Add FX support for torch.baddbmm and torch.Tensor.baddbmm

src/transformers/utils/fx.py

@@ -305,12 +305,22 @@ def torch_matmul(input, other, *, out=None):
 
 def torch_bmm(input, mat2, *, out=None):
     if out is not None:
-        raise ValueError("Don't support in-place abs for MetaTensor analysis")
+        raise ValueError("Don't support in-place bmm for MetaTensor analysis")
     batch_size, n, m = input.shape
     _, _, p = mat2.shape
     return torch.empty(batch_size, n, p, device="meta")
 
 
+def torch_baddbmm(input, batch1, batch2, *, beta=1, alpha=1, out=None):
+    if out is not None:
+        raise ValueError("Don't support in-place baddbmm for MetaTensor analysis")
+    return torch_bmm(batch1, batch2)
+
+
+def torch_tensor_baddbmm(self, batch1, batch2, *, beta=1, alpha=1, out=None):
+    return torch_baddbmm(self, batch1, batch2, beta=beta, alpha=alpha, out=out)
+
+
 def torch_einsum(equation, *operands):
     # TODO: infer shape without performing the computation, this might be quite hard.
     concrete_operands = (torch.empty_like(operand, device="cpu") for operand in operands)
@@ -495,6 +505,8 @@ def to_concrete(t):
     torch.Tensor.mul: torch_tensor_mul,
     torch.matmul: torch_matmul,
     torch.bmm: torch_bmm,
+    torch.baddbmm: torch_baddbmm,
+    torch.Tensor.baddbmm: torch_tensor_baddbmm,
     torch.einsum: torch_einsum,
     torch.Tensor.repeat: torch_tensor_repeat,
     torch.roll: torch_roll,