fixed module imports

official/vision/beta/projects/yolo/modeling/building_blocks/__init__.py

@@ -4,3 +4,4 @@ from .dark_tiny import DarkTiny
 from .csp_connect import CSPConnect
 from .csp_downsample import CSPDownSample
 from .csp_tiny import CSPTiny
+from .identity import Identity
\ No newline at end of file

official/vision/beta/projects/yolo/modeling/building_blocks/csp_connect_test.py

@@ -3,9 +3,7 @@ import tensorflow.keras as ks
 import numpy as np
 from absl.testing import parameterized
 
-from official.vision.beta.projects.yolo.modeling.building_blocks import CSPDownSample as layer
-from official.vision.beta.projects.yolo.modeling.building_blocks import CSPConnect as layer_companion
-
+from official.vision.beta.projects.yolo.modeling import building_blocks as nn_blocks
 
 class CSPConnect(tf.test.TestCase, parameterized.TestCase):
 
@@ -13,8 +11,8 @@ class CSPConnect(tf.test.TestCase, parameterized.TestCase):
                                   ("downsample", 224, 224, 64, 2))
   def test_pass_through(self, width, height, filters, mod):
     x = ks.Input(shape=(width, height, filters))
-    test_layer = layer(filters=filters, filter_reduce=mod)
-    test_layer2 = layer_companion(filters=filters, filter_reduce=mod)
+    test_layer = nn_blocks.CSPDownSample(filters=filters, filter_reduce=mod)
+    test_layer2 = nn_blocks.CSPConnect(filters=filters, filter_reduce=mod)
     outx, px = test_layer(x)
     outx = test_layer2([outx, px])
     print(outx)
@@ -29,8 +27,8 @@ class CSPConnect(tf.test.TestCase, parameterized.TestCase):
   def test_gradient_pass_though(self, filters, width, height, mod):
     loss = ks.losses.MeanSquaredError()
     optimizer = ks.optimizers.SGD()
-    test_layer = layer(filters, filter_reduce=mod)
-    path_layer = layer_companion(filters, filter_reduce=mod)
+    test_layer = nn_blocks.CSPDownSample(filters, filter_reduce=mod)
+    path_layer = nn_blocks.CSPConnect(filters, filter_reduce=mod)
 
     init = tf.random_normal_initializer()
     x = tf.Variable(

official/vision/beta/projects/yolo/modeling/building_blocks/csp_downsample_test.py

@@ -3,9 +3,7 @@ import tensorflow.keras as ks
 import numpy as np
 from absl.testing import parameterized
 
-from official.vision.beta.projects.yolo.modeling.building_blocks import CSPDownSample as layer
-from official.vision.beta.projects.yolo.modeling.building_blocks import CSPConnect as layer_companion
-
+from official.vision.beta.projects.yolo.modeling import building_blocks as nn_blocks
 
 class CSPDownSample(tf.test.TestCase, parameterized.TestCase):
 
@@ -13,7 +11,7 @@ class CSPDownSample(tf.test.TestCase, parameterized.TestCase):
                                   ("downsample", 224, 224, 64, 2))
   def test_pass_through(self, width, height, filters, mod):
     x = ks.Input(shape=(width, height, filters))
-    test_layer = layer(filters=filters, filter_reduce=mod)
+    test_layer = nn_blocks.CSPDownSample(filters=filters, filter_reduce=mod)
     outx, px = test_layer(x)
     print(outx)
     print(outx.shape.as_list())
@@ -27,8 +25,8 @@ class CSPDownSample(tf.test.TestCase, parameterized.TestCase):
   def test_gradient_pass_though(self, filters, width, height, mod):
     loss = ks.losses.MeanSquaredError()
     optimizer = ks.optimizers.SGD()
-    test_layer = layer(filters, filter_reduce=mod)
-    path_layer = layer_companion(filters, filter_reduce=mod)
+    test_layer = nn_blocks.CSPDownSample(filters, filter_reduce=mod)
+    path_layer = nn_blocks.CSPConnect(filters, filter_reduce=mod)
 
