fr-analysis

Skill: fr_analysis

Analyze PyTorch NCCL flight-recorder (FR) dumps to identify the collective operation hang and isolate the ranks responsible, using CollectiveAnalyzer.

Script: scripts/fr_attribution.py → attribution/trace_analyzer/fr_attribution.py

What it does

1. Loads all FR dump files matching a glob pattern under --fr-path.
2. Parses each dump into Collective records (op type, ranks, process group, timing, state).
3. Groups collectives by process group and sequence ID across ranks to detect mismatches.
4. Identifies the wavefront — the process group boundary where collectives diverge — and returns the missing ranks at that boundary as the root-cause suspects.
5. Optionally runs an LLM pass (--llm-analyze) over the structured findings for a human-readable summary.

CLI

python scripts/fr_attribution.py \
    --fr-path /path/to/fr_dumps/ \
    [-p "_dump_*"] \
    [--verbose] \
    [--health-check] \
    [--llm-analyze] \
    [--model MODEL] \
    [--debug]

Programmatic API

from nvidia_resiliency_ext.attribution.trace_analyzer.fr_attribution import CollectiveAnalyzer

analyzer = CollectiveAnalyzer({
    "fr_path": "/path/to/fr_dumps/",
    "pattern": "_dump_*",
    "verbose": False,
    "health_check": False,
    "llm_analyze": False,
    "model": "nvidia/nemotron-3-super-120b-a12b",
})
results = analyzer.run_sync({
    "fr_path": "/path/to/fr_dumps/",
})
# results: tuple[FRAnalysisResult | str, AttributionState]

Output

Returns (result, AttributionState) where result is the FR analysis table and describes:

• The selected wavefront/front process group
• Missing ranks at that process group (root-cause suspects)
• Per-rank collective status tables (when --verbose)
• Node health summary (when --health-check)
• LLM narrative (when --llm-analyze)

AttributionState.STOP indicates the hang is unrecoverable; CONTINUE indicates the job may be restartable after isolating the identified ranks.

Dump file formats

FR dumps are typically written to the directory specified by TORCH_NCCL_DEBUG_INFO_TEMP_FILE or triggered automatically on NCCL timeout.

Prerequisites

• FR dump files produced by PyTorch NCCL (set TORCH_NCCL_TRACE_BUFFER_SIZE > 0)
• LLM_API_KEY required only when using --llm-analyze
• langchain-openai required only when using --llm-analyze
• FR_DEBUG=1 env var enables verbose debug logging in the script