Use IDs for placeholder/outputs of implicit q/dq nodes (#141)

Summary:
Pull Request resolved: #141

In XNNPACK from the partitioner we encounter quantize node that are outputs to the graph. Since these quantize nodes are implicit and not actually quantizing the values, we must give its externalness to the node with which it quantizes.

### Outputs
```
                 |
conv --> q ----> | ----> dq -->
                 |
         Delegate boundary
```
In this example the output of the delegate is the output of quantized Conv. In XNNPACK, we must designate the output of Conv to be an external value. In this graph, however, the q node is the output. Since q is an implicit node (it does not actually running quantize) We pass its externalness to the conv node.

### Inputs
i.e.
```
        |
q ----> | ----> dq --> conv -->
        |
Delegate boundary
```
In this example q is the placeholder node and is given as input to the first dq. But we are running a quantized conv within the delegate. Since dq is implicit, we pass the externalness of the q place holder to dq, so that the input to Conv is external.

Reviewed By: digantdesai

Differential Revision: D48667676

fbshipit-source-id: 0be5bf8c3da227463e12eea46d72b76561a34e40

Author: mcr229

backends/xnnpack/operators/node_visitor.py

@@ -130,21 +130,33 @@ def gen_ids_and_flags(
         # This will break if we change the way q/dq are partitioned
 
         # Tensor can still be input if its quantizing node is an input
-        is_input = (self).is_graph_input(tensor)
+        if self.is_graph_input(tensor) or (
+            quant_params.is_input if quant_params else False
+        ):
+            tensor_input = tensor
+            if quant_params:
+                if quant_params.is_input and not self.is_graph_input(tensor):
+                    tensor_input = quant_params.q_input
+            assert (
+                tensor_input in self.external_ids.keys()
+            ), f"Tensor {tensor_input}, is_input. ext_ids: {self.external_ids.keys()}"
+            ext_id = self.external_ids[tensor_input].external_id
+            xnn_graph.input_ids.append(id_out)
+            flag = self.external_ids[tensor_input].io_type
         # Tensor can still be output if its quantizing node is an output
-        is_output = self.is_graph_output(tensor)
-        # handle logic for input/output tensors
-        if is_input or is_output:
+        elif self.is_graph_output(tensor) or (
+            quant_params.is_output if quant_params else False
+        ):
+            tensor_output = tensor
+            if quant_params:
+                if quant_params.is_output and not self.is_graph_output(tensor):
+                    tensor_output = list(tensor.users)[0]
             assert (
-                tensor in self.external_ids.keys()
-            ), f"Tensor {tensor}, is_input: {is_input}, is_output: {is_output}, ext_ids: {self.external_ids.keys()}"
-            ext_id = self.external_ids[tensor].external_id
-            if is_input:
-                xnn_graph.input_ids.append(id_out)
-                flag = XNN_VALUE_FLAG_EXTERNAL_INPUT
-            if is_output:
-                xnn_graph.output_ids.append(id_out)
-                flag = XNN_VALUE_FLAG_EXTERNAL_OUTPUT
+                tensor_output in self.external_ids.keys()
+            ), f"Tensor {tensor_output} is_output: ext_ids: {self.external_ids.keys()}"
+            ext_id = self.external_ids[tensor_output].external_id
+            xnn_graph.output_ids.append(id_out)
+            flag = self.external_ids[tensor_output].io_type
 
         return ext_id, id_out, flag