Closed j2whiting closed 6 months ago
We have a first pass for this data structure, but there are duplicate triplets which are problematic for the downstream configuration tasks.
TODO:
Data Structure Sample
[
{
"id": "N",
"subject": "S_1",
"outcome": "E_1",
"controllers": [
"I_1"
]
},
{
"id": "d",
"subject": "S_1",
"outcome": "E_1",
"controllers": [
"I_1"
]
},
{
"id": "c",
"subject": "S_1",
"outcome": "E_1",
"controllers": [
"I_1"
]
},
{
"id": "e",
"subject": "S_1",
"outcome": "E_1",
"controllers": [
"I_1"
]
},
{
"id": "f_0",
"subject": "S_1",
"outcome": "E_1",
"controllers": [
"I_1"
]
},
{
"id": "f_1",
"subject": "S_1",
"outcome": "E_1",
"controllers": [
"I_2"
]
},
{
"id": "f_2",
"subject": "S_1",
"outcome": "E_1",
"controllers": [
"I_3"
]
},
{
"id": "f_3",
"subject": "S_2",
"outcome": "E_2",
"controllers": [
"I_2"
]
},
{
"id": "f_4",
"subject": "S_2",
"outcome": "E_2",
"controllers": [
"I_1"
]
},
{
"id": "f_5",
"subject": "S_2",
"outcome": "E_2",
"controllers": [
"I_3"
]
},
{
"id": "f_6",
"subject": "S_3",
"outcome": "E_3",
"controllers": [
"I_3"
]
},
{
"id": "f_7",
"subject": "S_3",
"outcome": "E_3",
"controllers": [
"I_1"
]
},
{
"id": "f_8",
"subject": "S_3",
"outcome": "E_3",
"controllers": [
"I_2"
]
},
{
"id": "a",
"subject": "E_1",
"outcome": "I_1",
"controllers": []
},
{
"id": "b",
"subject": "I_1",
"outcome": "R_1",
"controllers": []
}
]
AMR strings sent to GoLLM from HMI will now contain mapping metadata which will be formatted into the prompt.
The dataset -> config chain prompt has been updated on the GoLLM side to utilize the metadata.
@mwdchang is this ready to be closed as complete?
We need to account for 3 different forms of parameter matrices to reduce ambiguity in configuration pipelines:
Update output from HMI and the GoLLM prompts with few shot examples.