Extend SliceNdLayer to use a layer as the size argument

returnn/tf/layers/basic.py

@@ -859,39 +859,48 @@ class SliceNdLayer(_ConcatInputLayer):
   def __init__(self, start, size, min_size=None, **kwargs):
     """
     :param LayerBase start: (B,...)
-    :param int|None size: if None, it uses the max possible size, and it becomes a dynamic axis
+    :param int|LayerBase|None size: if None, it uses the max possible size,
+      and it becomes a dynamic axis.
     :param int|None min_size: if size is None, but we want to have a min-size
     """
     super(SliceNdLayer, self).__init__(**kwargs)
     from returnn.tf.util.basic import where_bc, expand_multiple_dims
-    x = self.input_data.copy_as_batch_major()
-    seq_lens = x.get_sequence_lengths() if x.is_time_axis_dynamic() else None  # (B,) or None
+    from returnn.tf.util.data import Data
+    x = self.input_data.copy()
+    seq_lens_data = x.get_time_dim_tag().dyn_size_ext  # (B,) or None
     self.start = start
-    start_data = start.output.copy_as_batch_major()  # e.g. (B,) or (B,T)
+    self.size = size
+    start_data = start.output.copy()  # e.g. (B,) or (B,T)
     start_t = start_data.placeholder
     if size is None:
       if min_size is None:
         min_size = 0
-      if seq_lens is None:
+      if seq_lens_data is None:
         assert isinstance(x.batch_shape[x.time_dim_axis], int)
-        size = tf.maximum(tf.reduce_max(x.batch_shape[x.time_dim_axis] - start_t), min_size)  # scalar
+        size_t = x.batch_shape[x.time_dim_axis] - start_t
       else:
-        # make seq_lens compatible with start_t
-        seq_lens = expand_multiple_dims(  # e.g. (B,) or (B,1)
-          x=seq_lens,
-          axes=[-1] * (len(start_t.shape) - len(seq_lens.shape)))
-        size = tf.maximum(tf.reduce_max(seq_lens - start_t), min_size)  # scalar
+        seq_lens_t = seq_lens_data.copy_compatible_to(start_data, check_sparse=False).placeholder
+        size_t = seq_lens_t - start_t
+      size = tf.maximum(tf.reduce_max(size_t), min_size)  # scalar
+    elif isinstance(size, LayerBase):
+      size_data = size.output.copy()
+      common_data = Data.get_common_data([start_data, size_data])
+      size_data = size_data.copy_compatible_to(common_data)
+      size_t = size_data.placeholder
+      start_data = start_data.copy_compatible_to(common_data)
+      start_t = start_data.placeholder
+      min_size = 0
+      size = tf.maximum(tf.reduce_max(size_t), min_size)  # scalar
+    else:
+      size_t = None
     # for each start index in start_data, we want to gather a slice
     # therefore, the output's first axes are the same as the ones from start_data
     # and the next axis will therefore be the slice axis
     slice_tag = self.output.dim_tags[start_data.batch_ndim]
     assert slice_tag.description.startswith("sliced-time:")
     if not isinstance(size, int):
       # in this case, size is not known before runtime and becomes dynamic and we need to set dyn_size
-      if seq_lens is None:
-        dyn_size = tf.maximum(x.batch_shape[x.time_dim_axis] - start_t, min_size)  # (B,) or (B,T)
-      else:
-        dyn_size = tf.maximum(seq_lens - start_t, min_size)  # (B,) or (B,T)
+      dyn_size = tf.maximum(size_t, min_size)  # (B,) or (B,T)
       dyn_size_ext = Data(
         name=("%s:dyn_size" % slice_tag.description),
         dtype=Data.size_dtype,
@@ -909,20 +918,25 @@ def __init__(self, start, size, min_size=None, **kwargs):
       axis=start_data.batch_ndim)
     # [start+0, start+1, ...]
     gather_positions = tf.expand_dims(start_t, -1) + tf.range(0, size)  # e.g. (B, size) or (B, T, size)
-    if seq_lens is not None:
-      # broadcast from (B,) to the shape of the indices
-      seq_lens = expand_multiple_dims(  # e.g. (B,1) or (B,1,1)
-        x=seq_lens,
-        axes=[-1] * (len(gather_positions.shape) - len(seq_lens.shape)))
+    if seq_lens_data is not None:
+      seq_lens_t = seq_lens_data.copy_compatible_to(
+        gather_positions_data,
+        check_sparse=False).placeholder
       pad_mask = tf.logical_or(  # shape like gather_positions
-        tf.greater(gather_positions, seq_lens - 1),
+        tf.greater(gather_positions, seq_lens_t - 1),
         tf.less(gather_positions, 0))
-      gather_positions = tf.clip_by_value(gather_positions, 0, seq_lens - 1)
+      gather_positions = tf.clip_by_value(gather_positions, 0, seq_lens_t - 1)
     else:
       pad_mask = tf.logical_or(  # shape like gather_positions
         tf.greater(gather_positions, x.batch_shape[1] - 1),
         tf.less(gather_positions, 0))
       gather_positions = tf.clip_by_value(gather_positions, 0, x.batch_shape[1] - 1)
+    if isinstance(self.size, LayerBase):
+      pad_mask = tf.logical_or(tf.greater(gather_positions, tf.expand_dims(start_t + size_t - 1, -1)), pad_mask)
+    pad_mask_data = gather_positions_data.copy_template(
+      name="%s_gather_positions" % self.name,
+      dtype="bool")
+    pad_mask_data.placeholder = pad_mask
     gather_positions_data.placeholder = gather_positions
     position = InternalLayer(
       network=self.network,
@@ -944,28 +958,34 @@ def __init__(self, start, size, min_size=None, **kwargs):
     # the gradient flow would go into wrong frames
     # and might lead to unexpected behavior.
     # So to be on the safe side, we do the masking here.
-    pad_mask = expand_multiple_dims(pad_mask, [-1] * (len(placeholder.shape) - len(pad_mask.shape)))
+    pad_mask_data = pad_mask_data.copy_compatible_to(gather_layer.output, check_sparse=False, check_dtype=False)
+    pad_mask = pad_mask_data.placeholder
     self.output.placeholder = where_bc(pad_mask, tf.zeros_like(placeholder), placeholder)
 
