[3/N] HiSparse: host-resident sparse-MLA decode hot-buffering (#53781)

Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Signed-off-by: Matthew Bonanni <matthew@neuralmagic.com>
Signed-off-by: Nicolò Lucchesi <nicolo.lucchesi@mistral.ai>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Lucas Wilkinson <lwilkinson@neuralmagic.com>
Co-authored-by: Nicolò Lucchesi <nicolo.lucchesi@mistral.ai>
Co-authored-by: OpenAI Codex <noreply@openai.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
This commit is contained in:
Matthew Bonanni
2026-09-12 00:36:47 -04:00
committed by GitHub
co-authored by OpenAI Codex Lucas Wilkinson Lucas Wilkinson Nicolò Lucchesi OpenAI Codex Claude Opus 5 Lucas Wilkinson
parent 120ec4ebd2
commit d43bb2f37f
124 changed files with 14262 additions and 906 deletions
+2
View File
@@ -34,6 +34,7 @@ from vllm.platforms import current_platform
from vllm.sequence import IntermediateTensors
from vllm.utils.math_utils import round_up
from vllm.utils.torch_utils import current_stream
from vllm.v1.hisparse.binding import release_hisparse_profiling_cache
from vllm.v1.kv_cache_interface import KVCacheConfig
from vllm.v1.spec_decode.dynamic.utils import build_dynamic_sd_schedule_lookup
from vllm.v1.worker.gpu.attn_utils import build_slot_mappings_by_layer
@@ -931,6 +932,7 @@ def _teardown_profiling_state(runner: "GPUModelRunner") -> None:
if (model := getattr(runner, "model", None)) is not None:
layers = itertools.chain(layers, model.modules())
clear_layer_kv_caches(layers)
release_hisparse_profiling_cache(runner.compilation_config.static_forward_context)
runner.cache_config.num_gpu_blocks = None
runner.maybe_remove_all_loras(runner.lora_config)
gc.collect()