Enable Flashinfer TRTLLM-GEN-MoE FP8 blockwise kernel for Qwen3-Next on Blackwell (#12543)

2025-11-13 19:44:44 +08:00
parent aead0ef5e5
commit e7e89349c9
7 changed files with 107 additions and 9 deletions
@@ -217,6 +217,7 @@ suites = {
        TestFile("test_cpp_radix_cache.py", 60),
    ],
    "nightly-4-gpu-b200": [
+        TestFile("nightly/test_flashinfer_trtllm_gen_moe_backend.py", 300),
        TestFile("test_fp4_moe.py", 300),
        TestFile("nightly/test_gpt_oss_4gpu_perf.py", 600),
        TestFile("nightly/test_flashinfer_trtllm_gen_attn_backend.py", 300),