vLLM v0.25.0 makes Model Runner V2 the default and drops PagedAttention
vLLM v0.25.0 ships MRv2 as the default execution path for all dense models and permanently removes the legacy PagedAttention implementation.
vLLM v0.25.0 ships MRv2 as the default execution path for all dense models and permanently removes the legacy PagedAttention implementation.