{"id":65385,"date":"2024-06-06T10:49:58","date_gmt":"2024-06-06T07:19:58","guid":{"rendered":"https:\/\/nabfollower.com\/blog\/using-tensorflow-28-on-an-apple-silicon-arm64-chip-771\/"},"modified":"2024-06-06T10:49:58","modified_gmt":"2024-06-06T07:19:58","slug":"using-tensorflow-28-on-an-apple-silicon-arm64-chip-771","status":"publish","type":"post","link":"https:\/\/nabfollower.com\/blog\/using-tensorflow-28-on-an-apple-silicon-arm64-chip-771\/","title":{"rendered":"\u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 TensorFlow 2.8 \u062f\u0631 \u062a\u0631\u0627\u0634\u0647 Apple Silicon arm64"},"content":{"rendered":"
\n

\u0631\u0627\u06cc\u0627\u0646\u0647 \u0645\u0646 \u0627\u062e\u06cc\u0631\u0627\u064b \u06cc\u06a9 \u0631\u0627\u0628\u0637 \u0646\u0627\u062e\u0648\u0634\u0627\u06cc\u0646\u062f \u0628\u0627 \u062f\u06cc \u0647\u06cc\u062f\u0631\u0648\u0698\u0646 \u0645\u0648\u0646\u0648\u06a9\u0633\u06cc\u062f \u062f\u0627\u0634\u062a. \u0628\u0631\u0627\u06cc \u062a\u0639\u06cc\u06cc\u0646 \u0627\u06cc\u0646\u06a9\u0647 \u0622\u06cc\u0627 \u062f\u0648\u0628\u0627\u0631\u0647 \u0628\u0647 \u0632\u0646\u062f\u06af\u06cc \u0628\u0627\u0632 \u0645\u06cc \u06af\u0631\u062f\u062f \u0627\u0645\u0627 \u0638\u0627\u0647\u0631 \u062e\u0648\u0628\u06cc \u0646\u062f\u0627\u0631\u062f. \u0628\u0646\u0627\u0628\u0631\u0627\u06cc\u0646\u060c \u0645\u0646 \u06cc\u06a9 \u0645\u06a9 \u0628\u0648\u06a9 \u062c\u062f\u06cc\u062f \u062e\u0631\u06cc\u062f\u0645\u060c \u06cc\u0639\u0646\u06cc \u0686\u06cc\u067e M3 (arm64). \u0645\u0646 \u062a\u062c\u0631\u0628\u0647 \u062e\u0648\u0628\u06cc \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 M1 \u062f\u0631 \u06a9\u0627\u0631 \u0642\u0628\u0644\u06cc \u062e\u0648\u062f \u062f\u0627\u0634\u062a\u0645\u060c \u0628\u0646\u0627\u0628\u0631\u0627\u06cc\u0646 \u0645\u0634\u062a\u0627\u0642\u0627\u0646\u0647 \u0645\u0646\u062a\u0638\u0631 \u0622\u0646 \u0628\u0648\u062f\u0645.<\/p>\n

\u0627\u0644\u0628\u062a\u0647 \ud83d\ude29 x86<\/code> \u062f\u0631 \u0645\u0642\u0627\u0628\u0644 arm<\/code> \u0645\u0634\u06a9\u0644\u0627\u062a \u0645\u0639\u0645\u0627\u0631\u06cc \u0628\u0644\u0627\u0641\u0627\u0635\u0644\u0647 \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 TensorFlow \u0634\u0631\u0648\u0639 \u0634\u062f.<\/p>\n

\u062f\u0631 \u0627\u06cc\u0646\u062c\u0627 \u0646\u062d\u0648\u0647 \u0631\u0641\u0639 \u0622\u0646 \u0627\u0633\u062a. \u062f\u0631\u062e\u0648\u0627\u0633\u062a \u06a9\u0634\u0634: \u062a\u0635\u0648\u06cc\u0631\u0628\u0631\u062f\u0627\u0631\u06cc \u0633\u0644\u0648\u0644 \u0639\u0645\u06cc\u0642 #229<\/p>\n

DeepCell \u0627\u0632 TF 2.8 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0645\u06cc \u06a9\u0646\u062f\u060c \u0628\u0646\u0627\u0628\u0631\u0627\u06cc\u0646 \u0645\u0627 \u0628\u0627\u06cc\u062f \u0627\u0632 \u0622\u0646 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u06a9\u0646\u06cc\u0645. \u0645\u062a\u0623\u0633\u0641\u0627\u0646\u0647 \u0628\u0633\u062a\u0647 2.8.4 \u0628\u0627 \u0628\u0627\u06cc\u0646\u0631\u06cc \u0647\u0627\u06cc ARM \u0627\u0631\u0627\u0626\u0647 \u0646\u0645\u06cc \u0634\u0648\u062f. \u0627\u062a\u0641\u0627\u0642\u0627 TF 2.16.1 \u0645\u06cc\u06a9\u0646\u062f<\/em> \u062f\u0627\u0631\u0627\u06cc \u0628\u0627\u06cc\u0646\u0631\u06cc arm64 … \u0627\u0645\u0627 \u0645\u0646 \u0646\u0645\u06cc \u062a\u0648\u0627\u0646\u0645 \u0627\u0632 \u0622\u0646 \u062f\u0631 \u0627\u06cc\u0646\u062c\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u06a9\u0646\u0645 \ud83d\ude11<\/p>\n

