stormlog.tensorflow.analyzer

TensorFlow Memory Analysis

Classes

MemoryAnalyzer([sensitivity, ...])

Advanced TensorFlow memory analysis and optimization.

class stormlog.tensorflow.analyzer.MemoryAnalyzer(sensitivity=0.05, collective_sensitivity='medium', collective_threshold_overrides=None)[source]

Bases: object

Advanced TensorFlow memory analysis and optimization.

Parameters:

sensitivity (float)
collective_sensitivity (str)
collective_threshold_overrides (Mapping[str, Any] | None)

detect_memory_leaks(tracking_results)[source]

Detect potential memory leaks using statistical analysis.

Parameters:: tracking_results (Any)
Return type:: List[Dict[str, Any]]

analyze_fragmentation(profile_result)[source]

Analyze memory fragmentation patterns.

Parameters:: profile_result (Any)
Return type:: Dict[str, float]

detect_patterns(tracking_results)[source]

Detect memory usage patterns.

Parameters:: tracking_results (Any)
Return type:: List[Dict[str, Any]]

analyze_efficiency(profile_result)[source]

Analyze memory usage efficiency (0-10 scale).

Parameters:: profile_result (Any)
Return type:: float

correlate_with_performance(profile_result)[source]

Correlate memory usage with performance metrics.

Parameters:: profile_result (Any)
Return type:: Dict[str, Any]

analyze_memory_gaps(events, *, phase_resolver=None)[source]

Classify allocator-vs-device hidden memory gaps over time.

Parameters:

events (List[TelemetryEventV2]) – Chronologically ordered telemetry samples.
phase_resolver (PhaseReplayIndex | None)

Returns:

Prioritized list of gap findings (severity desc, confidence desc).

Return type:

List[GapFinding]

analyze_collective_attribution(events, *, phase_resolver=None)[source]

Attribute hidden-memory spikes to collective communication phases.

Parameters:

events (List[TelemetryEventV2])
phase_resolver (PhaseReplayIndex | None)

Return type:

List[CollectiveAttributionResult]

score_optimization(profile_result, events=None)[source]

Score optimization opportunities.

Parameters:

profile_result (Any) – TensorFlow profiling result object.
events (List[TelemetryEventV2] | None) – Optional telemetry event series for gap analysis. When provided, the result includes a gap_analysis section.

Return type:

Dict[str, Any]