[GLM5.3 Perf] Size the GLM-5 sparse indexer decode workspace, 3072 MiB GPU memory saved (#57701)

Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye
2026-09-19 16:57:01 -04:00
committed by GitHub
parent 133b71e0be
commit 36fa72d2d0
3 changed files with 17 additions and 17 deletions
+2 -2
View File
@@ -296,7 +296,7 @@ class Indexer(nn.Module):
cache_config=cache_config,
index_kpool=self.index_kpool,
)
self.max_model_len = vllm_config.model_config.max_model_len
self.max_pool_len = vllm_config.model_config.max_model_len // self.index_kpool
self.prefix = prefix
from vllm.v1.attention.backends.mla.indexer import get_max_prefill_buffer_size
@@ -307,7 +307,7 @@ class Indexer(nn.Module):
self.scale_fmt,
self.topk_tokens,
self.head_dim,
self.max_model_len,
self.max_pool_len,
self.max_total_seq_len,
self.topk_indices_buffer,
tail_cache=self.tail_cache,