outputs.py", line 43, in build_outputs
**kwargs
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/models/outputs.py", line 93, in build_single_output
**kwargs
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/features/base_feature.py", line 314, in concat_dependencies_and_build_output
**kwargs
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/features/sequence_feature.py", line 250, in build_output
kwarg=kwargs
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/features/sequence_feature.py", line 320, in build_sequence_output
eval_loss
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/features/sequence_feature.py", line 445, in sequence_measures
self.name
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/ludwig/models/modules/measure_modules.py", line 73, in masked_accuracy
maxlen=correct_predictions.shape[1], dtype=tf.int32)
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/tensorflow/python/ops/array_ops.py", line 2504, in sequence_mask
maxlen = ops.convert_to_tensor(maxlen)
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 1050, in convert_to_tensor
as_ref=False)
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 1146, in internal_convert_to_tensor
ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
File "/Users/liusitong/lstworkspace/wowo-server/services/crawler_py/venv/lib/python3.6/site-packages/tensorflow/python/framework/constant_op.py", line 282, in _dimension_tensor_conversion_function
raise ValueError("Cannot convert an unknown Dimension to a Tensor: %s" % d)
ValueError: Cannot convert an unknown Dimension to a Tensor: ?

Feature Request: Support TSV and JSON file formats as input data

CSV is good for numerical data, but when you have text data that may contain , and ", escaping the values in the columns can be tricky and identification of delimiter comma is harder for CSV parsers.

Can you add support for data file formats, TSV and JSON which do not have the problems above as much?

Missing reuters-allcats.csv in the Example page

I cannot find reuters-allcats.csv file link in https://uber.github.io/ludwig/examples/ .

I think the Example page should contain this file, so a user can actually follow the example to try Ludwig out.

Thanks for open sourcing this project.

model_definition_file in programmatic API takes an open file object instead of file path

When I try to use the programmatic API,

model = LudwigModel(model_definition={}, model_definition_file="some/path/to/file.yml")

it throws AttributeError: 'str' object has no attribute 'get'. It seems like when loading the model_definition_file, Ludwig is using yaml.load, which, according to the pyyaml documentation, takes a byte string, a Unicode string, an open binary file object, or an open text file object. However, the ludwig documentation says that model_definition_file should be a file path. I tried passing in an open file object and it works.

I think either the documentation needs to be changed or the model initialization should be modified to actually take a file path instead.

ImportError: cannot import name '_validate_lengths'

E:\repos\tensorflow>pushd E:\repos\tensorflow

E:\repos\tensorflow>ludwig experiment --data_csv translation.csv --model_definition_file model_definition.yaml
e:\programdata\anaconda3\lib\site-packages\h5py_init_.py:36: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type.
from .conv import register_converters as register_converters
Traceback (most recent call last):
File "e:\programdata\anaconda3\lib\runpy.py", line 193, in run_module_as_main
"main", mod_spec)
File "e:\programdata\anaconda3\lib\runpy.py", line 85, in run_code
exec(code, run_globals)
File "E:\ProgramData\Anaconda3\Scripts\ludwig.exe_main.py", line 5, in
File "e:\programdata\anaconda3\lib\site-packages\ludwig_init.py", line 16, in
from ludwig.api import LudwigModel
File "e:\programdata\anaconda3\lib\site-packages\ludwig\api.py", line 39, in
from ludwig.data.postprocessing import postprocess_df, postprocess
File "e:\programdata\anaconda3\lib\site-packages\ludwig\data\postprocessing.py", line 19, in
from ludwig.features.feature_registries import output_type_registry
File "e:\programdata\anaconda3\lib\site-packages\ludwig\features\feature_registries.py", line 33, in
from ludwig.features.image_feature import ImageBaseFeature
File "e:\programdata\anaconda3\lib\site-packages\ludwig\features\image_feature.py", line 24, in
from skimage.io import imread
File "e:\programdata\anaconda3\lib\site-packages\skimage_init.py", line 158, in
from .util.dtype import *
File "e:\programdata\anaconda3\lib\site-packages\skimage\util_init.py", line 7, in
from .arraycrop import crop
File "e:\programdata\anaconda3\lib\site-packages\skimage\util\arraycrop.py", line 8, in
from numpy.lib.arraypad import _validate_lengths
ImportError: cannot import name '_validate_lengths'

Cannot feed value of shape for Tensor when predicting model with image

I get this error when trying to perform prediction with a new image on trained model:

Traceback (most recent call last):
  File "/home/andrey/.venvs/ludwig-learn/bin/ludwig", line 11, in <module>
    sys.exit(main())
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/cli.py", line 86, in main
    CLI()
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/cli.py", line 64, in __init__
    getattr(self, args.command)()
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/cli.py", line 73, in predict
    predict.cli(sys.argv[2:])
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/predict.py", line 379, in cli
    full_predict(**vars(args))
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/predict.py", line 104, in full_predict
    debug
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/predict.py", line 173, in predict
    gpu_fraction=gpu_fraction
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/models/model.py", line 1182, in predict
    only_predictions=only_predictions
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/ludwig/models/model.py", line 756, in batch_evaluation
    is_training=is_training
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 929, in run
    run_metadata_ptr)
  File "/home/andrey/.venvs/ludwig-learn/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1128, in _run
    str(subfeed_t.get_shape())))
ValueError: Cannot feed value of shape (1,) for Tensor 'image_path/image_path:0', which has shape '(?, 100, 100, 3)'

The model definition I use:

input_features:
    -
        name: image_path
        type: image
        encoder: stacked_cnn
        in_memory: false
        height: 100
        width: 100

output_features:
    -
        name: tags
        type: set`

Example of testing csv-file:

image_path,tags
testdata/cat.101.jpg

I've tried output features with category and set types but eventually get that error.

Support for External Data Sources

Would it be possible to have S3 or any other object storage support? E.g reading input data from S3 or saving model to S3.

I know there could be a workaround using S3FS, but is this supported outta-the-box?

Thanks

Feature request: output probabilities

This is either because I couldn't 👀 instructions in documentation or something missing from current functionality: is there a way to have something like .predict_proba set of methods that would not only output target label(s) but also confidence/probability model assigns to it?

can ludwig use to estimate time of arrival (eta) ?

the ludwig looks good, and I want to know how to use ludwig to estimate time of arrival (eta), many thanks.

Feature Request: Add Features at Runtime

Currently adding a new feature / encoder / decoder requires forking the package.

It would be awesome to be able to add them at run time as part of the API and/or a conf.py e.g.

ludwig.add_base_feature(NotNumericalBaseFeature)

pytorch support?

Great project!

Given the stress on modular design, I'm wondering how hard it is to "swap-in" the pytorch library and the models.

Is the library tied deeply to tensorflow?
Is pytorch support planned in the roadmap?

Question: are unittests going to be open sourced?

Are unittests going to be open sourced? At the moment only integration tests have been published.

On the positive side, the Developers Guide says:

Current test coverage is limited to several integration tests which ensure end-to-end functionality but we are planning to expand it.

which hints that more testing is yet to be added.

"Getting started" issue

To help CLI users get started with Ludwig, it would be great if you could include an example with an associated downloadable .csv file and a matching model file to train it with.

The current "Getting started" documentation (and the associated examples section) all give a sketch of what the .csv file could look like for a particular scenario, but without a specific file that users can try. Providing a starting "hello world" pair of .csv /model files help eliminate these as a source of error when users are trying to establish their first successful run with ludwig.

TPU support?

Since Ludwig is running on TensorFlow, it can run on a TPU, which can be useful especially combined with the free TPU provided by Colaboratory.

Is there a way to provide the appropriate TPU-connector URL to a Ludwig model? (for both training and inference).

Ludwig does not use tensorflow-gpu?

I have tensorflow-gpu installed and Keras can use the GPU effectively. I only have one GPU.
With ludwig, I tried a regression problem and found the training is very slow.
train_stats = ludwig_model.train(data_df=df, logging_level=logging.ERROR, gpus=[0])
By watch -n 1 nvidia-smi, I found the training did not actually utilize the GPU but stored the data in the GPU memory anyway.

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 822 G /usr/bin/gnome-shell 186MiB |
| 0 4687 C /home/yshi1/anaconda3/bin/python 22929MiB |
| 0 10166 C /home/yshi1/anaconda3/bin/python 977MiB |
| 0 23459 G /usr/bin/X 191MiB |
+-----------------------------------------------------------------------------+

KeyError: 'text_char' on preprocessing.py

Error Description

The error is raised when you run the ludwig experiment --data_csv src/data/input/training.csv for the second time with the same file name .csv input file. you have to change the file name every time you run the experiment. When it creates the .json and the .hdf5 for the first time it works, but when it loads the pre created .json and the .hdf5 files it raise the following error

Code

ludwig experiment --data_csv src/data/input/training.csv --model_definition_file model_definition.yaml

Error

Found hdf5 and json with the same filename of the csv, using them instead
Using full hdf5 and json
Loading data from: src/data/input/training.hdf5
Loading metadata from: src/data/input/training.json
usr/lib/python3.5/site-packages/h5py/_hl/dataset.py:313: H5pyDeprecationWarning: dataset.value has been deprecated. Use dataset[()] instead.
"Use dataset[()] instead.", H5pyDeprecationWarning)
Traceback (most recent call last):
File "/usr/local/bin/ludwig", line 11, in
load_entry_point('ludwig==0.1.0', 'console_scripts', 'ludwig')()
File "usr/lib/python3.5/site-packages/ludwig/cli.py", line 86, in main
CLI()
File "usr/lib/python3.5/site-packages/ludwig/cli.py", line 64, in init
getattr(self, args.command)()
File "usr/lib/python3.5/site-packages/ludwig/cli.py", line 67, in experiment
experiment.cli(sys.argv[2:])
File "usr/lib/python3.5/site-packages/ludwig/experiment.py", line 548, in cli
experiment(**vars(args))
File "usr/lib/python3.5/site-packages/ludwig/experiment.py", line 234, in experiment
random_seed=random_seed
File "usr/lib/python3.5/site-packages/ludwig/data/preprocessing.py", line 562, in preprocess_for_training
[training_set, validation_set, test_set]
File "usr/lib/python3.5/site-packages/ludwig/data/preprocessing.py", line 777, in replace_text_feature_level
level)
KeyError: 'text_char'

Training using training data and validation CSV caused errors

When using "train" command with training and validation data will cause inappropriate attribute access violation for Dataframe.
I was running ludwig on Windows 10 and Python 3.6.8.

ludwig train --data_train_csv data_train.csv --data_validation_csv data_val.csv -mdf model.yaml

Using training raw csv, no hdf5 and json file with the same name have been found
Building dataset (it may take a while)
Loading training csv...
done
Loading validation csv..
done
Loading test csv..
done
Concatenating csvs..
done
Traceback (most recent call last):
File "d:\python\python36\lib\runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "d:\python\python36\lib\runpy.py", line 85, in run_code
exec(code, run_globals)
File "D:\Python\python36\Scripts\ludwig.exe_main.py", line 9, in
File "d:\python\python36\lib\site-packages\ludwig\cli.py", line 86, in main
CLI()
File "d:\python\python36\lib\site-packages\ludwig\cli.py", line 64, in init
getattr(self, args.command)()
File "d:\python\python36\lib\site-packages\ludwig\cli.py", line 70, in train
train.cli(sys.argv[2:])
File "d:\python\python36\lib\site-packages\ludwig\train.py", line 663, in cli
full_train(**vars(args))
File "d:\python\python36\lib\site-packages\ludwig\train.py", line 224, in full_train
random_seed=random_seed
File "d:\python\python36\lib\site-packages\ludwig\data\preprocessing.py", line 487, in preprocess_for_training
random_seed=random_seed
File "d:\python\python36\lib\site-packages\ludwig\data\preprocessing.py", line 90, in build_dataset_df
global_preprocessing_parameters
File "d:\python\python36\lib\site-packages\ludwig\data\preprocessing.py", line 165, in build_data
preprocessing_parameters
File "d:\python\python36\lib\site-packages\ludwig\features\image_feature.py", line 57, in add_feature_data
csv_path = os.path.dirname(os.path.abspath(dataset_df.csv))
File "d:\python\python36\lib\site-packages\pandas\core\generic.py", line 5067, in getattr
return object.getattribute(self, name)
AttributeError: 'DataFrame' object has no attribute 'csv'

Allow turning off saving best model after epochs (DONE) and regression prediction instability (TODO)

Saving the current best model after epochs does not consume too much time if the training is for big models, but it could be a high cost for training small models. I have found no way to turn off the saving best model feature. Tried to set skip_save_progress_weights=True, skip_save_processed_input=True but it did not help. Thanks

Can we have some more documentation of the predict function/ data format

I'm new to machine learning and have managed to train ludwig using an f1 data set:

team,surname,position,track,year
Mercedes,Rosberg,1,Albert Park Grand Prix Circuit,2014
McLaren,Magnussen,2,Albert Park Grand Prix Circuit,2014
McLaren,Button,3,Albert Park Grand Prix Circuit,2014
...

my model is as follows:

input_features:
 -
   name: team
   type: category
 -
   name: track
   type: category
 -
   name: surname
   type: category
 -
   name: year
   type: category
output_features:
 -
   name: position
   type: numerical
training:
 epochs: 10

When I run the predict function I am trying to get a driver / position prediction for each track. What do I need to do in order to make this happen. I have tried to make a new data file without the position bugt it falls over with a missing 'position' key so clearly that's needed even though that is the field I am trying to predict:

team,surname,track,year
Mercedes,Hamilton,Albert Park Grand Prix Circuit,2019
McLaren,Magnussen,Albert Park Grand Prix Circuit,2019
McLaren,Button,Albert Park Grand Prix Circuit,2019

Note the above data is truncated from the last 5 years with positions of 10th and above. I can expand this if it helps training!

When I add some positions in the prediction just returns a list of numerical values:

1.7585578
2.0917244
1.6508131

I'm assuming this is the position that is would be expecting for that driver/track/team combo but I'm not sure.

Can someone explain this to me further? Do I need a model without the 'position' key in for the prediction? Is there any way to tell Ludwig to use full Integers for positions and maybe only assign one of each per track (/per year)

Or can I be pointed to somewhere I can learn this myself?

Really slow startup time

Starting ludwig is really slow, taking more than 10 seconds to start on a macbook pro. This is probably due to some model imports.

$ time ludwig --help
usage: ludwig <command> [<args>]

Available sub-commands:
   experiment            Runs a full experiment training a model and testing it
   train                 Trains a model
   predict               Predicts using a pretrained model
   visualize             Visualizes experimental results
   collect_weights       Collects tensors containing a pretrained model weights
   collect_activations   Collects tensors for each datapoint using a pretrained model

ludwig cli runner

positional arguments:
  command     Subcommand to run

optional arguments:
  -h, --help  show this help message and exit

real	0m11.582s
user	0m8.958s
sys	0m1.447s

Import Error: Cannot load backend 'TkAgg' which requires the 'tk' interactive framework, as 'headless' is currently running

I'm facing import error while executing the ludwig from google colab.. !!

Below is the code which I executed in steps:

Loading Ludwig
!pip install ludwig
Upload file from local to Drive - to use in Colab
from google.colab import files
uploaded = files.upload()
Train DataSet using ludwig cmd
!ludwig train --data_csv ludwig_traindataset_1.csv --model_definition_file model_def.yaml

My model_def.yaml

input_features:
  - name: utterance
    type: sequence
    encoder: rnn
    cell_type: lstm
    bidirectional: true
    num_layers: 2
    reduce_output: None

output_features:
  - name: intent
    type: category
    reduce_input: sum
    num_fc_layers: 1
    fc_size: 64
  - name: slots
    type: sequence
    decoder: tagger

Dummy Dataset:

Attached the screenshot error:

Can anyone help me on this in Google Colab GPU / Python 3.6