【Hackathon 9th No.28】add test_cutlass_fp8_fp8_fp8_dual_gemm_fused (#3935)

* add test_cutlass_fp8_fp8_fp8_dual_gemm_fused

* fix the version

* fix code style

---------

Co-authored-by: Tao Luo <[email protected]>

Author: WanRui37

tests/operators/test_cutlass_fp8_fp8_fp8_dual_gemm_fused.py

@@ -0,0 +1,120 @@
+# Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import unittest
+from itertools import product
+
+import numpy as np
+import paddle
+
+from fastdeploy.model_executor.ops.gpu import cutlass_fp8_fp8_fp8_dual_gemm_fused
+
+
+class TestFp8Fp8Fp8DualGemm(unittest.TestCase):
+    def setUp(self):
+        """
+        Initialize the test environment,
+        including setting random seeds and environment variables.
+        """
+        paddle.seed(2024)
+        self.prop = paddle.device.cuda.get_device_properties()
+        self.sm_version = self.prop.major * 10 + self.prop.minor
+        print(f"sm version: {self.sm_version}")
+        self.E4M3_MAX_POS = 448.0
+        os.environ["FLAGS_cuda_core_fp8_gemm"] = "1"
+        print(paddle.device.cuda.get_device_properties())
+        print(paddle.__git_commit__)
+
+    def test_dual_gemm_case(self):
+        """
+        Check if the cutlass_fp8_fp8_fp8_dual_gemm_fused function works properly.
+        """
+        if self.sm_version < 90:
+            self.skipTest("cutlass_fp8_fp8_fp8_dual_gemm_fused only support sm90+")
+        nks = [
+            [2048, 2048],
+            [2048, 5504],
+            [6144, 2048],
+            [4096, 4096],
+            [4096, 12800],
+            [6144, 4096],
+            [5120, 5120],
+            [5120, 13824],
+            [15360, 5120],
+        ]
+        m_values = [1, 2, 3, 4]
+        transpose_combinations = [(False, True)]
+        activation_types = [""]
+
+        combinations = product(m_values, nks, transpose_combinations, activation_types)
+        for m, (n, k), (trans_x, trans_y), act_type in combinations:
+            x = (
+                paddle.rand([m, k] if not trans_x else [k, m])
+                .clip(min=-self.E4M3_MAX_POS, max=self.E4M3_MAX_POS)
+                .to(paddle.float8_e4m3fn)
+            )
+
+            y0 = (
+                paddle.rand([k, n] if not trans_y else [n, k])
+                .clip(min=-self.E4M3_MAX_POS, max=self.E4M3_MAX_POS)
+                .to(paddle.float8_e4m3fn)
+            )
+
+            y1 = (
+                paddle.rand([k, n] if not trans_y else [n, k])
+                .clip(min=-self.E4M3_MAX_POS, max=self.E4M3_MAX_POS)
+                .to(paddle.float8_e4m3fn)
+            )
+
+            scale0 = 1.2
+            scale1 = 0.8
+            scale_out = 1.0
+
+            x_bf16 = x.astype("bfloat16")
+            y0_bf16 = y0.astype("bfloat16")
+            y1_bf16 = y1.astype("bfloat16")
+
+            gemm0 = paddle.matmul(x_bf16, y0_bf16, transpose_x=trans_x, transpose_y=trans_y)
+            gemm1 = paddle.matmul(x_bf16, y1_bf16, transpose_x=trans_x, transpose_y=trans_y)
+
+            gemm0 = gemm0 * scale0
+            gemm1 = gemm1 * scale1
+
+            if act_type == "" or act_type == "swiglu":
+                ref_out = gemm0 * paddle.nn.functional.sigmoid(gemm1)
+
+            ref_out = ref_out.clip(min=-self.E4M3_MAX_POS, max=self.E4M3_MAX_POS).to(paddle.float8_e4m3fn)
+
+            result = cutlass_fp8_fp8_fp8_dual_gemm_fused(
+                x,
+                y0,
+                y1,
+                bias0=None,
+                bias1=None,
+                transpose_x=trans_x,
+                transpose_y=trans_y,
+                scale0=scale0,
+                scale1=scale1,
+                scale_out=scale_out,
+                activation_type=act_type,
+            )
+
+            np.testing.assert_allclose(
+                ref_out.astype("float32").numpy(), result.astype("float32").numpy(), rtol=5e-3, atol=5e-3
+            )
+
+
+if __name__ == "__main__":
+    unittest.main()