ramble.experimental namespace

Submodules

ramble.experimental.uploader module

class ramble.experimental.uploader.BigQueryUploader

Bases: Uploader

Class to handle upload of FOMs to BigQuery

chunked_upload(table_id, data)

insert_data(uri: str, results) → None

perform_upload(uri, results)

class ramble.experimental.uploader.Experiment(name, workspace_hash, data, timestamp)

Bases: object

Class representation of experiment data

generate_hash()

get_hash()

to_json()

class ramble.experimental.uploader.PrintOnlyUploader

Bases: Uploader

An uploader that only prints out formatted data without actually uploading.

perform_upload(uri, results)

class ramble.experimental.uploader.Uploader

Bases: object

perform_upload(uri, data)

ramble.experimental.uploader.determine_node_type(experiment, contexts)

Extract node type from available FOMS.

First prio is machine specific data, such as GCP meta data Second prio is more general data like CPU type

ramble.experimental.uploader.format_data(data_in)

Goal: convert results to a more searchable and decomposed format for insertion into data store (etc)

Input:

{ experiment_name:
    { "CONTEXTS": {
        "context_name": "FOM_name { unit: "value", "value":value" }
    ...}
    }
}

Output: The general idea is the decompose the results into a more “database” like format, where the runs and FOMs are in a different table.

ramble.experimental.uploader.get_user()

ramble.experimental.uploader.upload_results(results)

class ramble.experimental.uploader.uploader_types(*values)

Bases: Enum

BigQuery = 1

PrintOnly = 2