Copy of the inputs for the evaluation during compilation

forge/forge/tvm_to_python.py

@@ -2071,7 +2071,8 @@ def generate_forge_module(
             reload = False
 
     if verify_cfg is not None and verify_cfg.verify_forge_codegen_vs_framework:
-        framework_outputs = framework_mod.cpu_eval_forward(*pytorch_inputs)
+        eval_inputs = [eval_input.detach().clone() for eval_input in pytorch_inputs]
+        framework_outputs = framework_mod.cpu_eval_forward(*eval_inputs)
 
     if not reload:
         module_name = graph_name if counter == 0 else f"{graph_name}_{counter}"

forge/test/mlir/operators/tm/test_in_place.py

@@ -0,0 +1,116 @@
+# SPDX-FileCopyrightText: (c) 2025 Tenstorrent AI ULC
+#
+# SPDX-License-Identifier: Apache-2.0
+import pytest
+import torch
+import tensorflow as tf
+
+import forge
+from forge.verify.compare import compare_with_golden
+
+
+@pytest.mark.parametrize(
+    "shape",
+    [
+        (3, 3),
+    ],
+)
+@pytest.mark.push
+def test_in_place_torch(shape):
+    class Inplace(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+
+        def forward(self, x):
+            y = x + 1
+            x += 2
+
+            return x + y
+
+    input = torch.zeros(shape, requires_grad=False)
+    framework_input = input.detach().clone()
+    tt_inputs = [input]
+
+    framework_model = Inplace()
+    y = framework_model(framework_input)
+
+    compiled_model = forge.compile(framework_model, sample_inputs=tt_inputs, module_name="inplace")
+    tty = compiled_model(*tt_inputs)[0]
+
+    compare_with_golden(golden=y, calculated=tty)
+    print(framework_input)
+    print(tt_inputs[0])
+
+
+@pytest.mark.parametrize(
+    "shape",
+    [
+        (3, 3),
+    ],
+)
+@pytest.mark.push
+def test_in_place_tf(shape):
+    class Inplace(tf.keras.Model):
+        def __init__(self):
+            super().__init__()
+
+        def call(self, x):
+            y = x + 1
+            x += 2
+            return x + y
+
+    input = tf.zeros(shape)
+    framework_input = tf.identity(input)
+    tt_inputs = [input]
+
+    framework_model = Inplace()
+    y = framework_model(framework_input)
+
+    compiled_model = forge.compile(framework_model, sample_inputs=tt_inputs, module_name="inplace")
+    tty = compiled_model(*tt_inputs)[0]
+
+    # convert tensor from tf to torch
+    y = torch.tensor(y.numpy())
+
+    compare_with_golden(golden=y, calculated=tty)
+
+    print(framework_input)
+    print(tt_inputs[0])
+
+
+@pytest.mark.parametrize(
+    "shape",
+    [
+        (1, 1024),
+    ],
+)
+@pytest.mark.push
+@pytest.mark.xfail(
+    reason="TT model silently computes the result for the tensor with requires_grad=True on which in-place operation has been performed"
+)
+def test_in_place_backward(shape):
+    class MatmulParam(torch.nn.Module):
+        def __init__(self):
+            super().__init__()
+            self.p = torch.nn.Parameter(torch.rand(1024, 1024))
+            torch.nn.init.xavier_uniform_(self.p)
+
+        def forward(self, x):
+            y = torch.matmul(x, self.p)
+            x += 2  # In-place modification causing runtime error
+            return y
+
+    inputs = torch.rand(shape, requires_grad=True)
+
+    model = MatmulParam()
+    tt_model = forge.compile(model, sample_inputs=[torch.rand(shape)])
+
+    # Expect PyTorch model to raise a RuntimeError
+    with pytest.raises(
+        RuntimeError, match="a leaf Variable that requires grad is being used in an in-place operation."
+    ):
+        model(inputs)
+
+    # Expect tt_model to also raise a RuntimeError
+    with pytest.raises(RuntimeError):
+        tt_model(inputs)  # If no error is raised, the test will fail

forge/test/models/pytorch/vision/mlp_mixer/test_mlp_mixer.py

@@ -49,7 +49,7 @@ def test_mlp_mixer_timm_pytorch(record_forge_property, variant):
     record_forge_property("model_name", module_name)
 
     framework_model = download_model(timm.create_model, variant, pretrained=True)
-    config = resolve_data_config({}, model=model)
+    config = resolve_data_config({}, model=framework_model)
     transform = create_transform(**config)
 
     try: