Add model_dir to all tests to avoid "resource not found error". (#6143)

* fix test benchmark_graph_1_gpu_no_dist_strat failing

- Failure only occurs when all 1_gpu tests are run
together with the error:
tensorflow.python.framework.errors_impl.NotFoundError:
Resource localhost/logdir:/tmp/cifar10_model/
N10tensorflow22SummaryWriterInterfaceE does not exist.
[Op:WriteScalarSummary] name: epoch_loss/

Another fix might be to generate a different model_dir
in the core code, but that has other draw backs such as
restarting from the checkpoint.

* Model_dir for all tests.

official/resnet/keras/keras_cifar_benchmark.py

@@ -51,7 +51,7 @@ class Resnet56KerasAccuracy(keras_benchmark.KerasBenchmark):
     FLAGS.data_dir = DATA_DIR
     FLAGS.batch_size = 128
     FLAGS.train_epochs = 182
-    FLAGS.model_dir = self._get_model_dir('keras_resnet56_1_gpu')
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_1_gpu')
     FLAGS.dtype = 'fp32'
     self._run_and_report_benchmark()
 
@@ -62,7 +62,7 @@ class Resnet56KerasAccuracy(keras_benchmark.KerasBenchmark):
     FLAGS.data_dir = DATA_DIR
     FLAGS.batch_size = 128
     FLAGS.train_epochs = 182
-    FLAGS.model_dir = self._get_model_dir('keras_resnet56_eager_1_gpu')
+    FLAGS.model_dir = self._get_model_dir('benchmark_1_gpu')
     FLAGS.dtype = 'fp32'
     FLAGS.enable_eager = True
     self._run_and_report_benchmark()
@@ -74,7 +74,7 @@ class Resnet56KerasAccuracy(keras_benchmark.KerasBenchmark):
     FLAGS.data_dir = DATA_DIR
     FLAGS.batch_size = 128
     FLAGS.train_epochs = 182
-    FLAGS.model_dir = self._get_model_dir('keras_resnet56_eager_2_gpu')
+    FLAGS.model_dir = self._get_model_dir('benchmark_2_gpu')
     FLAGS.dtype = 'fp32'
     FLAGS.enable_eager = True
     self._run_and_report_benchmark()
@@ -86,7 +86,7 @@ class Resnet56KerasAccuracy(keras_benchmark.KerasBenchmark):
     FLAGS.data_dir = DATA_DIR
     FLAGS.batch_size = 128
     FLAGS.train_epochs = 182
-    FLAGS.model_dir = self._get_model_dir('keras_resnet56_2_gpu')
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_2_gpu')
     FLAGS.dtype = 'fp32'
     self._run_and_report_benchmark()
 
@@ -98,7 +98,7 @@ class Resnet56KerasAccuracy(keras_benchmark.KerasBenchmark):
     FLAGS.data_dir = DATA_DIR
     FLAGS.batch_size = 128
     FLAGS.train_epochs = 182
-    FLAGS.model_dir = self._get_model_dir('keras_resnet56_no_dist_strat_1_gpu')
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_1_gpu_no_dist_strat')
     FLAGS.dtype = 'fp32'
     self._run_and_report_benchmark()
 
@@ -145,6 +145,7 @@ class Resnet56KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = True
     FLAGS.turn_off_distribution_strategy = True
+    FLAGS.model_dir = self._get_model_dir('benchmark_1_gpu_no_dist_strat')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -153,6 +154,7 @@ class Resnet56KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = False
     FLAGS.turn_off_distribution_strategy = True
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_1_gpu_no_dist_strat')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -161,6 +163,7 @@ class Resnet56KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = True
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_1_gpu')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -169,6 +172,7 @@ class Resnet56KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = False
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_1_gpu')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -177,6 +181,7 @@ class Resnet56KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 2
     FLAGS.enable_eager = True
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_2_gpu')
     FLAGS.batch_size = 128 * 2  # 2 GPUs
     self._run_and_report_benchmark()
 
@@ -185,6 +190,7 @@ class Resnet56KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 2
     FLAGS.enable_eager = False
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_2_gpu')
     FLAGS.batch_size = 128 * 2  # 2 GPUs
     self._run_and_report_benchmark()
 

official/resnet/keras/keras_imagenet_benchmark.py

@@ -50,7 +50,7 @@ class Resnet50KerasAccuracy(keras_benchmark.KerasBenchmark):
     FLAGS.data_dir = DATA_DIR
     FLAGS.batch_size = 128 * 8
     FLAGS.train_epochs = 90
-    FLAGS.model_dir = self._get_model_dir('keras_resnet50_8_gpu')
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_8_gpu')
     FLAGS.dtype = 'fp32'
     self._run_and_report_benchmark()
 
@@ -61,7 +61,7 @@ class Resnet50KerasAccuracy(keras_benchmark.KerasBenchmark):
     FLAGS.data_dir = DATA_DIR
     FLAGS.batch_size = 128 * 8
     FLAGS.train_epochs = 90
-    FLAGS.model_dir = self._get_model_dir('keras_resnet50_eager_8_gpu')
+    FLAGS.model_dir = self._get_model_dir('benchmark_8_gpu')
     FLAGS.dtype = 'fp32'
     FLAGS.enable_eager = True
     self._run_and_report_benchmark()
@@ -113,6 +113,7 @@ class Resnet50KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = True
     FLAGS.turn_off_distribution_strategy = True
+    FLAGS.model_dir = self._get_model_dir('benchmark_1_gpu_no_dist_strat')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -122,6 +123,7 @@ class Resnet50KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = False
     FLAGS.turn_off_distribution_strategy = True
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_1_gpu_no_dist_strat')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -131,6 +133,7 @@ class Resnet50KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = True
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_1_gpu')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -140,6 +143,7 @@ class Resnet50KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 1
     FLAGS.enable_eager = False
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_1_gpu')
     FLAGS.batch_size = 128
     self._run_and_report_benchmark()
 
@@ -149,6 +153,7 @@ class Resnet50KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 8
     FLAGS.enable_eager = True
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_8_gpu')
     FLAGS.batch_size = 128 * 8  # 8 GPUs
     self._run_and_report_benchmark()
 
@@ -158,6 +163,7 @@ class Resnet50KerasBenchmarkBase(keras_benchmark.KerasBenchmark):
     FLAGS.num_gpus = 8
     FLAGS.enable_eager = False
     FLAGS.turn_off_distribution_strategy = False
+    FLAGS.model_dir = self._get_model_dir('benchmark_graph_8_gpu')
     FLAGS.batch_size = 128 * 8  # 8 GPUs
     self._run_and_report_benchmark()