[Bugfix] Fix Marlin MoE act order when is_k_full == False (#8741)

Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>

csrc/core/exception.hpp

+#pragma once
+
+#define VLLM_IMPLIES(p, q) (!(p) || (q))

csrc/moe/marlin_moe_ops.cu

@@ -25,6 +25,7 @@
 
 #include <iostream>
 
+#include "core/exception.hpp"
 #include "core/scalar_type.hpp"
 #include "marlin_kernels/marlin_moe_kernel_ku4b8.h"
 #include "marlin_kernels/marlin_moe_kernel_ku8b128.h"
@@ -189,7 +190,7 @@ int get_scales_cache_size(thread_config_t const& th_config, int prob_m,
     int load_groups =
         tb_groups * STAGES * 2;          // Chunk size is 2x pipeline over dim K
     load_groups = max(load_groups, 32);  // We load at least 32 scale groups
-    return load_groups * tb_n * 2;
+    return load_groups * tb_n * 4;
 
   } else {
     int tb_scales = tb_groups * tb_n * 2;
@@ -433,11 +434,7 @@ void marlin_mm_moe(const void* A, const void* B, void* C,
     int4* C_ptr = (int4*)C;
     const float* topk_weights_ptr = (const float*)topk_weights;
     const int* sorted_ids_ptr = (const int*)sorted_ids;
-    const int4* s_ptr =
-        (const int4*)s +
-        (((group_size == -1 || group_size == 0) ? 1 : prob_k / group_size) *
-         prob_n / 8) *
-            expert_idx;
+    const int4* s_ptr = (const int4*)s + num_groups * prob_n / 8 * expert_idx;
     const int* g_idx_ptr = (const int*)g_idx + prob_k * expert_idx;
     const int* perm_ptr = (const int*)perm + prob_k * expert_idx;
     int* locks = (int*)workspace;
@@ -521,6 +518,9 @@ torch::Tensor marlin_gemm_moe(
               " is not size_n = ", size_n);
   num_groups = b_scales.size(1);
 
+  TORCH_CHECK(VLLM_IMPLIES(!is_k_full, has_act_order),
+              "if is_k_full is false, has_act_order must be true");
+
   if (has_act_order) {
     if (is_k_full) {
       TORCH_CHECK(num_groups > 1, "For act_order, num_groups must be > 1");

tests/kernels/test_moe.py

@@ -145,6 +145,7 @@ def compute_max_diff(output, output_ref):
 @pytest.mark.parametrize("group_size", [-1, 32, 64, 128])
 @pytest.mark.parametrize("act_order", [True, False])
 @pytest.mark.parametrize("num_bits", [4, 8])
+@pytest.mark.parametrize("is_k_full", [True, False])
 def test_fused_marlin_moe(
     m: int,
     n: int,
@@ -154,6 +155,7 @@ def test_fused_marlin_moe(
     group_size: int,
     act_order: bool,
     num_bits: int,
+    is_k_full: bool,
 ):
     seed_everything(7)
 
@@ -166,6 +168,9 @@ def test_fused_marlin_moe(
             return
         if group_size in (k, n):
             return
+    else:
+        if not is_k_full:
+            return
 
     quant_type = (scalar_types.uint4b8
                   if num_bits == 4 else scalar_types.uint8b128)
@@ -246,6 +251,7 @@ def test_fused_marlin_moe(
         w1_scale=scales1,
         w2_scale=scales2,
         num_bits=num_bits,
+        is_k_full=is_k_full,
     )
 
     assert compute_max_diff(marlin_output, triton_output) < 4e-2
@@ -290,6 +296,7 @@ def test_fused_marlin_moe(
 @pytest.mark.parametrize("group_size", [-1, 32, 64, 128])
 @pytest.mark.parametrize("act_order", [True, False])
 @pytest.mark.parametrize("num_bits", [4, 8])
+@pytest.mark.parametrize("is_k_full", [True, False])
 def test_single_marlin_moe_multiply(
     m: int,
     n: int,
@@ -299,6 +306,7 @@ def test_single_marlin_moe_multiply(
     group_size: int,
     act_order: bool,
     num_bits: int,
+    is_k_full: bool,
 ):
     if topk > e:
         return
@@ -309,6 +317,9 @@ def test_single_marlin_moe_multiply(
             return
         if group_size == k:
             return
+    else:
+        if not is_k_full:
+            return
 
     quant_type = (scalar_types.uint4b8
                   if num_bits == 4 else scalar_types.uint8b128)
@@ -339,15 +350,18 @@ def test_single_marlin_moe_multiply(
     sort_indices = stack_and_dev(sort_indices_l)
 
     score = torch.randn((m, e), device="cuda", dtype=dtype)
-    marlin_output = single_marlin_moe(a,
-                                      qweight,
-                                      scales,
-                                      score,
-                                      g_idx,
-                                      sort_indices,
-                                      topk,
-                                      renormalize=False,
-                                      num_bits=num_bits)
+    marlin_output = single_marlin_moe(
+        a,
+        qweight,
+        scales,
+        score,
+        g_idx,
+        sort_indices,
+        topk,
+        renormalize=False,
+        num_bits=num_bits,
+        is_k_full=is_k_full,
+    )
     torch_output = torch_moe_single(a, w_ref.transpose(1, 2), score, topk)
 
     assert compute_max_diff(marlin_output, torch_output) < 1e-2

vllm/model_executor/layers/fused_moe/fused_marlin_moe.py

@@ -21,6 +21,7 @@ def single_marlin_moe(
     renormalize: bool,
     override_config: Optional[Dict[str, Any]] = None,
     num_bits: int = 8,
+    is_k_full: bool = True,
 ) -> torch.Tensor:
     """
     This function computes the multiplication of hidden_states with expert
@@ -86,7 +87,7 @@ def single_marlin_moe(
 
     intermediate_cache = torch.ops._moe_C.marlin_gemm_moe(
         hidden_states, w, sorted_token_ids, topk_weights, topk_ids, scales,
-        g_idx, perm, workspace, scalar_type, M, N, K, True, E, topk,
+        g_idx, perm, workspace, scalar_type, M, N, K, is_k_full, E, topk,
         block_size_m, True, False)
 
     return torch.sum(intermediate_cache.view(*intermediate_cache.shape), dim=1)
@@ -107,6 +108,7 @@ def fused_marlin_moe(
     w1_scale: Optional[torch.Tensor] = None,
     w2_scale: Optional[torch.Tensor] = None,
     num_bits: int = 8,
+    is_k_full: bool = True,
 ) -> torch.Tensor:
     """
     This function computes a Mixture of Experts (MoE) layer using two sets of
@@ -199,7 +201,7 @@ def fused_marlin_moe(
         M,
         2 * N,
         K,
-        True,
+        is_k_full,
         E,
         topk,
         block_size_m,
@@ -223,7 +225,7 @@ def fused_marlin_moe(
         M,
         K,
         N,
-        True,
+        is_k_full,
         E,
         topk,
         block_size_m,