iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[Bugfix] Move extract_layer_index back inside is_v32 guard (#46199)

Browse Source 
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Tyler Michael Smith
2026-06-20 21:19:10 +00:00
committed by GitHub
co-authored by Claude Opus 4.6
parent 891cc4b9c5
commit 77148992cf
1 changed files with 22 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/model_executor/models/deepseek_v2.py
+22 -16
View File
@@ -1001,24 +1001,30 @@ class DeepseekV2MLAAttention(nn.Module):
# IndexCache config
# Refer: https://arxiv.org/abs/2603.12201 for more details.
_skip_topk = False
_index_topk_freq = getattr(config, "index_topk_freq", 1)
_index_topk_pattern = getattr(config, "index_topk_pattern", None)
_index_skip_topk_offset = getattr(config, "index_skip_topk_offset", 2)
layer_id = extract_layer_index(prefix)
is_mtp_layer = False
if self.is_v32:
_index_topk_freq = getattr(config, "index_topk_freq", 1)
_index_topk_pattern = getattr(config, "index_topk_pattern", None)
_index_skip_topk_offset = getattr(config, "index_skip_topk_offset", 2)
layer_id = extract_layer_index(prefix)
if _index_topk_pattern is None:
_skip_topk = (
max(layer_id - _index_skip_topk_offset + 1, 0) % _index_topk_freq != 0
if _index_topk_pattern is None:
_skip_topk = (
max(layer_id - _index_skip_topk_offset + 1, 0) % _index_topk_freq
!= 0
)
elif 0 <= layer_id < len(_index_topk_pattern):
_skip_topk = _index_topk_pattern[layer_id] == "S"
# The skip pattern only governs backbone layers. MTP/nextn
# layers (layer_id >= num_hidden_layers) always build a full
# indexer: they compute indices at draft step 0 and toggle
# at runtime via set_skip_topk
# (index_share_for_mtp_iteration).
_num_hidden_layers = getattr(config, "num_hidden_layers", None)
is_mtp_layer = (
_num_hidden_layers is not None and layer_id >= _num_hidden_layers
)
elif 0 <= layer_id < len(_index_topk_pattern):
_skip_topk = _index_topk_pattern[layer_id] == "S"
# The skip pattern only governs backbone layers. MTP/nextn layers
# (layer_id >= num_hidden_layers) always build a full indexer: they
# compute indices at draft step 0 and toggle at runtime via
# set_skip_topk (index_share_for_mtp_iteration).
_num_hidden_layers = getattr(config, "num_hidden_layers", None)
is_mtp_layer = _num_hidden_layers is not None and layer_id >= _num_hidden_layers
if self.is_v32 and (not _skip_topk or is_mtp_layer):
self.indexer_rope_emb = get_rope(