ldmlogger 0.0.11

Python client side functions for working with LDM framework.

LDMLogger

Class LDMLogger contains methods for working with LDM framework. Basically it is python wrapper around LDM framework REST API.

By using this class provided methods users can:

• start/finish runs
• log message to a current run
• upload file to a current run
• upload sources (directory of files, excluding files matched by ldmignore file) to a current run
• upload silver data for a given run

Detailed description of these methods (in a form of pydoc documentation is given here).

Tutorial:

To start using functionality provided by ldmlogger you need to install package ldmlogger. It can be done in a following way:

 pip install ldmlogger

Also you need to have:

• an IP adress of a running instance of LDM framework

• a permission to create a user or credentials of already created user (to login to LDM framework)
• id of a project in LDM framework, that you want to work with
• user token that is issued by the LDM platform for the specific user.

After all aforementioned things have been acquired the first thing that we can do is create a run.

Creating runs

from logger import LDMLogger

user_token_loc = "<InsertYourTokenHere>"
project_id_loc = "<InsertYourProjectID>"
ldm_server_address = "<InsertServerAddress>"

#create instance of LDMLogger
lgr = LDMLogger(user_token_loc, 
                project_id_loc, 
                ldm_server_address
)

lgr.start_run("my first run")

lgr.finish_run()

Code seen in a listing above connects to LDM framework instance located at the address ldm_server_address, "logins" to this framework with token user_token_loc, and creates (starts and finishes) a run for project project_id_loc.

Logging messages

from logger import LDMLogger

user_token_loc = "<InsertYourTokenHere>"
project_id_loc = "<InsertYourProjectID>"
ldm_server_address = "<InsertServerAddress>"

#create instance of LDMLogger
lgr = LDMLogger(user_token_loc, 
                project_id_loc, 
                ldm_server_address
)

lgr.start_run("my first run")

lgr.log({"msg": "first message"})

lgr.finish_run()

Code seen in a listing above is appended with just one additional line compared to previous example. Line lgr.log({"msg": "first message"}) creates a log message inside a current run. Method log gets a free form JSON object as a parameter.

Uploading files

from logger import LDMLogger

user_token_loc = "<InsertYourTokenHere>"
project_id_loc = "<InsertYourProjectID>"
ldm_server_address = "<InsertServerAddress>"

#create instance of LDMLogger
lgr = LDMLogger(user_token_loc, 
                project_id_loc, 
                ldm_server_address
)

lgr.start_run("my first run")

lgr.upload_file("./abc.txt")

lgr.finish_run()

Line lgr.upload_file("./abc.txt") takes file from path passed as a parameter ("./abc.txt" in this case) and uploads this file to LDM server (Uploaded file will be attached to a current run). Path is resolved against current directory.

Uploading sources

TODO: update this when LDMLogger has new version of upload sources

from logger import LDMLogger

user_token_loc = "<InsertYourTokenHere>"
project_id_loc = "<InsertYourProjectID>"
ldm_server_address = "<InsertServerAddress>"

#create instance of LDMLogger
lgr = LDMLogger(user_token_loc, 
                project_id_loc, 
                ldm_server_address
)

lgr.start_run("my first run")

lgr.upload_sources()

lgr.finish_run()

Code seen in a listing above is appended with just one additional line compared to previous example. Line lgr.upload_file("./abc.txt") takes file from path passed as a parameter ("./abc.txt" in this case) and uploads this file to LDM server (within a current run). Path is resolved against current directory.

Uploading silver data

from logger import LDMLogger

user_token_loc = "<InsertYourTokenHere>"
project_id_loc = "<InsertYourProjectID>"
ldm_server_address = "<InsertServerAddress>"

#create instance of LDMLogger
lgr = LDMLogger(user_token_loc, 
                project_id_loc, 
                ldm_server_address
)

lgr.start_run("my first run")

results_on_test_data_set = [
    {
        'file': 'file1.jpeg',
        'silver': 'A'
    },
    {
        'file': 'file2.jpeg',
        'silver': 'B'
    },
]

lgr.validate(results_on_test_data_set, 'Test')

lgr.finish_run()

TODO: vai LDMLogger.validate ir labs metodes vards ? Vai nebut labak to aaukt par upload_silver_data ? Reali jau nekada validaciju tas neveic.

Code seen in a listing above uploads silver data for Test dataset in a specific run of a specific project to LDM server. Method validate takes 2 parameters. The second tells for what kind of data set we upload silver data (Test in our case). The first one specifies silver data to be uploaded. This silver data is in fact a list of objects, where each object has 2 fields: 'file' - specifing name of the file for which we provide silver label and 'silver', specifiying the value of the silver label for this file. Variable results_on_test_data_set contains an example of this kind of list. There are 2 objects. The first one saying that for file 'file1.jpeg', value of silver label is 'A'. The second one saying that for file 'file2.jpeg' value of silver label is 'B'.

