[Model] Drop the 4x M-RoPE cache headroom for Qwen3-VL, Qwen3.5 and Qwen3.8 (#59853)

Signed-off-by: lesj0610 <lesj0610@users.noreply.github.com>
Signed-off-by: lesj0610 <lesj0610@gmail.com>
Co-authored-by: lesj0610 <lesj0610@users.noreply.github.com>
This commit is contained in:
SeongJun Lee
2026-10-10 06:51:52 +00:00
committed by GitHub
co-authored by lesj0610
parent c1692c85f1
commit a98247ab4d
11 changed files with 56 additions and 6 deletions
+4
View File
@@ -508,6 +508,10 @@ class ModelConfig:
# here early.
if self.multimodal_config:
factors["language_model_only"] = self.multimodal_config.language_model_only
# Sizes Qwen3.5's M-RoPE cache.
factors["video_pruning"] = (
self.multimodal_config.is_multimodal_pruning_enabled()
)
return hash_factors(factors)
def _update_nested(