official/modeling/hyperparams/params_dict.py

@@ -28,7 +28,7 @@ import yaml
 # values (e.g. floats, ints, etc.), and a lists within brackets.
 _PARAM_RE = re.compile(
     r"""
-  (?P<name>[a-zA-Z][\w\.]*)    # variable name: "var" or "x"
+  (?P<name>[a-zA-Z][\w\.]*)(?P<bracketed_index>\[?[0-9]*\]?)  # variable name: "var" or "x" followed by optional index: "[0]" or "[23]"
   \s*=\s*
   ((?P<val>\'(.*?)\'           # single quote
   |
@@ -223,7 +223,7 @@ class ParamsDict(object):
     """Validate the parameters consistency based on the restrictions.
 
     This method validates the internal consistency using the pre-defined list of
-    restrictions. A restriction is defined as a string which specfiies a binary
+    restrictions. A restriction is defined as a string which specifies a binary
     operation. The supported binary operations are {'==', '!=', '<', '<=', '>',
     '>='}. Note that the meaning of these operators are consistent with the
     underlying Python immplementation. Users should make sure the define
@@ -385,6 +385,8 @@ def nested_csv_str_to_json_str(csv_str):
   if not csv_str:
     return ''
 
+  array_param_map = collections.defaultdict(str)
+  max_index_map = collections.defaultdict(str)
   formatted_entries = []
   nested_map = collections.defaultdict(list)
   pos = 0
@@ -398,6 +400,27 @@ def nested_csv_str_to_json_str(csv_str):
     m_dict = m.groupdict()
     name = m_dict['name']
     v = m_dict['val']
+    bracketed_index = m_dict['bracketed_index']
+    # If we reach the name of the array.
+    if bracketed_index and '.' not in name:
+      # Extract the array's index by removing '[' and ']'
+      index = int(bracketed_index[1:-1])
+      if '.' in v:
+        numeric_val = float(v)
+      else:
+        numeric_val = int(v)
+      # Add the value to the array.
+      if name not in array_param_map:
+        max_index_map[name] = index
+        array_param_map[name] = [None] * (index + 1)
+        array_param_map[name][index] = numeric_val
+      elif index < max_index_map[name]:
+        array_param_map[name][index] = numeric_val
+      else:
+        array_param_map[name] += [None] * (index - max_index_map[name])
+        array_param_map[name][index] = numeric_val
+        max_index_map[name] = index
+      continue
 
     # If a GCS path (e.g. gs://...) is provided, wrap this in quotes
     # as yaml.load would otherwise throw an exception
@@ -407,7 +430,10 @@ def nested_csv_str_to_json_str(csv_str):
     name_nested = name.split('.')
     if len(name_nested) > 1:
       grouping = name_nested[0]
-      value = '.'.join(name_nested[1:]) + '=' + v
+      if bracketed_index:
+        value = '.'.join(name_nested[1:]) + bracketed_index + '=' + v
+      else:
+        value = '.'.join(name_nested[1:]) + '=' + v
       nested_map[grouping].append(value)
     else:
       formatted_entries.append('%s : %s' % (name, v))
@@ -416,6 +442,13 @@ def nested_csv_str_to_json_str(csv_str):
     value = ','.join(value)
     value = nested_csv_str_to_json_str(value)
     formatted_entries.append('%s : %s' % (grouping, value))
+
+  # Add array parameters and check that the array is fully initialized.
+  for name in array_param_map:
+    if any(v is None for v in array_param_map[name]):
+      raise ValueError('Did not pass all values of array: %s' % name)
+    formatted_entries.append('%s : %s' % (name, array_param_map[name]))
+
   return '{' + ', '.join(formatted_entries) + '}'
 
 

official/modeling/hyperparams/params_dict_test.py

@@ -176,7 +176,7 @@ class ParamsDictTest(tf.test.TestCase):
     # Valid rule.
     params = params_dict.ParamsDict({'a': 1, 'c': {'a': 1}}, ['a == c.a'])
 