     init = tf.random_normal_initializer()
     x = tf.Variable(

official/vision/beta/projects/yolo/modeling/building_blocks/dark_conv_test.py

@@ -2,8 +2,7 @@ import tensorflow as tf
 import tensorflow.keras as ks
 import tensorflow_datasets as tfds
 from absl.testing import parameterized
-
-from official.vision.beta.projects.yolo.modeling.building_blocks import DarkConv
+from official.vision.beta.projects.yolo.modeling import building_blocks as nn_blocks
 
 
 class DarkConvTest(tf.test.TestCase, parameterized.TestCase):
@@ -17,7 +16,7 @@ class DarkConvTest(tf.test.TestCase, parameterized.TestCase):
     else:
       pad_const = 0
     x = ks.Input(shape=(224, 224, 3))
-    test_layer = DarkConv(filters=64,
+    test_layer = nn_blocks.DarkConv(filters=64,
                           kernel_size=kernel_size,
                           padding=padding,
                           strides=strides,
@@ -37,7 +36,7 @@ class DarkConvTest(tf.test.TestCase, parameterized.TestCase):
     loss = ks.losses.MeanSquaredError()
     optimizer = ks.optimizers.SGD()
     with tf.device("/CPU:0"):
-      test_layer = DarkConv(filters, kernel_size=(3, 3), padding="same")
+      test_layer = nn_blocks.DarkConv(filters, kernel_size=(3, 3), padding="same")
 
     init = tf.random_normal_initializer()
     x = tf.Variable(initial_value=init(shape=(1, 224, 224,

official/vision/beta/projects/yolo/modeling/building_blocks/dark_residual_test.py

@@ -3,8 +3,7 @@ import tensorflow.keras as ks
 import numpy as np
 from absl.testing import parameterized
 
-from official.vision.beta.projects.yolo.modeling.building_blocks import DarkResidual as layer
-
+from official.vision.beta.projects.yolo.modeling import building_blocks as nn_blocks
 
 class DarkResidualTest(tf.test.TestCase, parameterized.TestCase):
 
@@ -16,7 +15,7 @@ class DarkResidualTest(tf.test.TestCase, parameterized.TestCase):
     if downsample:
       mod = 2
     x = ks.Input(shape=(width, height, filters))
-    test_layer = layer(filters=filters, downsample=downsample)
+    test_layer = nn_blocks.DarkResidual(filters=filters, downsample=downsample)
     outx = test_layer(x)
     print(outx)
     print(outx.shape.as_list())
@@ -31,7 +30,7 @@ class DarkResidualTest(tf.test.TestCase, parameterized.TestCase):
   def test_gradient_pass_though(self, filters, width, height, downsample):
     loss = ks.losses.MeanSquaredError()
     optimizer = ks.optimizers.SGD()
-    test_layer = layer(filters, downsample=downsample)
+    test_layer = nn_blocks.DarkResidual(filters, downsample=downsample)
 
     if downsample:
       mod = 2

official/vision/beta/projects/yolo/modeling/building_blocks/dark_tiny_test.py

@@ -3,7 +3,7 @@ import tensorflow.keras as ks
 import numpy as np
 from absl.testing import parameterized
 
-from official.vision.beta.projects.yolo.modeling.building_blocks import DarkTiny
+from official.vision.beta.projects.yolo.modeling import building_blocks as nn_blocks
 
 
 class DarkTinyTest(tf.test.TestCase, parameterized.TestCase):
@@ -12,7 +12,7 @@ class DarkTinyTest(tf.test.TestCase, parameterized.TestCase):
                                   ("last", 224, 224, 1024, 1))
   def test_pass_through(self, width, height, filters, strides):
     x = ks.Input(shape=(width, height, filters))
-    test_layer = DarkTiny(filters=filters, strides=strides)
+    test_layer = nn_blocks.DarkTiny(filters=filters, strides=strides)
     outx = test_layer(x)
     self.assertEqual(width % strides, 0, msg="width % strides != 0")
     self.assertEqual(height % strides, 0, msg="height % strides != 0")
@@ -24,7 +24,7 @@ class DarkTinyTest(tf.test.TestCase, parameterized.TestCase):
   def test_gradient_pass_though(self, width, height, filters, strides):
     loss = ks.losses.MeanSquaredError()
     optimizer = ks.optimizers.SGD()
-    test_layer = DarkTiny(filters=filters, strides=strides)
+    test_layer = nn_blocks.DarkTiny(filters=filters, strides=strides)
 
