Skip to content

Commit

Permalink
chore(validator): MAPE and json generation
Browse files Browse the repository at this point in the history
This commit fixes MAPE calculation and its tests.
Additionally, the json generation is cleaned and added to `gen-report`
as well.

Signed-off-by: Sunil Thaha <[email protected]>
  • Loading branch information
sthaha committed Oct 24, 2024
1 parent 8770bf5 commit d0fe966
Show file tree
Hide file tree
Showing 4 changed files with 111 additions and 128 deletions.
134 changes: 64 additions & 70 deletions e2e/tools/validator/src/validator/cli/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,6 @@

logger = logging.getLogger(__name__)
pass_config = click.make_pass_decorator(config.Validator)
data_dict = {}


@dataclass
Expand Down Expand Up @@ -415,8 +414,6 @@ def stress(cfg: config.Validator, script_path: str, report_dir: str):
click.secho(" * Generating report dir and tag", fg="green")
results_dir, tag = create_report_dir(report_dir)
click.secho(f"\tresults dir: {results_dir}, tag: {tag}", fg="bright_green")
filepath = results_dir + "/" + tag + ".json"
data_dict.update({"file_path": filepath})

res = TestResult(tag)

Expand All @@ -437,7 +434,7 @@ def stress(cfg: config.Validator, script_path: str, report_dir: str):
time.sleep(10)

res.validations = run_validations(cfg, stress_test, results_dir)
create_json(res)
write_json_report(results_dir, res)
write_md_report(results_dir, res)


Expand Down Expand Up @@ -473,6 +470,7 @@ def gen_report(cfg: config.Validator, start: datetime.datetime, end: datetime.da
script_result = ScriptResult(start, end)
res.validations = run_validations(cfg, script_result, results_dir)

write_json_report(results_dir, res)
write_md_report(results_dir, res)


Expand Down Expand Up @@ -612,84 +610,80 @@ def validate_acpi(cfg: config.Validator, duration: datetime.timedelta, report_di
return int(res.validations.passed)


def create_json(res):
def update_list_json(new_value: list, new_key: str):
data_dict[new_key] = new_value
def write_json_report(results_dir: str, res: TestResult):
pattern = re.compile(r'[{]?(\w+)=("[^"]*"|[^,]+)[},]?')

def custom_encode(input_string):
pattern = re.compile(r'(\w+)=("[^"]*"|[^,]+)')
def extract_label_value(input_string):
matches = pattern.findall(input_string)
parsed_dict = {key: value.strip('"') for key, value in matches}
return json.dumps(parsed_dict)

result = []

for i in res.validations.results:
value = {}
if i.mse_passed:
value["mse"] = float(i.mse.value)
else:
value["mse"] = float(i.mse.error)
if i.mape_passed:
value["mape"] = float(i.mape.value)
else:
value["mape"] = float(i.mape.error)
if i.mae_passed:
value["mae"] = float(i.mae.value)
else:
value["mae"] = float(i.mae.error)
value["status"] = (
"mape passed: "
+ str(i.mape_passed)
+ ", mse passed: "
+ str(i.mse_passed)
+ ", mae passed: "
+ str(i.mae_passed)
)
m_name = i.name.replace(" - ", "_")

result.append({m_name: value})
return {key: value.strip('"') for key, value in matches}

data_dict = {}
results = []

for r in res.validations.results:
value = {
"mae": str(r.mae),
"mape": str(r.mape),
"mse": str(r.mse),
"status": (
"mape passed: "
+ str(r.mape_passed).lower()
+ ", mse passed: "
+ str(r.mse_passed).lower()
+ ", mae passed: "
+ str(r.mae_passed).lower()
),
}
results.append({r.name: value})

build_info = []
for i in res.build_info:
tmp = i.replace("kepler_exporter_build_info", "")
build_info.append(custom_encode(tmp))
for r in res.build_info:
selector = r.replace("kepler_exporter_build_info", "")
build_info.append(extract_label_value(selector))

node_info = []
for i in res.node_info:
tmp = i.replace("kepler_exporter_node_info", "")
node_info.append(custom_encode(tmp))
for r in res.node_info:
selector = r.replace("kepler_exporter_node_info", "")
node_info.append(extract_label_value(selector))

update_list_json(build_info, "build_info")
update_list_json(node_info, "node_info")
data_dict["build_info"] = build_info
data_dict["node_info"] = node_info

machine_spec = []
machine_spec.append(
machine_specs = []
machine_specs.append(
{
"type": "host",
"model": res.host_spec[0][0],
"cores": res.host_spec[0][1],
"threads": res.host_spec[0][2],
"sockets": res.host_spec[0][3],
"flags": res.host_spec[0][4],
"dram": res.host_spec[1],
}
)
machine_spec.append(
{
"type": "vm",
"model": res.vm_spec[0][0],
"cores": res.vm_spec[0][1],
"threads": res.vm_spec[0][2],
"sockets": res.vm_spec[0][3],
"flags": res.vm_spec[0][4],
"dram": res.vm_spec[1],
"model": res.host_spec.cpu_spec.model,
"cores": res.host_spec.cpu_spec.cores,
"threads": res.host_spec.cpu_spec.threads,
"sockets": res.host_spec.cpu_spec.sockets,
"flags": res.host_spec.cpu_spec.flags,
"dram": res.host_spec.dram_size,
}
)
update_list_json(machine_spec, "machine_spec")
if res.vm_spec is not None:
machine_specs.append(
{
"type": "vm",
"model": res.vm_spec.cpu_spec.model,
"cores": res.vm_spec.cpu_spec.cores,
"threads": res.vm_spec.cpu_spec.threads,
"sockets": res.vm_spec.cpu_spec.sockets,
"flags": res.vm_spec.cpu_spec.flags,
"dram": res.vm_spec.dram_size,
}
)

data_dict["machine_specs"] = machine_specs
data_dict["results"] = results

update_list_json(result, "result")
json_template = JsonTemplate(**data_dict)
file_name = data_dict["file_path"]
with open(file_name, "w") as file:

path = os.path.join(results_dir, f"{res.tag}.json")
with open(path, "w") as file:
json.dump(json_template, file, cls=CustomEncoder, indent=2)

#
# TODO: remove all the above in favor of below
# with open(file_name+"sane.json", "w") as file:
# json.dump(dataclasses.asdict(res), file, cls=CustomEncoder, indent=2)
17 changes: 1 addition & 16 deletions e2e/tools/validator/src/validator/prometheus/__init__.py
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,6 @@
from datetime import datetime
from typing import NamedTuple, Protocol

import numpy as np
import numpy.typing as npt
from prometheus_api_client import PrometheusConnect
from sklearn.metrics import mean_absolute_error, mean_absolute_percentage_error, mean_squared_error
Expand Down Expand Up @@ -91,20 +90,6 @@ class Result(NamedTuple):
mae: ValueOrError


def validate_arrays(actual: npt.ArrayLike, predicted: npt.ArrayLike) -> tuple[npt.ArrayLike, npt.ArrayLike]:
actual, predicted = np.array(actual), np.array(predicted)

if len(actual) != len(predicted):
msg = f"actual and predicted must be of equal length: {len(actual)} != {len(predicted)}"
raise ValueError(msg)

if len(actual) == 0 or len(predicted) == 0:
msg = f"actual ({len(actual)}) and predicted ({len(predicted)}) must not be empty"
raise ValueError(msg)

return (actual, predicted)


def mse(actual: npt.ArrayLike, predicted: npt.ArrayLike) -> ValueOrError:
try:
return ValueOrError(value=mean_squared_error(actual, predicted))
Expand All @@ -116,7 +101,7 @@ def mse(actual: npt.ArrayLike, predicted: npt.ArrayLike) -> ValueOrError:

def mape(actual: npt.ArrayLike, predicted: npt.ArrayLike) -> ValueOrError:
try:
return ValueOrError(value=mean_absolute_percentage_error(actual, predicted))
return ValueOrError(value=mean_absolute_percentage_error(actual, predicted) * 100)

# ruff: noqa: BLE001 (Suppressed as we want to catch all exceptions here)
except Exception as e:
Expand Down
56 changes: 29 additions & 27 deletions e2e/tools/validator/src/validator/report/__init__.py
Original file line number Diff line number Diff line change
@@ -1,3 +1,4 @@
import datetime
import json
from typing import Any

