add tapas masklm models

Author: co63oc

samples/transformers-auto-model/google/tapas-base-masklm/graph_hash.txt

@@ -0,0 +1 @@
+8e2f2f1b2cd29f1d0f726b2818b4ae3aac430fe15412a65835665131fc87436e

samples/transformers-auto-model/google/tapas-base-masklm/graph_net.json

@@ -0,0 +1,6 @@
+{
+    "framework": "torch",
+    "num_devices_required": 1,
+    "num_nodes_required": 1,
+    "dynamic": true
+}

samples/transformers-auto-model/google/tapas-base-masklm/input_meta.py

@@ -0,0 +1,39 @@
+import torch
+
+from torch import device
+
+
+class GraphModule(torch.nn.Module):
+    def forward(
+        self,
+        s0: torch.SymInt,
+        L_index_indices: torch.Tensor,
+        L_index_num_segments: torch.Tensor,
+    ):
+        l_index_indices = L_index_indices
+        l_index_num_segments = L_index_num_segments
+        tensor = torch.tensor([1])
+        batch_size = torch.prod(tensor)
+        tensor = None
+        arange = torch.arange(start=0, end=batch_size, device=device(type="cpu"))
+        offset = arange * l_index_num_segments
+        arange = None
+        offset_1 = offset.view((1,))
+        offset = None
+        offset_2 = offset_1.unsqueeze(-1)
+        offset_1 = None
+        indices = offset_2 + l_index_indices
+        offset_2 = l_index_indices = None
+        view_1 = indices.view(-1)
+        indices = None
+        mul_1 = l_index_num_segments * batch_size
+        l_index_num_segments = batch_size = None
+        as_tensor = torch.as_tensor(view_1, device=device(type="cpu"))
+        view_1 = None
+        as_tensor_1 = torch.as_tensor(mul_1, device=device(type="cpu"))
+        mul_1 = None
+        as_tensor_2 = torch.as_tensor([-1], dtype=torch.int64)
+        as_tensor_3 = torch.as_tensor((), dtype=torch.int64)
+        flattened_shape = torch.cat([as_tensor_2, as_tensor_3], dim=0)
+        as_tensor_2 = as_tensor_3 = None
+        return (flattened_shape, as_tensor_1, as_tensor)