diff options
author | Mark Peschel | 2023-07-08 22:34:48 -0400 |
---|---|---|
committer | Mark Peschel | 2023-07-08 22:36:02 -0400 |
commit | d529c0ede8354b1c3751d9a324d3d6ffcd6db5da (patch) | |
tree | e63026638947c407da888f3bc1e83f6a797dec7a | |
parent | d2a73892d9dafe76e86b9aa5a6b93e400f26ee6c (diff) | |
download | aur-d529c0ede8354b1c3751d9a324d3d6ffcd6db5da.tar.gz |
Make test.py slightly less non-deterministic.
-rw-r--r-- | next.txt | 3 | ||||
-rw-r--r-- | test.py | 20 |
2 files changed, 13 insertions, 10 deletions
@@ -1,6 +1,3 @@ -it occurs to me, about the opt things, we can just do march=native. - build with match=x86_64 -O2 and march=native -O2 and benchmark the results. - remove the optimization notes from github readme clean chroot build party. I am so glad this is over. @@ -33,12 +33,13 @@ def is_gfx1031(): # if len(architectures) != 1, we ignore it since idk what to do then. return {'gfx1031'} == architectures -# You can use environment variables if your GPU works but is not officially supported. -# For example, I have a Radeon RX 6750 XT, which is detected as the unsupported gfx1031 arch, -# but the supported gfx1030 architecture works just dandy. -# This test, when run on such a machine, will override the architecture. -if is_gfx1031(): - os.environ['HSA_OVERRIDE_GFX_VERSION'] = '10.3.0' +def set_gfx_override(): + # You can use environment variables if your GPU works but is not officially supported. + # For example, I have a Radeon RX 6750 XT, which is detected as the unsupported gfx1031 arch, + # but the supported gfx1030 architecture works just dandy. + # This test, when run on such a machine, will override the architecture. + if is_gfx1031(): + os.environ['HSA_OVERRIDE_GFX_VERSION'] = '10.3.0' def uninstall(package_name): subprocess.run(['sudo', 'pacman', '-Rdd', package_name]) @@ -122,12 +123,16 @@ def test_tf_short(): tf.autograph.to_graph(temp) def test_tf_mnist(): + seed = 531 import tensorflow as tf + tf.random.set_seed(seed) + tf.keras.utils.set_random_seed(531) + # Taken from tensorflow.org/tutorials/quickstart/beginner. mnist = tf.keras.datasets.mnist (x_train, y_train), (x_test, y_test) = mnist.load_data() x_train, x_test = x_train / 255., x_test / 255. - + import time start = time.time() @@ -155,6 +160,7 @@ def test_tf_mnist(): print('Your mileage may vary.') def check_packages(tensorflow_path, python_tensorflow_path): + set_gfx_override() uninstall_all() install(tensorflow_path) |