Open MithrilTuxedo opened 3 years ago
I have run this on an iMac mini with M1 processor and 16GB of RAM. I observe similar issues when using high resolution images (or changing the number of layers you let contribute). You can observe the memory allocated to Python with activity monitor, I noticed things started going wrong when approaching 12 GB allocated to Python. But if you stay within the memory limits it runs very fast!
If I adjust the image size down to 600x300 it locks up and logs out. If I adjust the image size down to 300x200 it uses <2GB of memory, but still locks up.
The last time I ran it (300x200) it actually restarted the computer and prompted to submit a system bug report when I logged in.
When running a Keras example (https://keras.io/examples/generative/deep_dream/) using slightly larger input images for input and output (1200x800) my system locks up (cursor still moves but otherwise unresponsive) and eventually returns me to the login screen (some sort of crash, though I get no dialog) when tensorflow-macos is installed in my venv.
macOS Big Sur 11.2.1 iMac (Retina 5K, 27-inch, 2017) 3.4 GHz Quad-Core Intel Core i5 40 GB 2400 MHz DDR4 Radeon Pro 570 4 GB
Python 3.8.2 (default, Dec 21 2020, 15:06:04) [Clang 12.0.0 (clang-1200.0.32.29)] on darwin
pip list
It works: