[converter&docs] fix misc & converter tests (#283)

* fix misc tests

* debug

* fix constant list unpacking

docs/op_matrix.md

@@ -94,6 +94,8 @@ Operators that are implemented in Python
 | `aten::hardtanh_` |  |
 | `aten::im2col` | only 4-D input tensors (batched image-like tensors) are supported |
 | `aten::index` | Multiple indices for aten::index is not supported |
+| `aten::index_put` | aten::index_put_ with accumulate=True is not supported |
+| `aten::index_put_` | aten::index_put_ with accumulate=True is not supported |
 | `aten::index_select` |  |
 | `aten::instance_norm` |  |
 | `aten::layer_norm` |  |

tinynn/converter/operators/torch/aten.py

@@ -3077,7 +3077,10 @@ def parse(self, node, attrs, args, graph_converter):
                 input_names, self.input_tensors[1], graph_converter=graph_converter, non_existent_as_buffer=True
             )
         else:
-            indices_tensors = [self.find_or_create_input(1, graph_converter)]
+            if type(self.input_tensors[1]) in (tuple, list):
+                indices_tensors = [self.create_attr_tensor(x) for x in self.input_tensors[1]]
+            else:
+                indices_tensors = [self.find_or_create_input(1, graph_converter)]
 
         dim = input_tensor.tensor.ndim