From ef5c8ab97996fa6a43fe8020feda52d43be185c0 Mon Sep 17 00:00:00 2001 From: Brian Nguyen Date: Fri, 14 Aug 2026 09:58:41 -0500 Subject: [PATCH] [TRTLLM-15177][test] Wire remaining Kimi K3 MoE unit tests into L0 (Hopper) test_kimi_k3_mlp.py, test_kimi_k3_moe_gate.py, and test_kimi_k3_situ_and_mul.py existed in the tree but were referenced by no L0 test list, so they never ran in pre-merge CI. They are any-CUDA (Triton trtllm::situ_and_mul op plus a torch reference, gated on requires_cuda only, not SM100/SM103), so wire them into the H100 pre-merge single-GPU list next to the other MoE-component unit tests rather than spend scarce Blackwell capacity. The Blackwell-gated K3 suites (kimi_kda, kimi_k3_attn_res, kimi_k3_situ_moe, KDA parity) are already wired in l0_b200.yml and l0_gb300_multi_gpus.yml. Closes item 1.4 of TRTLLM-15177 for the previously-unwired K3 MoE files. Signed-off-by: Brian Nguyen --- tests/integration/test_lists/test-db/l0_h100.yml | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/tests/integration/test_lists/test-db/l0_h100.yml b/tests/integration/test_lists/test-db/l0_h100.yml index 7b26114e6126..f4a962de8456 100644 --- a/tests/integration/test_lists/test-db/l0_h100.yml +++ b/tests/integration/test_lists/test-db/l0_h100.yml @@ -41,6 +41,14 @@ l0_h100: # ------------- MoE: test_moe_backend (by backend) --------------- # ------------- MoE: test_single_gpu (by backend) --------------- - unittest/_torch/modules/moe/test_moe_module.py::test_configurable_moe_single_gpu -k "CUTLASS and not None" + # ------------- Kimi K3 MoE unit tests --------------- + # Any-CUDA (Triton trtllm::situ_and_mul op + torch reference, requires_cuda + # only; not SM100/SM103-gated), so they run on Hopper rather than consuming + # scarce Blackwell pre-merge capacity. The Blackwell-gated K3 suites live in + # l0_b200.yml / l0_gb300_multi_gpus.yml. (TRTLLM-15177) + - unittest/_torch/modules/moe/test_kimi_k3_mlp.py + - unittest/_torch/modules/moe/test_kimi_k3_moe_gate.py + - unittest/_torch/modules/moe/test_kimi_k3_situ_and_mul.py - unittest/_torch/multimodal -k "not nemotron_nano_v2_vl_fp8" # test_speculative_d2h_parity_real_predictor (both no_stop_token and # stop_token_15) is a HW-agnostic sampler D2H-parity invariant (plain CUDA