     init = tf.random_normal_initializer()
     x = tf.Variable(

training_dir/params.yaml

+runtime:
+  all_reduce_alg: null
+  batchnorm_spatial_persistent: false
+  dataset_num_private_threads: null
+  default_shard_dim: -1
+  distribution_strategy: mirrored
+  enable_xla: false
+  gpu_thread_mode: null
+  loss_scale: null
+  mixed_precision_dtype: float32
+  num_cores_per_replica: 1
+  num_gpus: 0
+  num_packs: 1
+  per_gpu_thread_count: 0
+  run_eagerly: false
+  task_index: -1
+  tpu: null
+  worker_hosts: null
+task:
+  gradient_clip_norm: 0.0
+  init_checkpoint: ''
+  logging_dir: null
+  losses:
+    l2_weight_decay: 0.0005
+    label_smoothing: 0.0
+    one_hot: true
+  model:
+    add_head_batch_norm: false
+    backbone:
+      darknet:
+        model_id: cspdarknettiny
+      type: darknet
+    dropout_rate: 0.0
+    input_size: [224, 224, 3]
+    norm_activation:
+      activation: relu
+      norm_epsilon: 0.001
+      norm_momentum: 0.99
+      use_sync_bn: false
+    num_classes: 1001
+  train_data:
+    block_length: 1
+    cache: false
+    cycle_length: 10
+    deterministic: null
+    drop_remainder: true
+    dtype: float16
+    enable_tf_data_service: false
+    global_batch_size: 128
+    input_path: imagenet-2012-tfrecord/train*
+    is_training: true
+    sharding: true
+    shuffle_buffer_size: 10000
+    tf_data_service_address: null
+    tf_data_service_job_name: null
+    tfds_as_supervised: false
+    tfds_data_dir: ''
+    tfds_download: false
+    tfds_name: ''
+    tfds_skip_decoding_feature: ''
+    tfds_split: ''
+  validation_data:
+    block_length: 1
+    cache: false
+    cycle_length: 10
+    deterministic: null
+    drop_remainder: false
+    dtype: float16
+    enable_tf_data_service: false
+    global_batch_size: 128
+    input_path: imagenet-2012-tfrecord/valid*
+    is_training: true
+    sharding: true
+    shuffle_buffer_size: 10000
+    tf_data_service_address: null
+    tf_data_service_job_name: null
+    tfds_as_supervised: false
+    tfds_data_dir: ''
+    tfds_download: false
+    tfds_name: ''
+    tfds_skip_decoding_feature: ''
+    tfds_split: ''
+trainer:
+  allow_tpu_summary: false
+  best_checkpoint_eval_metric: ''
+  best_checkpoint_export_subdir: ''
+  best_checkpoint_metric_comp: higher
+  checkpoint_interval: 10000
+  continuous_eval_timeout: 3600
+  eval_tf_function: true
+  max_to_keep: 5
+  optimizer_config:
+    ema: null
+    learning_rate:
+      polynomial:
+        cycle: false
+        decay_steps: 799000
+        end_learning_rate: 0.0001
+        initial_learning_rate: 0.1
+        name: PolynomialDecay
+        power: 4.0
+      type: polynomial
+    optimizer:
+      sgd:
+        clipnorm: null
+        clipvalue: null
+        decay: 0.0
+        momentum: 0.9
+        name: SGD
+        nesterov: false
+      type: sgd
+    warmup:
+      linear:
+        name: linear
+        warmup_learning_rate: 0
+        warmup_steps: 1000
+      type: linear
+  steps_per_loop: 10000
+  summary_interval: 10000
+  train_steps: 800000
+  train_tf_function: true
+  train_tf_while_loop: true
+  validation_interval: 10000
+  validation_steps: 400