Pydoc documentation

Detailed python method description (pydocs).

Class LDMLogger contains methods for working with LDM framework.

Methods:

• def __init__(self, 
               user_token, project_id=None, 
               server_url="http://localhost:5000",
               root_dir=Path(os.getcwd()).parent,
               should_upload_sources = False)
  

  Creates new instance of class LDMLogger.

  Parameters:

  • user_token:str - user token. Can be obtained insede LDM framework.
  • project_id:str - ID of the project for which we are creating run. This ID can be obtained in LDM.
  • server_url:str - IP address of LDM server instance.
  • root_dir:str - path to root_dir. Is used only in upload sources.
  • should_upload_sources:boolean. Set to True if sources shoud be uploaded when run is started.

  Returns: None

• def start_run(self, 
                comment = "", git_commit_url = "")
  

  Starts a new run.

  Parameters:

  • comment (string): Comment for a run. This parameter is optional and can be ommited.

  • git_commit_url (string): URL of a git commit representing the state of a code base used in this run. This parameter is optional and can be ommited.

  Returns: None

•   def log(self, log_msg_obj):
  

  Logs object, representing log message to server.

  Parameters:

  • log_msg_obj: Object representing message to log

  Returns: None

• def finish_run(self):
  

  Finish the current run.

  Returns: None

• def validate(self, results, dataset_type="Train"):
  

  Uploads silver data for (Train/Test/Validate) dataset to server.

  Parameters:

  • silver_data: list. List of objects representing silver data. Each object must have fields 'file', representing file name and 'silver', representing silver label for this file.

  • dataset_type: str . One of Train, Validation, Test.

  Returns:

  None

• def upload_file(self, file_name, comment = ""):
  

  Uploads file (file_name) to the logging server and attaches it to the current run.

  Parameters:

  • file_name (string): File path (on a local machine) of file to be uploaded.

  • comment (string): Comment for a file to be uploaded. This prm is optional and can be ommited.

  Returns:

  None

• def add_project(self, name, project_type, description=""):
  

  Creates project on LDM server.

  Parameters:

  • name:str - name of a project to be created
  • project_type:str - one of "ImageClassification", "ImageCaptioning", "VideoTranscription", "AudioTranscription"
  • description:str - project description. This prm is optional.

  Returns:

  project_id in case of successfull project creation or None in case of errors

• def upload_dataset(self, path_to_dataset,dataset_type_in="Train"):
  

  Creates .zip file of path_to_dataset dir and its subdirs and uploads this .zip as a dataset_type_in dataset.

  Parameters:

  • path_to_dataset - path to dir to be zipped.
  • dataset_type_in - one of Train, Test, Validation

  Returns:

  None

• def upload_sources(self, ldmignore_path = "ldmignore.txt"):
  

  Creates a .zip of root_dir (passed in constructor) and uploads this .zip to the LDM server. In case if root_dir is None no action is performed (method exits immediately).

  Parameters:

  • ldmignore_path - path to .gitignore style file, containing dirs and files to be ignored when uploading sources to LDM.

  Returns:

  None

• def download_dataset(self, path="", dataset_type_in="Train"):
  

  Downloads dataset from server.

  Parameters:

  • path - path to dir, where to put downloaded .zip file
  • dataset_type_in - one of Train, Test, Validation

  Returns:

  None

• def save_colab_notebook_history_to_file(self, file_name):
  

  Use this function to save the history of executed cells in IPython notebook to a new notebook (file_name).

  Parameters:

  • file_name (string): Path to file were notebook will be stored.

  Returns:

  None

Change log

[0.0.1] - 2020-12-09

[0.0.3] - 2021-02-10

[0.0.4] - 2021-02-12

Added

• Functions for
  • dataset upload
  • dataset download
  • project creation
  • result validation on a specific dataset

Changed

• changed package structure (only one package (logger) is uploaded)
• removed all examples from package logger
• class LDMLogger has package visibility

Removed

[0.0.6] - 2021-02-17

Changed

• start_run does not upload sources by default

[0.0.7] - 2021-03-19

Added

• Function to save notebook

[0.0.8] - 2021-03-20

Removed

• N/A

Changed

• Updated readme and docs

Added

• N/A

[0.0.9] - 2022-04-26

Small updates

[0.0.11] - 2022-04-26

Fixed typos