Linting your project

Linting is the process of statically analyzing code to find code style violations and to detect errors. mlf-core implements a custom linting system, but depending on the template external tools linting tools may additionally be called.

Usage

mlf-core lint can be invoked on an existing project using

$ mlf-core lint <OPTIONS> <PATH>

mlf-core’s linting is divided into three distinct phases.

  1. All linting functions, which all templates share are called and the results are collected.

  2. Template specific linting functions are invoked and the results are appended to the results of phase 1

  3. Template specific external linters are called (e.g. autopep8 for Python based projects)

The linting results of the first two phases are assigned into 3 groups:

  1. Passed

  2. Passed with warning

  3. Failed

If any of the checks failed linting stops and returns an error code.

Linting example

Linting applied to a newly created mlflow-pytorch project.

To examine the reason for a failed linting test please follow the URL. All reasons are explained in the section Linting codes.

Linting codes

The following error or warning numbers correspond to errors found during linting. If you are not sure why a specific linting error has occurred you may find more information using the respective error code.

General

general-1

File not found. This error occurs when your project does not include all of mlf-core’s files, which all templates share.
Please create the file and populate it with appropriate values. You should also critically reflect why it is missing, since at the time of the project creation using mlf-core this file should not have been missing!

general-2

Dockerfile invalid. This error usually originates from empty Dockerfiles or missing FROM statements.

general-3

TODO String found. The origin of this warning are mlf-core TODO: TODO mlf-core: or strings in the respective files. Usually, they point to things that should be manually configured or require other attention. You may remove them if there is no task for you to be solved.

general-4

Cookiecutter string found. This error occurs if something went wrong at the project creation stage. After a project has been created using mlf-core there should not be any jinja2 syntax left.

general-5

Versions not consistent. If the version of all files specified in the [bumpversion] sections defined in the qube.cfg file are not consistent, this error may be found. Please ensure that the version is consistent! If you need to exclude specific lines from this check please consult Bumping the version of an existing project. To prevent this error you should only increase the version of your project using mlf-core bump-version.

general-6

changelog.rst invalid. The changelog.rst file requires that every changelog section has a header with the version and the corresponding release date. The version above another changelog section should always be greater than the section below (e.g. 1.1.0 above 1.0.0). Every section must have the headings **Added**, **Fixed**, **Dependencies** and **Deprecated**.

mlflow-pytorch

mlflow-pytorch-1

File not found. This error occurs when your project does not include all of mlflow-pytorch’s expected files.
Please create the file and populate it with appropriate values. You should also critically reflect why it is missing, since at the time of the project creation using mlf-core this file should not have been missing!

mlflow-pytorch-2

Expected line not found. This error occurs when CPU/GPU deterministic training may no longer be guaranteed, since a required setting has been disabled or removed.
Currently, mlflow-pytorch expects the following lines in the main entry script:
1
2
3
4
 trainer.deterministic = True,
 trainer.benchmark = False,
 set_general_random_seeds(general_seed),
 set_pytorch_random_seeds(pytorch_seed, num_of_gpus)
Line 1 enables deterministic training operations
Line 2 disables the search for the optimal algorithm for specific operations, which may not necessarily be deterministic.
Line 3 sets the general random seeds (python random, numpy random and python general)
Line 4 sets the seed of Pytorch

mlflow-pytorch-3

Function operates non-deterministically.
Several functions and algorithms available in Pytorch are still based on atomic add or other non-deterministic operators. Hence, these functions are not allowed to be used.
Currently mlflow-pytorch reports:
1
2
3
4
5
6
7
8
'index_add',
'scatter_add',
'bincount',
'embedding_bag',
'ctc_loss',
'interpolate',
'repeat_interleave',
'index_select'

mlflow-tensorflow

mlflow-tensorflow-1

File not found. This error occurs when your project does not include all of mlflow-tensorflow’s expected files.
Please create the file and populate it with appropriate values. You should also critically reflect why it is missing, since at the time of the project creation using mlf-core this file should not have been missing!

mlflow-tensorflow-2

Expected line not found. This error occurs when CPU/GPU deterministic training may no longer be guaranteed, since a required setting has been disabled or removed.
Currently, mlflow-tensorflow expects:
1
2
3
4
5
6
7
set_general_random_seeds(dict_args["general_seed"]),
set_tensorflow_random_seeds(dict_args["tensorflow_seed"])
def set_tensorflow_random_seeds(seed):
    tf.random.set_seed(seed)
    tf.config.threading.set_intra_op_parallelism_threads = 1  # CPU only
    tf.config.threading.set_inter_op_parallelism_threads = 1  # CPU only
    os.environ['TF_DETERMINISTIC_OPS'] = '1'
Line 1 sets the general random seeds (python random, numpy random and python general)
Line 2 sets the seed of Tensorflow
Line 4 fixes the seed of Tensorflow
Line 5 sets the number of threads within an individual operation for parallelism to 1
Line 6 sets the number of threads between independent operations for parallelism to 1
Line 7 enables and forces all deterministic operations

mlflow-tensorflow-3

Function operates non-deterministically.
There are a couple of functions left in Tensorflow, which are known to be operating non-deterministically. They are not allowed to be used.
1
2
'softmax_cross_entropy_with_logits',
'sparse_softmax_cross_entropy_with_logits'

mlflow-xgboost

mlflow-xgboost-1

File not found. This error occurs when your project does not include all of mlflow-dask’s expected files.
Please create the file and populate it with appropriate values. You should also critically reflect why it is missing, since at the time of the project creation using mlf-core this file should not have been missing!

mlflow-xgboost-2

Expected line not found. This error occurs when CPU/GPU deterministic training may no longer be guaranteed, since a required setting has been disabled or removed.
Currently, mlflow-xgboost expects:
1
2
3
4
set_general_random_seeds(dict_args["general_seed"]),
set_xgboost_random_seeds(dict_args["xgboost_seed"], param)
def set_xgboost_random_seeds(seed, param):
    param['seed'] = seed
Line 1 sets the general random seeds (python random, numpy random and python general)
Line 2 sets the seed of XGBoost
Line 4 fixes the seed of XGBoost

mlflow-xgboost-3

The version of XGBoost has to be at least 1.1.0, since this is first version which includes all deterministic operations.
Refrain from using versions older than 1.1.0, especially when making use of GPUs.

mlflow-xgboost-4

The all_reduce algorithm in XGBoost may not operate deterministically.

mlflow-xgboost_dask

mlflow-xgboost_dask-1

File not found. This error occurs when your project does not include all of mlflow-xgboost_dask’s expected files.
Please create the file and populate it with appropriate values. You should also critically reflect why it is missing, since at the time of the project creation using mlf-core this file should not have been missing!

mlflow-xgboost_dask-2

Expected line not found. This error occurs when CPU/GPU deterministic training may no longer be guaranteed, since a required setting has been disabled or removed.
Currently, mlflow-xgboost_dask expects:
1
2
3
4
set_general_random_seeds(dict_args["general_seed"]),
set_xgboost_dask_random_seeds(dict_args["xgboost_seed"], param)
def set_xgboost_random_seeds(seed, param):
    param['seed'] = seed
Line 1 sets the general random seeds (python random, numpy random and python general)
Line 2 sets the seed of XGBoost
Line 4 fixes the seed of XGBoost

mlflow-xgboost_dask-3

The version of XGBoost has to be at least 1.1.0, since this is first version which includes all deterministic operations.
Refrain from using versions older than 1.1.0, especially when making use of GPUs.

mlflow-xgboost-4

The all_reduce algorithm in XGBoost may not operate deterministically.