perf: optimize TypeBasedDispatcher using dict for O(1) lookup (#12001)

This commit is contained in:
xlzheng
2025-11-15 22:02:42 +08:00
committed by GitHub
parent 10592e9c08
commit 37e8724ef3
4 changed files with 513 additions and 3 deletions
+2
View File
@@ -422,6 +422,7 @@ suite_amd = {
TestFile("quant/test_fused_rms_fp8_group_quant.py", 10),
TestFile("rl/test_update_weights_from_disk.py", 210),
TestFile("test_abort.py", 51),
TestFile("test_bench_typebaseddispatcher.py", 10),
TestFile("test_chunked_prefill.py", 410),
TestFile("test_create_kvindices.py", 2),
TestFile("test_eval_fp8_accuracy.py", 303),
@@ -454,6 +455,7 @@ suite_amd = {
# TestFile("test_triton_attention_kernels.py", 4),
TestFile("test_triton_attention_backend.py", 150),
TestFile("test_triton_sliding_window.py", 250),
TestFile("test_type_based_dispatcher.py", 10),
TestFile("test_wave_attention_kernels.py", 2),
# Disabled temporarily
# TestFile("test_vlm_input_format.py", 300),