diff --git a/vllm/kernels/helion/configs/rms_norm_per_block_quant/nvidia_b200.json b/vllm/kernels/helion/configs/rms_norm_per_block_quant/nvidia_b200.json index 6acf8f29da..afb30aab42 100644 --- a/vllm/kernels/helion/configs/rms_norm_per_block_quant/nvidia_b200.json +++ b/vllm/kernels/helion/configs/rms_norm_per_block_quant/nvidia_b200.json @@ -655,7 +655,7 @@ "range_warp_specializes": [ null, null, - true, + false, null ], "range_num_stages": [], @@ -1705,7 +1705,7 @@ "range_warp_specializes": [ null, false, - true, + false, null ], "range_num_stages": [], @@ -1838,14 +1838,14 @@ ], "range_unroll_factors": [ 0, - 4, + 2, 0, 1 ], "range_warp_specializes": [ null, null, - true, + false, null ], "range_num_stages": [], @@ -1872,7 +1872,7 @@ "last", "" ], - "num_warps": 2, + "num_warps": 4, "num_stages": 6, "indexing": [ "tensor_descriptor", @@ -1881,7 +1881,7 @@ "tensor_descriptor", "tensor_descriptor", "tensor_descriptor", - "pointer", + "tensor_descriptor", "tensor_descriptor", "pointer" ], @@ -2610,49 +2610,49 @@ ], "range_unroll_factors": [ 0, - 3, + 0, 0, 0 ], "range_warp_specializes": [ null, + false, null, - true, null ], "range_num_stages": [], "range_multi_buffers": [ null, - true, + null, null, null ], "range_flattens": [ null, + false, null, - true, null ], "static_ranges": [ true ], "load_eviction_policies": [ - "last", - "first", - "first", "first", "", + "", + "first", + "first", "last" ], - "num_warps": 1, - "num_stages": 3, + "num_warps": 4, + "num_stages": 7, "indexing": [ - "tensor_descriptor", - "tensor_descriptor", - "tensor_descriptor", - "tensor_descriptor", "pointer", "tensor_descriptor", + "pointer", + "pointer", + "tensor_descriptor", + "pointer", "tensor_descriptor", "tensor_descriptor", "tensor_descriptor" @@ -2680,14 +2680,14 @@ ], "range_unroll_factors": [ 0, - 4, + 0, 0, 0 ], "range_warp_specializes": [ - null, null, true, + null, null ], "range_num_stages": [], @@ -2707,14 +2707,14 @@ true ], "load_eviction_policies": [ - "", + "last", "", "first", "last", "last", "" ], - "num_warps": 1, + "num_warps": 2, "num_stages": 3, "indexing": [ "pointer", @@ -2723,7 +2723,7 @@ "tensor_descriptor", "tensor_descriptor", "pointer", - "tensor_descriptor", + "pointer", "pointer", "pointer" ], @@ -2809,7 +2809,7 @@ }, "config": { "block_sizes": [ - 8192, + 2048, 8 ], "loop_orders": [ @@ -2821,13 +2821,13 @@ "range_unroll_factors": [ 0, 3, - 0, - 2 + 1, + 0 ], "range_warp_specializes": [ null, null, - true, + false, null ], "range_num_stages": [], @@ -2841,10 +2841,10 @@ null, null, false, - false + null ], "static_ranges": [ - false + true ], "load_eviction_policies": [ "", @@ -2854,7 +2854,7 @@ "", "last" ], - "num_warps": 1, + "num_warps": 2, "num_stages": 3, "indexing": [ "pointer", @@ -2884,20 +2884,20 @@ ], "loop_orders": [ [ - 0, - 1 + 1, + 0 ] ], "range_unroll_factors": [ 0, - 3, + 1, 0, - 2 + 0 ], "range_warp_specializes": [ null, + false, null, - true, null ], "range_num_stages": [], @@ -2908,37 +2908,37 @@ null ], "range_flattens": [ - null, null, false, - false + false, + null ], "static_ranges": [ - false + true ], "load_eviction_policies": [ + "last", "", "", "first", - "last", - "", + "first", "last" ], - "num_warps": 1, - "num_stages": 3, + "num_warps": 2, + "num_stages": 7, "indexing": [ "pointer", "pointer", - "tensor_descriptor", + "pointer", + "pointer", "tensor_descriptor", "pointer", "tensor_descriptor", - "tensor_descriptor", "pointer", - "pointer" + "tensor_descriptor" ], "atomic_indexing": [], "pid_type": "flat" } } -] \ No newline at end of file +]