lint and revert some change

Signed-off-by: jiahanc <[email protected]>

Author: jiahanc

csrc/trtllm_batched_gemm_runner.cu

@@ -169,6 +169,7 @@ void TrtllmGenBatchedGemmRunner::run(
   auto const configs = bmm.getBatchedGemmConfigs();
 
   auto const& config = configs[configIndex];
+
   FLASHINFER_CHECK(numBatches > 0, "Batched GEMM requires numBatches > 0");
   if (!mOptions.staticBatch) {
     FLASHINFER_CHECK(totalNumPaddedTokens,

include/flashinfer/trtllm/fused_moe/RoutingKernelTopK.cuh

@@ -217,7 +217,7 @@ __forceinline__ __device__ void reduceTopK(cg::thread_block_tile<WarpSize> const
 
     for (int ii = 0; ii < numResults; ++ii) {
       topKBufferValue[ii] = minValue;
-      topKBufferIdx[ii] = ii * WarpSize - 1;  //@todo: check if this is correct
+      topKBufferIdx[ii] = ii * WarpSize - 1;
     }
     for (int loop = 0; loop < numLoops; ++loop) {
       int start = loop * 4;

tests/moe/test_trtllm_gen_fused_moe.py

@@ -1914,7 +1914,7 @@ def cache_permute_indices():
                 "routed_scaling": None,
                 "has_routing_bias": False,
                 "routing_method_type": RoutingMethodType.Renormalize,
-                "compatible_moe_impls": [FP8BlockScaleMoe, FP4Moe],
+                "compatible_moe_impls": [FP8BlockScaleMoe, FP8PerTensorMoe, FP4Moe],
             },
             id="Renorm",
             marks=pytest.mark.skip(