-    # Overridding violates the existing rule, raise error upon validate.
+    # Overriding violates the existing rule, raise error upon validate.
     params.override({'a': 11})
     with self.assertRaises(KeyError):
       params.validate()
@@ -393,6 +393,23 @@ class IOTest(tf.test.TestCase):
     converted_dict = yaml.load(converted_csv_str)
     self.assertDictEqual(converted_dict, expected_output)
 
+  def test_int_array_param_nested_csv_str_to_json_str(self):
+    csv_str = 'a.b[2]=3,a.b[0]=1,a.b[1]=2'
+    json_str = '{a : {b : [1, 2, 3]}}'
+    converted_csv_str = params_dict.nested_csv_str_to_json_str(csv_str)
+    self.assertEqual(converted_csv_str, json_str)
+
+  def test_float_array_param_nested_csv_str_to_json_str(self):
+    csv_str = 'a.b[1]=3.45,a.b[2]=1.32,a.b[0]=2.232'
+    json_str = '{a : {b : [2.232, 3.45, 1.32]}}'
+    converted_csv_str = params_dict.nested_csv_str_to_json_str(csv_str)
+    self.assertEqual(converted_csv_str, json_str)
+
+  def test_incomplete_array_param_nested_csv_str_to_json_str(self):
+    csv_str = 'a.b[0]=1,a.b[2]=2'
+    self.assertRaises(ValueError, params_dict.nested_csv_str_to_json_str,
+                      csv_str)
+
   def test_csv_str_load_supported_datatypes(self):
     csv_str = 'a=1,b=2.,c=[1,2,3],d=\'hello, there\',e=\"Hi.\"'
     converted_csv_str = params_dict.nested_csv_str_to_json_str(csv_str)

official/pip_package/setup.py

@@ -20,8 +20,8 @@ import sys
 from setuptools import find_packages
 from setuptools import setup
 
-version = '2.12.0'
-tf_version = '2.12.0'  # Major version.
+version = '2.13.0'
+tf_version = '2.13.0'  # Major version.
 
 project_name = 'tf-models-official'
 

official/projects/mae/optimization.py

@@ -92,13 +92,17 @@ class _ViTAdamW(nlp_optimization.AdamWeightDecay):
         and self._vars_substr is not None
         and self._layers_idx is not None
     ):
+      is_decayed = False
       for var_substr, idx in zip(self._vars_substr, self._layers_idx):
         if var_substr in var.name:
           decay_factor = self._layer_decay ** (self._max_idx - idx)
           lr_t = lr_t * decay_factor
+          is_decayed = True
           logging.debug(
               'Applying layer-wise lr decay: %s: %f', var.name, decay_factor)
           break
+      if not is_decayed:
+        logging.debug('Ignore layer-wise lr decay: %s', var.name)
     decay = self._decay_weights_op(var, lr_t, apply_state)
     with tf.control_dependencies([decay]):
       var_device, var_dtype = var.device, var.dtype.base_dtype
@@ -155,13 +159,17 @@ class _ViTAdamW(nlp_optimization.AdamWeightDecay):
         and self._vars_substr is not None
         and self._layers_idx is not None
     ):
+      is_decayed = False
       for var_substr, idx in zip(self._vars_substr, self._layers_idx):
         if var_substr in var.name:
           decay_factor = self._layer_decay ** (self._max_idx - idx)
           lr_t = lr_t * decay_factor
+          is_decayed = True
           logging.debug(
               'Applying layer-wise lr decay: %s: %f', var.name, decay_factor)
           break
+      if not is_decayed:
+        logging.debug('Ignore layer-wise lr decay: %s', var.name)
     decay = self._decay_weights_op(var, lr_t, apply_state)
     with tf.control_dependencies([decay]):
       var_device, var_dtype = var.device, var.dtype.base_dtype

official/vision/tasks/maskrcnn.py

@@ -545,7 +545,7 @@ class MaskRCNNTask(base_task.Task):
       state.update(visualization_utils.update_detection_state(step_outputs))
       # TODO(allenyan): Mapping `detection_masks` (w.r.t. the `gt_boxes`) back
       # to full masks (w.r.t. the image). Disable mask visualization fow now.
-      state.pop('detection_masks')
+      state.pop('detection_masks', None)
 
     if not state:
       # Create an arbitrary state to indicate it's not the first step in the