get_common_data, add name parameter

Before, the output name was just the first input's name.
This was very confusing when get_common_data errored because of implicit broadcasting:
Then you would think the error was in the first input,
while it was in the layer that combined the dims instead.

returnn/tf/layers/base.py

@@ -367,7 +367,7 @@ def _base_get_out_data_from_opts(cls, network, name,
       allow_broadcast_all_sources = True
     sources_data = Data.get_common_data(
       sources_data_list, ignore_feature_dim=True,
-      allow_broadcast_all_sources=allow_broadcast_all_sources) if sources_data_list else None
+      allow_broadcast_all_sources=allow_broadcast_all_sources, name="%s_sources" % name) if sources_data_list else None
     if sources_data and not sources_data.sparse and not out_type.get("sparse", False):
       out_type.setdefault("dtype", sources_data.dtype)
     # You are supposed to set self.output.{batch_dim_axis,time_dim_axis} explicitly,

returnn/tf/layers/basic.py

@@ -1051,7 +1051,7 @@ def __init__(self, start, size, min_size=None, out_spatial_dim=None, **kwargs):
     data_objs = [start_data]
     data_objs += [size.output] if isinstance(size, LayerBase) else []
     data_objs += [seq_lens_data] if isinstance(seq_lens_data, Data) else []
-    common_data = Data.get_common_data(data_objs)
+    common_data = Data.get_common_data(data_objs, name="%s_inputs")
     start_data = start_data.copy_compatible_to(common_data, check_sparse=False)
     start_t = start_data.placeholder
     if size is None:
@@ -3049,7 +3049,7 @@ def _set_output_sizes(self, merge_axes):
       if not out_size:
         out_size = in_size
       else:
-        new_data = Data.get_common_data([out_size, in_size])
+        new_data = Data.get_common_data([out_size, in_size], name="%s_output" % self.name)
         new_data.placeholder = (
           out_size.copy_compatible_to(new_data).placeholder
           * in_size.copy_compatible_to(new_data).placeholder)
@@ -5674,12 +5674,13 @@ def __init__(self, axis=None, out_spatial_dim=None, **kwargs):
     self.output.placeholder = tf.stack([src.placeholder for src in sources_], axis=axis)
 
   @classmethod
-  def _get_axis_and_common(cls, sources):
+  def _get_axis_and_common(cls, sources, name):
     """
     :param list[LayerBase] sources:
+    :param str name:
     :rtype: (int,Data)
     """
-    common_source = Data.get_common_data([src.output for src in sources]).copy_template()
+    common_source = Data.get_common_data([src.output for src in sources], name=name)
     dummy_tag = Dim(kind=Dim.Types.Spatial, dimension=1)
     return common_source.get_default_new_axis_for_dim_tag(dummy_tag), common_source
 
@@ -5692,7 +5693,7 @@ def get_out_data_from_opts(cls, name, sources, axis=None, out_spatial_dim=None,
     :param Dim|None out_spatial_dim:
     :rtype: Data
     """
-    axis_, common_source = cls._get_axis_and_common(sources)
+    axis_, common_source = cls._get_axis_and_common(sources, name="%s_sources" % name)
     if axis is None:
       axis = axis_
     out = common_source.copy_template(name="%s_output" % name)
@@ -6945,10 +6946,10 @@ def get_out_data_from_opts(cls, eval_locals=None, n_out=NotSpecified, sources=()
         allow_broadcast_all_sources = True
       out_type_.update(
         Data.get_common_data(
-          [s.output for s in sources], allow_broadcast_all_sources=allow_broadcast_all_sources).get_kwargs())
+          [s.output for s in sources], allow_broadcast_all_sources=allow_broadcast_all_sources).get_kwargs(),
+          name="%s_output" % kwargs["name"])
     if n_out is not NotSpecified:
       out_type_["dim"] = n_out
-    out_type_["name"] = "%s_output" % kwargs["name"]
     if out_type:
       if isinstance(out_type, dict):
         if "shape" in out_type:
@@ -7196,14 +7197,14 @@ def get_out_data_from_opts(cls, n_out=NotSpecified, out_type=None, out_shape=Non
         allow_broadcast_all_sources = True
       out_type_.update(
         Data.get_common_data(
-          [s.output for s in sources], allow_broadcast_all_sources=allow_broadcast_all_sources).get_kwargs())
+          [s.output for s in sources], allow_broadcast_all_sources=allow_broadcast_all_sources).get_kwargs(),
+          name="%s_output" % kwargs["name"])
     if n_out is not NotSpecified:
       out_type_["dim"] = n_out
     elif out_type_.get("sparse", False):
       out_type_["dim"] = 2
     out_type_["dtype"] = "bool"
     out_type_["vocab"] = None
-    out_type_["name"] = "%s_output" % kwargs["name"]
     if out_type:
       if isinstance(out_type, dict):
         out_type_.update(out_type)

returnn/tf/util/data.py

@@ -5039,11 +5039,12 @@ def get_batch_shape_dim_tags(self):
     return self.dim_tags
 
   @classmethod
-  def get_common_data(cls, sources, ignore_feature_dim=False, allow_broadcast_all_sources=NotSpecified):
+  def get_common_data(cls, sources, ignore_feature_dim=False, allow_broadcast_all_sources=NotSpecified, name=None):
     """
     :param list[Data] sources:
     :param bool ignore_feature_dim: when set, the feature dim does not have to match in the sources
     :param bool|NotSpecified allow_broadcast_all_sources:
+    :param str name:
     :return: some generic data where the sources should be compatible to (with copy_compatible_to),
       i.e. it contains the union of all axes from all sources (least common multiple).
       This is always a template, and a new copy.
@@ -5058,7 +5059,7 @@ def get_common_data(cls, sources, ignore_feature_dim=False, allow_broadcast_all_
     common_batch = BatchInfo.get_common_batch_info([src.batch for src in sources if src.batch])
     # Try with the (first) largest.
     common = [s for s in sources if s.batch_ndim == max_ndim][0]
-    common = common.copy_template()
+    common = common.copy_template(name=name)
     common.beam = None  # this will be reset
     if common_batch:
       common.batch = common_batch.copy_set_beam(None)  # the beam will be reset