wehub-resource-sync
89 lines
3.3 KiB
Python
89 lines
3.3 KiB
Python
import math
|
|
from dataclasses import dataclass
|
|
|
|
from mlflow.entities._mlflow_object import _MlflowObject
|
|
from mlflow.protos.service_pb2 import CalculateTraceFilterCorrelation
|
|
|
|
|
|
@dataclass
|
|
class TraceFilterCorrelationResult(_MlflowObject):
|
|
"""
|
|
Result of calculating correlation between two trace filter conditions.
|
|
|
|
This class represents the correlation analysis between two trace filters,
|
|
using Normalized Pointwise Mutual Information (NPMI) as the correlation metric.
|
|
|
|
NPMI ranges from -1 to 1:
|
|
- -1: Perfect negative correlation (filters never co-occur)
|
|
- 0: Independence (filters occur independently)
|
|
- 1: Perfect positive correlation (filters always co-occur)
|
|
- NaN: Undefined (when one or both filters have zero matches)
|
|
|
|
Attributes:
|
|
npmi: Normalized Pointwise Mutual Information score (unsmoothed).
|
|
Returns NaN when undefined (e.g., when filter1_count=0 or filter2_count=0).
|
|
Returns -1.0 when filters never co-occur but both have support.
|
|
Otherwise returns a value in [-1, 1].
|
|
npmi_smoothed: NPMI calculated with Jeffreys prior smoothing (alpha=0.5).
|
|
More robust for small sample sizes and confidence interval estimation.
|
|
Returns NaN when undefined.
|
|
filter1_count: Number of traces matching the first filter.
|
|
filter2_count: Number of traces matching the second filter.
|
|
joint_count: Number of traces matching both filters.
|
|
total_count: Total number of traces in the experiment(s).
|
|
confidence_lower: Lower bound of the confidence interval for NPMI (optional).
|
|
confidence_upper: Upper bound of the confidence interval for NPMI (optional).
|
|
"""
|
|
|
|
npmi: float
|
|
filter1_count: int
|
|
filter2_count: int
|
|
joint_count: int
|
|
total_count: int
|
|
npmi_smoothed: float | None = None
|
|
confidence_lower: float | None = None
|
|
confidence_upper: float | None = None
|
|
|
|
@classmethod
|
|
def from_proto(cls, proto):
|
|
"""
|
|
Create a TraceFilterCorrelationResult from a protobuf response.
|
|
|
|
Args:
|
|
proto: CalculateTraceFilterCorrelation.Response protobuf message
|
|
|
|
Returns:
|
|
TraceFilterCorrelationResult instance
|
|
"""
|
|
return cls(
|
|
npmi=proto.npmi if proto.HasField("npmi") else float("nan"),
|
|
npmi_smoothed=proto.npmi_smoothed if proto.HasField("npmi_smoothed") else None,
|
|
filter1_count=proto.filter1_count,
|
|
filter2_count=proto.filter2_count,
|
|
joint_count=proto.joint_count,
|
|
total_count=proto.total_count,
|
|
)
|
|
|
|
def to_proto(self):
|
|
"""
|
|
Convert this result to a protobuf response message.
|
|
|
|
Returns:
|
|
CalculateTraceFilterCorrelation.Response protobuf message
|
|
"""
|
|
|
|
response = CalculateTraceFilterCorrelation.Response()
|
|
|
|
if self.npmi is not None and not math.isnan(self.npmi):
|
|
response.npmi = self.npmi
|
|
|
|
if self.npmi_smoothed is not None and not math.isnan(self.npmi_smoothed):
|
|
response.npmi_smoothed = self.npmi_smoothed
|
|
|
|
response.filter1_count = self.filter1_count
|
|
response.filter2_count = self.filter2_count
|
|
response.joint_count = self.joint_count
|
|
response.total_count = self.total_count
|
|
|
|
return response
|