[New Sample] Add microsoft deberta large mnli

Author: GreatV

samples/transformers-auto-model/microsoft_deberta_large_mnli/graph_hash.txt

@@ -0,0 +1 @@
+5f2c70e2cf8e9a078d107f0461e6546e1f3f0ef020361c727bcd6bbadd3da96f

samples/transformers-auto-model/microsoft_deberta_large_mnli/graph_net.json

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

samples/transformers-auto-model/microsoft_deberta_large_mnli/input_meta.py

@@ -0,0 +1,85 @@
+import torch
+
+
+class GraphModule(torch.nn.Module):
+    def forward(
+        self,
+        L_stack0_0_: torch.Tensor,
+        L_self_modules_intermediate_modules_dense_parameters_weight_: torch.nn.parameter.Parameter,
+        L_self_modules_intermediate_modules_dense_parameters_bias_: torch.nn.parameter.Parameter,
+        L_self_modules_output_modules_dense_parameters_weight_: torch.nn.parameter.Parameter,
+        L_self_modules_output_modules_dense_parameters_bias_: torch.nn.parameter.Parameter,
+        L_self_modules_output_modules_LayerNorm_parameters_weight_: torch.nn.parameter.Parameter,
+        L_self_modules_output_modules_LayerNorm_parameters_bias_: torch.nn.parameter.Parameter,
+    ):
+        l_stack0_0_ = L_stack0_0_
+        l_self_modules_intermediate_modules_dense_parameters_weight_ = (
+            L_self_modules_intermediate_modules_dense_parameters_weight_
+        )
+        l_self_modules_intermediate_modules_dense_parameters_bias_ = (
+            L_self_modules_intermediate_modules_dense_parameters_bias_
+        )
+        l_self_modules_output_modules_dense_parameters_weight_ = (
+            L_self_modules_output_modules_dense_parameters_weight_
+        )
+        l_self_modules_output_modules_dense_parameters_bias_ = (
+            L_self_modules_output_modules_dense_parameters_bias_
+        )
+        l_self_modules_output_modules_layer_norm_parameters_weight_ = (
+            L_self_modules_output_modules_LayerNorm_parameters_weight_
+        )
+        l_self_modules_output_modules_layer_norm_parameters_bias_ = (
+            L_self_modules_output_modules_LayerNorm_parameters_bias_
+        )
+        hidden_states = torch._C._nn.linear(
+            l_stack0_0_,
+            l_self_modules_intermediate_modules_dense_parameters_weight_,
+            l_self_modules_intermediate_modules_dense_parameters_bias_,
+        )
+        l_self_modules_intermediate_modules_dense_parameters_weight_ = (
+            l_self_modules_intermediate_modules_dense_parameters_bias_
+        ) = None
+        hidden_states_1 = torch._C._nn.gelu(hidden_states)
+        hidden_states = None
+        hidden_states_2 = torch._C._nn.linear(
+            hidden_states_1,
+            l_self_modules_output_modules_dense_parameters_weight_,
+            l_self_modules_output_modules_dense_parameters_bias_,
+        )
+        hidden_states_1 = (
+            l_self_modules_output_modules_dense_parameters_weight_
+        ) = l_self_modules_output_modules_dense_parameters_bias_ = None
+        hidden_states_3 = torch.nn.functional.dropout(
+            hidden_states_2, 0.1, False, False
+        )
+        hidden_states_2 = None
+        add = hidden_states_3 + l_stack0_0_
+        hidden_states_3 = l_stack0_0_ = None
+        hidden_states_4 = add.float()
+        add = None
+        mean = hidden_states_4.mean(-1, keepdim=True)
+        sub = hidden_states_4 - mean
+        pow_1 = sub.pow(2)
+        sub = None
+        variance = pow_1.mean(-1, keepdim=True)
+        pow_1 = None
+        sub_1 = hidden_states_4 - mean
+        hidden_states_4 = mean = None
+        add_1 = variance + 1e-07
+        variance = None
+        sqrt = torch.sqrt(add_1)
+        add_1 = None
+        hidden_states_5 = sub_1 / sqrt
+        sub_1 = sqrt = None
+        hidden_states_6 = hidden_states_5.to(torch.float32)
+        hidden_states_5 = None
+        mul = (
+            l_self_modules_output_modules_layer_norm_parameters_weight_
+            * hidden_states_6
+        )
+        l_self_modules_output_modules_layer_norm_parameters_weight_ = (
+            hidden_states_6
+        ) = None
+        y = mul + l_self_modules_output_modules_layer_norm_parameters_bias_
+        mul = l_self_modules_output_modules_layer_norm_parameters_bias_ = None
+        return (y,)

samples/transformers-auto-model/microsoft_deberta_large_mnli/input_tensor_constraints.py

@@ -0,0 +1,68 @@
+class Program_weight_tensor_meta_L_stack0_0_:
+    name = "L_stack0_0_"
+    shape = [1, 10, 1024]
+    dtype = "torch.float32"
+    device = "cuda:0"
+    mean = -0.006
+    std = 1.098
+    data = None
+
+
+class Program_weight_tensor_meta_L_self_modules_intermediate_modules_dense_parameters_weight_:
+    name = "L_self_modules_intermediate_modules_dense_parameters_weight_"
+    shape = [4096, 1024]
+    dtype = "torch.float32"
+    device = "cuda:0"
+    mean = -0.000
+    std = 0.026
+    data = None
+
+
+class Program_weight_tensor_meta_L_self_modules_intermediate_modules_dense_parameters_bias_:
+    name = "L_self_modules_intermediate_modules_dense_parameters_bias_"
+    shape = [4096]
+    dtype = "torch.float32"
+    device = "cuda:0"
+    mean = -0.033
+    std = 0.022
+    data = None
+
+
+class Program_weight_tensor_meta_L_self_modules_output_modules_dense_parameters_weight_:
+    name = "L_self_modules_output_modules_dense_parameters_weight_"
+    shape = [1024, 4096]
+    dtype = "torch.float32"
+    device = "cuda:0"
+    mean = 0.000
+    std = 0.026
+    data = None
+
+
+class Program_weight_tensor_meta_L_self_modules_output_modules_dense_parameters_bias_:
+    name = "L_self_modules_output_modules_dense_parameters_bias_"
+    shape = [1024]
+    dtype = "torch.float32"
+    device = "cuda:0"
+    mean = 0.000
+    std = 0.014
+    data = None
+
+
+class Program_weight_tensor_meta_L_self_modules_output_modules_LayerNorm_parameters_weight_:
+    name = "L_self_modules_output_modules_LayerNorm_parameters_weight_"
+    shape = [1024]
+    dtype = "torch.float32"
+    device = "cuda:0"
+    mean = 1.131
+    std = 0.063
+    data = None
+
+
+class Program_weight_tensor_meta_L_self_modules_output_modules_LayerNorm_parameters_bias_:
+    name = "L_self_modules_output_modules_LayerNorm_parameters_bias_"
+    shape = [1024]
+    dtype = "torch.float32"
+    device = "cuda:0"
+    mean = -0.001
+    std = 0.032
+    data = None