   def get_dep_layers(self):
     """
     :rtype: list[LayerBase]
     """
-    return super(SliceNdLayer, self).get_dep_layers() + [self.start]
+    dep_layers = super(SliceNdLayer, self).get_dep_layers() + [self.start]
+    if isinstance(self.size, LayerBase):
+      dep_layers += [self.size]
+    return dep_layers
 
   @classmethod
   def get_out_data_from_opts(cls, name, sources=(), start=None, size=None, **kwargs):
     """
     :param str name:
     :param list[LayerBase] sources:
     :param LayerBase|None start:
-    :param int|None size:
+    :param int|LayerBase|None size:
     :rtype: Data
     """
     from ..util.data import DimensionTag
-    start_data = start.output.copy_as_batch_major()
-    input_data = sources[0].output.copy_as_batch_major()
+    start_data = start.output.copy()
+    input_data = sources[0].output.copy()
     gather_positions_data = start_data.copy_template(name="%s_gather_positions" % name)
+    if isinstance(size, LayerBase):
+      size = None
     # size might be None here in which case we set the dyn_size in __init__
     tag = DimensionTag(
       kind=DimensionTag.Types.Spatial,
@@ -991,6 +1011,8 @@ def transform_config_dict(cls, d, network, get_layer):
     """
     super(SliceNdLayer, cls).transform_config_dict(d, network=network, get_layer=get_layer)
     d["start"] = get_layer(d["start"])
+    if isinstance(d["size"], str):
+      d["size"] = get_layer(d["size"])
 
 
 class GatherLayer(_ConcatInputLayer):

tests/test_TFNetworkLayer.py

@@ -2820,6 +2820,60 @@ def test_SliceNdLayer_multidimensional_start():
           numpy.testing.assert_equal(orig_seq[t2], segments[b, t, t2])
 
 
+def test_SliceNdLayer_multidimensional_size():
+  with make_scope() as session:
+    n_out = 5
+    n_batch = 3
+    max_seq_len = 10
+    config = Config({
+      "debug_print_layer_output_template": True,
+      "extern_data": {
+        "data": {"dim": n_out},
+        "classes": {"dim": n_out, "sparse": True}
+      }})
+    net = TFNetwork(config=config, train_flag=True)
+    net.construct_from_dict({
+      "output": {
+        "class": "rec", "from": "data:data", "unit": {
+          "const1": {"class": "constant", "value": 1},
+          "start": {"class": "reinterpret_data", "from": "prev:choice", "set_sparse": False},
+          "size": {"class": "combine", "from": ["const1", "start"], "kind": "add"},
+          "slices": {"class": "slice_nd", "from": "base:data:data", "start": "start", "size": "size"},
+          "output": {"class": "reduce", "from": "slices", "mode": "max", "axes": "dyn:-1"},
+          "prob": {"class": "softmax", "from": "data:source", "target": "classes", "loss": "ce"},
+          'choice': {
+            'class': 'choice', 'target': "classes", 'beam_size': 3, 'from': "prob", "input_type": "prob",
+            "initial_output": 0,}}}})
+    session.run(tf_compat.v1.global_variables_initializer())
+    output_layer = net.layers["output"]
+    starts = output_layer.cell.output_layers_net.layers["start"].output.get_placeholder_as_batch_major()
+    sizes = output_layer.cell.output_layers_net.layers["size"].output.get_placeholder_as_batch_major()
+    segments = output_layer.cell.output_layers_net.layers["slices"].output.get_placeholder_as_batch_major()
+    feed = make_feed_dict(net.extern_data.data.values(), n_batch=n_batch, n_time=max_seq_len, same_time=True)
+    starts = session.run(starts, feed_dict=feed)
+    sizes = session.run(sizes, feed_dict=feed)
+    segments = session.run(segments, feed_dict=feed)
+    seq_lens = feed[net.extern_data.data["data"].size_placeholder[0]]
+    input_data = feed[net.extern_data.data["data"].placeholder]
+    max_size = numpy.amax(sizes)
+    max_size = max(max_size, 0)
+    assert segments.shape == (n_batch, max_seq_len, max_size, n_out)
+    for b in range(n_batch):
+      for t in range(max_seq_len):
+        s = starts[b, t]
+        size = sizes[b, t]
+        end = min(s + size, seq_lens[b])
+        orig_seq = input_data[b, s:end]
+        if len(orig_seq) < max_size:
+          orig_seq = numpy.pad(orig_seq, [(0, max_size - len(orig_seq)), (0, 0)], "constant")
+        elif len(orig_seq) > max_size:
+          orig_seq = orig_seq[:max_size]
+        assert orig_seq.shape == (max_size, n_out)
+        orig_seq = numpy.where((numpy.arange(s, s + max_size) >= seq_lens[b])[:, None], 0.0, orig_seq)
+        for t2 in range(max_size):
+          numpy.testing.assert_equal(orig_seq[t2], segments[b, t, t2])
+
+
 def test_SliceNdLayer_set_tag_on_size_tensor():
   with make_scope():
     n_out = 5