Expand Down Expand Up @@ -33,49 +34,50 @@ def __repr__(self):
class JsonTemplate:
def __init__(
self,
file_path: str,
build_info: list[Any],
node_info: list[Any],
machine_spec: list[Any],
result: list[dict[str, Any]],
machine_specs: list[Any],
results: list[dict[str, Any]],
):
if build_info is None:
build_info = []
if node_info is None:
node_info = []
if machine_spec is None:
machine_spec = []
if result is None:
result = []

self.file_path = file_path
self.build_info = build_info
self.node_info = node_info
self.machine_spec = machine_spec
self.result = []
for res in result:
self.machine_specs = machine_specs
self.results = []
for res in results:
for key, value in res.items():
self.result.append(Result(key, value))
self.results.append(Result(key, value))

def to_dict(self):
return {
"file_path": self.file_path,
"build_info": self.build_info,
"node_info": self.node_info,
"machine_spec": self.machine_spec,
"result": [res.to_dict() for res in self.result],
"machine_specs": self.machine_specs,
"results": [res.to_dict() for res in self.results],
}

def __repr__(self):
return (
f"JsonTemplate(file_path='{self.file_path}', build_info={self.build_info}, "
f"node_info={self.node_info}, machine_spec={self.machine_spec}, "
f"result={self.result})"
f"JsonTemplate('build_info={self.build_info}, "
f"node_info={self.node_info}, machine_spec={self.machine_specs}, "
f"result={self.results})"
)


class CustomEncoder(json.JSONEncoder):
def default(self, obj):
if hasattr(obj, "to_dict"):
return obj.to_dict()
return super().default(obj)
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)

def default(self, o):
if hasattr(o, "_asdict"):
return o._asdict()

if hasattr(o, "to_dict"):
return o.to_dict()

if type(o) == datetime.datetime:
return o.isoformat()

if type(o).__name__ == "bool":
return str(o).lower()

return super().default(o)
32 changes: 17 additions & 15 deletions e2e/tools/validator/tests/validator/prometheus/test_prometheus.py
Original file line number Diff line number Diff line change
Expand Up @@ -155,41 +155,43 @@ def test_mse():
"a": [ 1, 2, 3, ],
"b": [ 4, 5, 6, ],
"mse": 9.0, # (1 - 4)^2 + (2 - 5)^2 + (3 - 6)^2 / 3
"mape": 1.833333,
"mape": 183.3333,
"mae": 3.0, # (|1-4| + |2-5| + |3-6|) / 3
}, {
"a": [ 1.5, 2.5, 3.5 ],
"b": [ 1.0, 2.0, 3.0 ],
"mse": 0.25, # 3 x (0.5^2) / 3
"mape": 0.225396,
"mape": 22.5396,
"mae": 0.5, # |1.5 - 1.0| + |2.5 - 2.0| + |3.5 - 3.0|
}, {
"a": [ 1, -2, 3 ],
"b": [ -1, 2, -3 ],
"mse": 18.6666, # 2.0^2 + 4.0^2 + 6.0^2 / 3
"mape": 2.000,
"mape": 200.0,
"mae": 4.0 # (|1-(-1)| + |-2-2| + |3-(-3)|) / 3
}]
# fmt: on

for s in inputs:
for a, b in ([s["a"], s["b"]], [s["b"], s["a"]]):
expected_mse = s["mse"]
actual_mse = mse(a, b)
assert actual_mse.error is None
assert pytest.approx(actual_mse.value, rel=1e-3) == expected_mse

actual_mae = mae(a, b)
assert actual_mae.error is None
expected_mae = s["mae"]
assert pytest.approx(actual_mae.value, rel=1e-3) == expected_mae

# NOTE: MAPE(a , b) != MAPE(b, a) unlike MSE and MAE
for s in inputs:
a = s["a"]
b = s["b"]

expected_mse = s["mse"]
actual_mse = mse(a, b)
assert actual_mse.error is None
assert expected_mse == pytest.approx(actual_mse.value, rel=1e-3)

actual_mape = mape(a, b)
assert actual_mape.error is None
expected_mape = s["mape"]
assert expected_mape == pytest.approx(actual_mape.value, rel=1e-3)

actual_mae = mae(a, b)
assert actual_mae.error is None
expected_mae = s["mae"]
assert expected_mae == pytest.approx(actual_mae.value, rel=1e-3)
assert pytest.approx(actual_mape.value, rel=1e-3) == expected_mape


def test_mse_with_large_arrays():
Expand Down

0 comments on commit d0fe966

Please sign in to comment.