vllm.model_executor.model_loader.weight_cache.warmup_cache.base

Warmup cache lifecycle interface.

Classes:

• WarmupCache –

  One daemon warmup cache and its cache-specific D2D semantics.

WarmupCache

Bases: Seedable

One daemon warmup cache and its cache-specific D2D semantics.

Source code in vllm/model_executor/model_loader/weight_cache/warmup_cache/base.py

class WarmupCache(Seedable):
    """One daemon warmup cache and its cache-specific D2D semantics."""

    def adopt_before_load(
        self, daemon: "WeightCacheDaemon", transport: "SeedTransport"
    ) -> None:
        pass

    def warmup(
        self, daemon: "WeightCacheDaemon", transport: "SeedTransport"
    ) -> None: ...