diff --git a/README.md b/README.md
index 44db317..dd23b87 100644
--- a/README.md
+++ b/README.md
@@ -134,7 +134,10 @@ firefox https://github.com/bazelbuild/bazel/releases
 # Install Tensorflow
 git clone tensorflow...
 cd tensorflow
-git checkout remotes/origin/r2.1
+git checkout v2.1.0
+bazel clean
+# Get flags to pass:
+grep flags -m1 /proc/cpuinfo | cut -d ":" -f 2 | tr '[:upper:]' '[:lower:]' | { read FLAGS; OPT="-march=native"; for flag in $FLAGS; do case "$flag" in "sse4_1" | "sse4_2" | "ssse3" | "fma" | "cx16" | "popcnt" | "avx" | "avx2") OPT+=" -m$flag";; esac; done; MODOPT=${OPT//_/\.}; echo "$MODOPT"; }
 ./configure
 # Run Bazel to build pip package. Takes nearly 2 hours to build.
 bazel build --config=opt //tensorflow/tools/pip_package:build_pip_package
diff --git a/wut-worker-mas b/wut-worker-mas
index b0852e1..9def451 100755
--- a/wut-worker-mas
+++ b/wut-worker-mas
@@ -18,8 +18,8 @@
 HOSTNUM=`hostname | sed -e 's/ml//g'`
 
 #export TF_CONFIG='{"cluster": {"worker": [ "ml0-int:2222", "ml1-int:2222", "ml2-int:2222", "ml3-int:2222", "ml4-int:2222", "ml5-int:2222"]}, "task": {"index": '$HOSTNUM', "type": "worker"}}'
-#export TF_CONFIG='{"cluster": {"worker": [ "ml1-int:2222", "ml2-int:2222", "ml3-int:2222", "ml4-int:2222", "ml5-int:2222"]}}'
-export TF_CONFIG='{"cluster": {"chief": [ "ml0-int:2222" ], "worker": [ "ml1-int:2222", "ml2-int:2222", "ml3-int:2222", "ml4-int:2222", "ml5-int:2222"]}, "task": {"index": '$HOSTNUM', "type": "worker"}}'
+export TF_CONFIG='{"cluster": {"worker": [ "ml1-int:2222", "ml2-int:2222", "ml3-int:2222", "ml4-int:2222", "ml5-int:2222"]}}'
+#export TF_CONFIG='{"cluster": {"chief": [ "ml0-int:2222" ], "worker": [ "ml1-int:2222", "ml2-int:2222", "ml3-int:2222", "ml4-int:2222", "ml5-int:2222"]}, "task": {"index": '$HOSTNUM', "type": "worker"}}'
 
 echo $TF_CONFIG
 python3 wut-worker-mas.py
diff --git a/wut-worker-mas.py b/wut-worker-mas.py
index 2026d12..d0b90a2 100644
--- a/wut-worker-mas.py
+++ b/wut-worker-mas.py
@@ -26,11 +26,9 @@ from tensorflow.python.keras.preprocessing import image
 from tensorflow.python.keras.preprocessing.image import img_to_array
 from tensorflow.python.keras.preprocessing.image import ImageDataGenerator
 from tensorflow.python.keras.preprocessing.image import load_img
-#import tensorflow.python.distribute.cluster_resolver
-#from tensorflow.python.distribute.cluster_resolver import TFConfigClusterResolver
-#from tensorflow.python.distribute.cluster_resolver.TFConfigClusterResolver
 
 tf.keras.backend.clear_session()
+tf.config.optimizer.set_jit(True)
 options = tf.data.Options()
 os.environ["TF_CONFIG"] = json.dumps({
     "cluster": {
@@ -39,27 +37,16 @@ os.environ["TF_CONFIG"] = json.dumps({
     },
    "task": {"type": "chief", "index": 0 },
 })
-#os.environ["TF_CONFIG"] = json.dumps({
-#    "cluster": {
-#        "worker": [ "ml1-int:2222", "ml2-int:2222", "ml3-int:2222", "ml4-int:2222", "ml5-int:2222" ]
-#    }#,
-#   #"task": {"type": "worker", "index": 0 },
-#})
 
 print("Tensorflow Version: ", tf.__version__)
 print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))
 print("Num CPUs Available: ", len(tf.config.experimental.list_physical_devices('CPU')))
-#with tf.device("GPU:0"):
-#  tf.ones(())  # Make sure we can run on GPU
-
-# This ensures that XLA and ptxas work well together, and helps with scaling.
 print("XLA_FLAGS='{}'".format(os.getenv("XLA_FLAGS")))
 
 IMG_HEIGHT = 416
 IMG_WIDTH= 804
 batch_size = 32
 epochs = 4
-
 BUFFER_SIZE = 10000
 NUM_WORKERS = 6
 GLOBAL_BATCH_SIZE = 64 * NUM_WORKERS
@@ -68,12 +55,9 @@ GLOBAL_BATCH_SIZE = 64 * NUM_WORKERS
 POSITIVE_DIRECTORY = '/home/jebba/devel/spacecruft/satnogs-wut/data/pos'
 pos_dir = '/home/jebba/devel/spacecruft/satnogs-wut/data/posdir'
 
-from tensorflow.python.distribute.cluster_resolver import SimpleClusterResolver
-
 strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy(
     tf.distribute.experimental.CollectiveCommunication.RING)
 
-
 def get_bytes_and_label(filepath):
   raw_bytes = tf.io.read_file(filepath)
   label = tf.strings.regex_full_match(
@@ -117,13 +101,9 @@ dataset = dataset.map(process_image, num_parallel_calls=AUTOTUNE)
 dataset = dataset.batch(batch_size=32)
 dataset = dataset.prefetch(buffer_size=AUTOTUNE)
 
-
 os.makedirs(profile_dir, exist_ok=True)
 
 # tf.data.Dataset.from_generator
-
-tf.config.optimizer.set_jit(True)
-
 #tf.summary.trace_on(profiler=True)