fix: isolate simulator predictor cache
This commit is contained in:
@@ -611,6 +611,12 @@ def main() -> None:
|
||||
command = configure_stage_batch_ledger(
|
||||
command, enabled=args.store_stage_batch_ledger
|
||||
)
|
||||
command.extend(
|
||||
[
|
||||
"--metrics_config_cache_dir",
|
||||
str(args.output_root / "cache"),
|
||||
]
|
||||
)
|
||||
if args.align_real_graph_runtime:
|
||||
command.extend(
|
||||
[
|
||||
@@ -640,7 +646,7 @@ def main() -> None:
|
||||
)
|
||||
fallback_evidence = (
|
||||
collective_fallback_evidence(run_dir)
|
||||
if args.cc_backend == "vidur"
|
||||
if args.cc_backend == "vidur" and config.tp > 1
|
||||
else []
|
||||
)
|
||||
if completed.returncode == 0 and fallback_evidence:
|
||||
@@ -823,6 +829,7 @@ def main() -> None:
|
||||
"arrival": "original_trace_timestamp_and_order",
|
||||
"input_output": "exact_source_values",
|
||||
"store_stage_batch_ledger": args.store_stage_batch_ledger,
|
||||
"predictor_cache_root": str(args.output_root / "cache"),
|
||||
"ttft_slo": "1000ms + 1000ms * input_tokens / 8000",
|
||||
"tpot_slos_ms": TPOT_SLOS_MS,
|
||||
"primary_tpot_slo_ms": 150.0,
|
||||
|
||||
Reference in New Issue
Block a user