- Python 3
- Python packages, can be installed with
pip install -r requirements.txtnix-shell// example: parse.json
{
"name": "benchmarks",
"dir": {
"data": "./data",
"output": "./output"
},
"data": [
{
"name": "1T-16MB-1",
"file": "./raw_data/randomize/1T-16MB/1/ucare.log",
"old_format": true
},
{
"name": "3_30006",
"file": "./raw_data/stringtable/3_30006/ucare.log",
"old_format": true
},
{
"name": "3_120026",
"file": "./raw_data/stringtable/3_120026/ucare.log",
"old_format": true
},
{
"name": "renaissance",
"file": "./raw_data/benchmarks/renaissance/ucare.log"
},
{
"name": "dacapo",
"file": "./raw_data/benchmarks/dacapo/ucare.log"
},
{
"name": "specjvm",
"file": "./raw_data/benchmarks/specjvm/ucare.log"
}
]
}old_formatkey is for backward compatibility with old version ofucare.log
// example: training.json
{
"name": "benchmarks",
"skip_value": 7,
"sm_add_constant": false,
"dir": {
"data": "./data/benchmarks",
"output": "./output"
},
"models": [
"ransac",
"lreg"
],
"subtitle": "",
"data": {
"main": [
"1T-16MB-1",
"..."
],
"stringtable": [
"1",
"..."
]
}
}modelskey can beransac,lreg, andsvrdataconsists of two key which entries will be prepended bydir/datakey :mainstringtable
// example: inference.json
{
"name": "benchmarks",
"skip_value": 0,
"sm_add_constant": false,
"dir": {
"data": "./data/benchmarks",
"output": "./output"
},
"model": {
"main": {
"name": "ransac",
"file": "./output/benchmarks/train/main/model/ransac.joblib"
},
"stringtable": {
"name": "ransac",
"file": "./output/benchmarks/train/stringtable/model/ransac.joblib"
}
},
"combined_plot": {
"max": 500,
"min": -500,
"subtitle": "Heap Size 4G"
},
"data": [
{
"name": "renaissance",
"color": "green",
"label": "Renaissance",
"subtitle": ""
},
{
"name": "dacapo",
"color": "blue",
"label": "DaCapo",
"subtitle": ""
},
{
"name": "specjvm",
"color": "red",
"label": "Specjvm2008",
"subtitle": ""
}
]
}dir_outputwill be appended withnamekey
the output dir will be
${dir_output}/${name}
skip_valueis number of value that will be skipped in cdf plot (in case of anomaly)
python parse.py -c <parse.json>python train.py \
-c <train.json> -t [main|stringtable]
# equals with
python train.py \
--config <train.json> --type [main|stringtable]python inference.py -c <inference.json>- Ray Andrew
- Cesar Stuardo