datmo.snapshot module

class datmo.snapshot.Snapshot(snapshot_entity, home=None)[source]

Snapshot is an entity object to enable user access to properties

Parameters:
  • snapshot_entity (datmo.core.entity.snapshot.Snapshot) – core snapshot entity to reference
  • home (str, optional) – root directory of the project (default is CWD, if not provided)
id

str – the id of the entity

model_id

str – the parent model id for the entity

session_id

str – id of session associated with task

id

str – the id of the entity

model_id

str – the parent model id for the entity

session_id

str – session id within which snapshot is created

message

str – long description of snapshot

code_id

str – code reference associated with the snapshot

environment_id

str – id for environment used to create snapshot

file_collection_id

str – file collection associated with the snapshot

config

dict – key, value pairs of configurations

stats

dict – key, value pairs of metrics and statistics

task_id

str – task id associated with snapshot

label

str – short description of snapshot

created_at

datetime.datetime

Raises:InvalidArgumentType
files
get_files(mode='r')[source]

Returns a list of file objects for the snapshot

Parameters:mode (str) – file object mode (default is “r” which signifies read mode)
Returns:list of file objects associated with the snapshot
Return type:list
datmo.snapshot.create(message, label=None, home=None, task_id=None, commit_id=None, environment_id=None, filepaths=None, config=None, stats=None)[source]

Create a snapshot within a project

The project must be created before this is implemented. You can do that by using the following command:

$ datmo init
Parameters:
  • message (str) – a description of the snapshot for later reference
  • label (str, optional) – a short description of the snapshot for later reference (default is None, which means a blank label is stored)
  • home (str, optional) – absolute home path of the project (default is None, which will use the CWD as the project path)
  • task_id (str, optional) –

    task object id to use to create snapshot if task id is passed then subsequent parameters would be ignored. when using task id, it will overwrite the following inputs

    commit_id: taken form the source code after the task is run

    environment_id: used to run the task,

    filepaths: this is the set of all files saved during the task

    config: nothing is passed into this variable. the user may add something to the config by passing in a dict for the config

    stats: the task.results are added into the stats variable of the snapshot.

  • commit_id (str, optional) – provide the exact commit hash associated with the snapshot (default is None, which means it automatically creates a commit)
  • environment_id (str, optional) – provide the environment object id to use with this snapshot (default is None, which means it creates a default environment)
  • filepaths (list, optional) – provides a list of absolute filepaths to files or directories that are relevant (default is None, which means we have an empty
  • config (dict, optional) – provide the dictionary of configurations (default is None, which means it is empty)
  • stats (dict, optional) – provide the dictionary of relevant statistics or metrics (default is None, which means it is empty)
Returns:

returns a Snapshot entity as defined above
Return type:

Snapshot

Examples

You can use this function within a project repository to save snapshots for later use. Once you have created this, you will be able to view the snapshot with the datmo snapshot ls cli command

>>> import datmo
>>> datmo.snapshot.create(message="my first snapshot", filepaths=["/path/to/a/large/file"], config={"test": 0.4, "test2": "string"}, stats={"accuracy": 0.94})

You can also use the result of a task run in order to create a snapshot

>>> datmo.snapshot.create(message="my first snapshot from task", task_id="1jfkshg049")
datmo.snapshot.ls(session_id=None, filter=None, home=None)[source]

List snapshots within a project

The project must be created before this is implemented. You can do that by using the following command:

$ datmo init
Parameters:
  • session_id (str, optional) – session to filter output snapshots (default is None, which means no session filter is given)
  • filter (str, optional) – a string to use to filter from message and label (default is to give all snapshots, unless provided a specific string. eg: best)
  • home (str, optional) – absolute home path of the project (default is None, which will use the CWD as the project path)
Returns:

returns a list of Snapshot entities (as defined above)
Return type:

list

Examples

You can use this function within a project repository to list snapshots.

>>> import datmo
>>> snapshots = datmo.snapshot.ls()