vllm kvprune wo:v1.1.1

d4645504 · chenzk · d29c39ca · d29c39ca · d29c39ca · d29c39ca
Commit d4645504 authored Apr 30, 2026 by chenzk
20 changed files
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/tensor_details/layout_details/hopper_scale.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/tensor_details/layout_details/hopper_scale.py
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/tensor_details/layout_details/hopper_value.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/tensor_details/layout_details/hopper_value.py
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/tensor_details/layout_details/strided.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/tensor_details/layout_details/strided.py
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/testing.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/testing.py
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk.py
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk_details/__init__.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk_details/__init__.py
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk_details/_topk_backward.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk_details/_topk_backward.py
--- a/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk_details/_topk_forward.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/triton_kernels/topk_details/_topk_forward.py
--- a/vllm/compactor-vllm/src/compactor_vllm/utils/__init__.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/utils/__init__.py
--- a/vllm/compactor-vllm/src/compactor_vllm/utils/arguments.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/utils/arguments.py
--- a/vllm/compactor-vllm/src/compactor_vllm/utils/context.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/utils/context.py
--- a/vllm/compactor-vllm/src/compactor_vllm/utils/helpers.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/utils/helpers.py
--- a/vllm/compactor-vllm/src/compactor_vllm/utils/sequence.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/utils/sequence.py
--- a/vllm/compactor-vllm/src/compactor_vllm/utils/triton_compat.py
+++ b/vllm/compactor-vllm/src/compactor_vllm/utils/triton_compat.py
--- a/vllm/compactor-vllm/tests/test_store_kv.py
+++ b/vllm/compactor-vllm/tests/test_store_kv.py
--- a/vllm/compactor-vllm/tests/test_triton_attention.py
+++ b/vllm/compactor-vllm/tests/test_triton_attention.py
--- a/vllm/compactor-vllm/vllm_memory_comparison.png
+++ b/vllm/compactor-vllm/vllm_memory_comparison.png
--- a/vllm/compactor-vllm/vllm_throughput_comparison.png
+++ b/vllm/compactor-vllm/vllm_throughput_comparison.png
--- a/vllm/kvprune_legacy_save/__init__.py
+++ b/vllm/kvprune_legacy_save/__init__.py
--- a/vllm/kvprune_legacy_save/attention/__init__.py
+++ b/vllm/kvprune_legacy_save/attention/__init__.py