biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V0 Deprecation] Deprecate virtual engine (#37195)

Browse Source 
Signed-off-by: yewentao256 <zhyanwentao@126.com>
This commit is contained in:
Wentao Ye
2026-03-18 17:30:14 -04:00
committed by GitHub
parent 6ae4c8d6fc
commit 0d81a1fe61
23 changed files with 23 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
tests/compile/passes/test_rope_kvcache_fusion.py
View File

@@ -295,7 +295,7 @@ def test_rope_kvcache_fusion(
}
q_unfused, k_unfused, v_unfused, dummy = model(qkv_unfused, pos_unfused)
attn_layer = forward_context.no_compile_layers[model.layer_name]
kv_cache_unfused = attn_layer.kv_cache[forward_context.virtual_engine]
kv_cache_unfused = attn_layer.kv_cache[0]
del dummy
torch._dynamo.mark_dynamic(qkv, 0)
@@ -309,7 +309,7 @@ def test_rope_kvcache_fusion(
}
q_fused, k_fused, v_fused, dummy = model_fused(qkv, pos)
attn_layer = forward_context.no_compile_layers[model.layer_name]
kv_cache_fused = attn_layer.kv_cache[forward_context.virtual_engine]
kv_cache_fused = attn_layer.kv_cache[0]
del dummy
assert fusion_pass.matched_count == 1