Parameters

Setting/ Loading/ Retrieving Parameters

Both pipelines and components use the following methods.

Setting Parameters

set_params() sets parameters in pipelines or components.

Example:

pipeline.set_params(raw_params={
    "sub_pipeline;component:param": 2,  # Sets param in component within sub_pipeline
    "sub_pipeline:param": 3,            # Sets param in sub_pipeline
    "param": 4                          # Sets param at root level
}, base_priority=1000)

Loading Parameters

load_params() loads parameters from files.

Example:

pipeline.load_params(params_file='params.toml', placeholders_file='placeholders.toml')

note:

1. both can use .py file as parameters/placeholders as well

2. By default, parameters files are indicated in gpt_graph/config/config.toml

3. You also need to set GPT_GRAPH_FOLDER in gpt_graph/config/env.toml

Retrieving Parameters

get_all_params() retrieves all parameters.

Example:

all_params = pipeline.get_all_params()
print(all_params)

Parameter Priority

Parameters can have different priorities:

Priority	Description
0	Default
1+	Loading from config file
1000+	Manually set params
…	…

Parameter Types

Parameters have several attributes:

1. Type:

   • input

   • param

   • placeholder

2. Status:

   • input

   • empty

   • default

   • assigned

   • ult_input

   • config

   • filter_cri

   • cache

3. Value:

   • The actual parameter value

   • <CACHE>

   • <INPUT>

   • <EMPTY>

4. priority:

   • 1000+ indicates a manual change, rather than a config file change or default value

5. placeholder:

   • Represented as [some thing]