SolitaryThinker
0047d54a1b
First step toward lean pooled run-to-completion serving: rip out the GPU-time cost/budget machinery entirely (it priced nothing useful for the target design). Removed: CostModel + LoopSpec.step_cost_model; ResourceRequest.compute_seconds; StepResult.actual_seconds; AdmissionController's compute budget + SchedulerMetrics .gpu_seconds (the memory/OOM reservation guard stays); the Profiler observer (cost calibration); per-step timing in RuntimeLoopContext; cost-based fleet/Dynamo routing (now a coarse step-count load proxy); DeploymentCard.cost_model. Stripped cost from all 9 recipe cards + their loops. Tests: dropped the 3 cost-specific tests (cost routing, cost_model aliasing, compute-budget gate); refocused 2 (loop cache validation, NaNWatch-clean). CPU mini: 151 passed / 2 skipped. Interleave removal + pooled serving (P2) and the native omni port (P3) follow. (pyproject kernel hack excluded as always.)
86 lines
3.9 KiB
Python
86 lines
3.9 KiB
Python
"""RuntimeLoopContext — the concrete inversion point.
|
|
|
|
This is the single seam: ``execute(plan)`` runs the kernel thunk built by the model's
|
|
``next()``, wrapped by interceptors (compute-altering) and observers (read-only). The model
|
|
never sees the scheduler; the scheduler never sees the model's math. Per-request loop state
|
|
is bound here so interceptors read state from ``LoopState`` (never module globals).
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
from typing import Any
|
|
|
|
from v2._enums import ExecutionProfile
|
|
from v2.loop.contracts import StepResult, WorkPlan
|
|
from v2.request.streams import OmniEvent, Stream, StreamChunk
|
|
|
|
|
|
class RuntimeLoopContext:
|
|
|
|
def __init__(self, instance: Any, *, observers, interceptors, slots: dict[str, Any], stream: Stream, cancel_scope,
|
|
profile: ExecutionProfile, metrics: dict, request_id: str):
|
|
self.instance = instance
|
|
self.observers = observers
|
|
self.interceptors = interceptors
|
|
self.slots = slots # the program's typed named edges
|
|
self.stream = stream
|
|
self.cancel_scope = cancel_scope
|
|
self.profile = profile
|
|
self.metrics = metrics
|
|
self.request_id = request_id
|
|
self.state = None # bound per step by the LoopRunner
|
|
|
|
def bind_state(self, state) -> None:
|
|
self.state = state
|
|
|
|
def execute(self, plan: WorkPlan) -> StepResult:
|
|
self.observers.emit("step_start", plan=plan)
|
|
# Interceptors may supply an override for the *expensive forward* (e.g. a cached
|
|
# prediction). The step body still runs the cheap scheduler step (combine + solver),
|
|
# so the override changes compute, not the loop's state-transition shape.
|
|
override = None
|
|
if self.interceptors.active and self.state is not None:
|
|
override = self.interceptors.before(plan, self.state)
|
|
# only count a skip the step body will actually honor (must carry a forward result)
|
|
if override is not None and "noise_pred" in override:
|
|
self.metrics["skipped_steps"] = self.metrics.get("skipped_steps", 0) + 1
|
|
|
|
def _eager(ov):
|
|
return plan.run(self.instance, ov) if plan.run is not None else {}
|
|
|
|
out = self._run_step(plan, override, _eager)
|
|
result = out if isinstance(out, StepResult) else StepResult(output=dict(out))
|
|
|
|
if self.interceptors.active and self.state is not None:
|
|
self.interceptors.after(plan, self.state, result)
|
|
|
|
self.observers.emit("step_complete", plan=plan, result=result)
|
|
self.metrics["steps"] = self.metrics.get("steps", 0) + 1
|
|
return result
|
|
|
|
def _run_step(self, plan: WorkPlan, override, eager):
|
|
"""Execute the step body, under piecewise CUDA-graph capture/replay when the loop declares
|
|
graph_capture="breakable_cudagraph". Otherwise plain eager — the unchanged path. Replay
|
|
still runs the current thunk (CPU models the lifecycle, not the speedup); see
|
|
runtime/cudagraph.py."""
|
|
if plan.run is None:
|
|
return {}
|
|
loop = self.instance.card.loops.get(plan.loop_id) if self.instance.card else None
|
|
if getattr(loop, "graph_capture", "eager") != "breakable_cudagraph":
|
|
return eager(override)
|
|
if self.instance.graphs is None:
|
|
from v2.runtime.cudagraph import GraphCapturer
|
|
self.instance.graphs = GraphCapturer()
|
|
return self.instance.graphs.dispatch(plan, self.instance, override, eager)
|
|
|
|
def emit(self, chunk: StreamChunk) -> None:
|
|
self.stream.emit(
|
|
OmniEvent(type="media.chunk", request_id=self.request_id, seq=chunk.seq, payload={"chunk": chunk}))
|
|
self.metrics["stream_chunks"] = self.metrics.get("stream_chunks", 0) + 1
|
|
|
|
def check_cancel(self) -> None:
|
|
if self.cancel_scope is not None:
|
|
self.cancel_scope.check() # raises request.Cancelled at the step boundary
|
|
|
|
def observe(self, event: str, **kw) -> None:
|
|
self.observers.emit(event, **kw)
|