[Relax][Frontend][ONNX] Fix FastGelu when bias does not set (#18358)

* [#17877][Relax][Frontend][ONNX] Fix  when bias does not set

* [#17877][FRONTEND][ONNX] Fix Error converting operator FastGelu, with inputs: [x, bias]

* [#17877][FRONTEND][ONNX] Fix Warning: Detected pow(x, y) where y >= 3, it is recommended to avoid

* [#17877][FRONTEND][ONNX] Fix tvm.error.InternalError: Check failed: (ptr) is false: The struct_info is not populated, check if you have normalized the expr

Author: cchung100m

python/tvm/relax/frontend/onnx/onnx_frontend.py

@@ -1155,11 +1155,12 @@ class FastGelu(OnnxOpConverter):
 
     @classmethod
     def _impl_v1(cls, bb, inputs, attr, params):
-        if inputs[1]:
+        x = inputs[0]
+        if len(inputs) > 1 and inputs[1] is not None:
             bias = inputs[1]
             bias_shape = bias.struct_info.shape
             assert len(bias_shape) == 1, "bias term must be a 1D tensor"
-            x += bias
+            x = bb.emit(relax.op.add(x, bias))
 
         # Declare consts
         const_dtype = x.struct_info.dtype
@@ -1169,11 +1170,13 @@ def _impl_v1(cls, bb, inputs, attr, params):
         const2 = relax.const(0.044715 * math.sqrt(2 / math.pi), dtype=const_dtype)
 
         # Compute FastGelu
-        term1 = relax.op.multiply(half, x)
-        term2 = relax.op.multiply(const1, x)
-        term3 = relax.op.multiply(const2, relax.op.power(x, relax.const(3, const_dtype)))
-        tanh = relax.op.tanh(relax.op.add(term2, term3))
-        return relax.op.multiply(term1, relax.op.add(one, tanh))
+        term1 = bb.emit(relax.op.multiply(half, x))
+        term2 = bb.emit(relax.op.multiply(const1, x))
+        # use x^3 = x * x * x instead of pow(x, 3) for better performance
+        x_cubed = bb.emit(relax.op.multiply(relax.op.multiply(x, x), x))
+        term3 = bb.emit(relax.op.multiply(const2, x_cubed))
+        tanh = bb.emit(relax.op.tanh(relax.op.add(term2, term3)))
+        return bb.emit(relax.op.multiply(term1, relax.op.add(one, tanh)))
 
 
 class BiasGelu(OnnxOpConverter):

tests/python/relax/test_frontend_onnx.py

@@ -828,6 +828,36 @@ def test_bias_gelu():
     verify_binary("BiasGelu", [32, 32], [32], [32, 32], domain="com.microsoft")
 
 
+def test_fast_gelu():
+    """Test FastGelu with and without bias"""
+    # Test FastGelu without bias
+    fast_gelu_node = helper.make_node("FastGelu", ["x"], ["y"], domain="com.microsoft")
+    graph = helper.make_graph(
+        [fast_gelu_node],
+        "fast_gelu_test",
+        inputs=[helper.make_tensor_value_info("x", TensorProto.FLOAT, [32, 32])],
+        outputs=[helper.make_tensor_value_info("y", TensorProto.FLOAT, [32, 32])],
+    )
+    model = helper.make_model(graph, producer_name="fast_gelu_test")
+    check_correctness(model)
+
+    # Test FastGelu with bias
+    fast_gelu_with_bias_node = helper.make_node(
+        "FastGelu", ["x", "bias"], ["y"], domain="com.microsoft"
+    )
+    graph_with_bias = helper.make_graph(
+        [fast_gelu_with_bias_node],
+        "fast_gelu_with_bias_test",
+        inputs=[
+            helper.make_tensor_value_info("x", TensorProto.FLOAT, [32, 32]),
+            helper.make_tensor_value_info("bias", TensorProto.FLOAT, [32]),
+        ],
+        outputs=[helper.make_tensor_value_info("y", TensorProto.FLOAT, [32, 32])],
+    )
+    model_with_bias = helper.make_model(graph_with_bias, producer_name="fast_gelu_with_bias_test")
+    check_correctness(model_with_bias)
+
+
 def test_where():
     where_node = helper.make_node("Where", ["a", "b", "c"], ["d"])