\u0627\u067e\u0644 \u0645\u062f\u0627\u0631\u06a9\u06cc \u0628\u0631\u0627\u06cc \u0646\u0635\u0628 TensorFlow \u0648 \u067e\u0644\u0627\u06af\u06cc\u0646 “\u0641\u0644\u0632\u06cc” \u062f\u0627\u0631\u062f. \u0628\u0647 \u062e\u0635\u0648\u0635\u060c<\/p>\n

\n

\u0628\u0631\u0627\u06cc TensorFlow \u0646\u0633\u062e\u0647 2.12 \u06cc\u0627 \u0642\u0628\u0644 \u0627\u0632 \u0622\u0646:<\/p>\n

python -m pip install tensorflow-macos\n<\/code><\/pre>\n<\/blockquote>\n
\u062f\u0631 \u0645\u0648\u0631\u062f \u0645\u0627 \u0628\u0647 tensorflow-macos==2.8.0 \u0646\u06cc\u0627\u0632 \u062f\u0627\u0631\u06cc\u0645 \u06a9\u0647 \u062f\u0631 \u062a\u0627\u0631\u06cc\u062e\u0686\u0647 \u0627\u0646\u062a\u0634\u0627\u0631 tensorflow-macos \u06cc\u0627\u0641\u062a \u0645\u06cc \u0634\u0648\u062f.  \u0645\u062a\u0623\u0633\u0641\u0627\u0646\u0647 \u0644\u06cc\u0633\u062a \u0641\u0627\u06cc\u0644 \u0647\u0627 \u0646\u0634\u0627\u0646 \u0645\u06cc \u062f\u0647\u062f \u06a9\u0647 \u0647\u06cc\u0686 \u062a\u0648\u0632\u06cc\u0639 \u067e\u0627\u06cc\u062a\u0648\u0646 3.10 \u0648\u062c\u0648\u062f \u0646\u062f\u0627\u0631\u062f\u060c \u0628\u0646\u0627\u0628\u0631\u0627\u06cc\u0646 \u0628\u0627\u06cc\u062f \u0628\u0647 \u067e\u0627\u06cc\u062a\u0648\u0646 3.9 \u062a\u0646\u0632\u0644 \u0628\u062f\u0647\u0645.<\/p>\n
\u062f\u0631 \u0645\u0648\u0631\u062f tensorflow-metal \u0645\u0633\u062a\u0646\u062f\u0627\u062a \u0628\u0633\u062a\u0647 \u0645\u06cc \u06af\u0648\u06cc\u062f \u06a9\u0647 \u0645\u0627 \u0628\u0647 \u0646\u0633\u062e\u0647 0.4.0 \u0646\u06cc\u0627\u0632 \u062f\u0627\u0631\u06cc\u0645.<\/p>\n
\u0645\u0646 \u06cc\u06a9 \u0641\u0627\u06cc\u0644 \u0645\u0648\u0631\u062f \u0646\u06cc\u0627\u0632 \u062c\u062f\u06cc\u062f \u0631\u0627 \u0628\u0631\u0627\u06cc \u06a9\u0627\u0631\u0628\u0631\u0627\u0646 Mac arm64 \u0628\u0633\u062a\u0647 \u0628\u0646\u062f\u06cc \u06a9\u0631\u062f\u0645:<\/p>\n
\n
$ cat requirements-mac-arm64.txt\ntensorflow-macos==2.8.0\ntensorflow-metal==0.4.0\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0633\u067e\u0633 \u0627\u0644\u0632\u0627\u0645\u0627\u062a \u0645\u06a9 \u0631\u0627 \u0646\u0635\u0628 \u0645\u06cc \u06a9\u0646\u06cc\u062f:<\/p>\n
\n
pip install<\/span> -r<\/span> requirements-mac-arm64.txt\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0627\u0644\u0628\u062a\u0647 \u0647\u0630\u06cc\u0627\u0646 \u0647\u0627 \u0645\u062a\u0648\u0642\u0641 \u0646\u0645\u06cc \u0634\u0648\u0646\u062f!  \u062f\u0631 \u062d\u0627\u0644 \u062f\u0648\u06cc\u062f\u0646 pip install -r requirements.txt<\/code> DeepCell \u0631\u0627 \u0646\u0635\u0628 \u0646\u0645\u06cc \u06a9\u0646\u062f\u060c \u0632\u06cc\u0631\u0627 \u0628\u0633\u062a\u06af\u06cc \u062f\u0627\u0631\u062f tensorflow<\/code> – \u0646\u0647 tensorflow-macos<\/code> (\u06a9\u0647 \u0647\u0645\u0627\u0646 \u0645\u0627\u0698\u0648\u0644 \u067e\u0627\u06cc\u062a\u0648\u0646 \u0631\u0627 \u0641\u0631\u0627\u0647\u0645 \u0645\u06cc \u06a9\u0646\u062f tensorflow<\/code>).<\/p>\n
\u0628\u0646\u0627\u0628\u0631\u0627\u06cc\u0646 \u0645\u0646 \u0622\u0646 \u0631\u0627 \u0628\u0647 \u0627\u06cc\u0646 \u062a\u0631\u062a\u06cc\u0628 \u0627\u062c\u0631\u0627 \u06a9\u0631\u062f\u0645 \u062a\u0627 \u067e\u0633 \u0627\u0632 \u0646\u0635\u0628 \u0648\u0627\u0628\u0633\u062a\u06af\u06cc\u200c\u0647\u0627\u06cc\u06cc \u06a9\u0647 \u0645\u06cc\u200c\u062a\u0648\u0627\u0646\u06cc\u0645\u060c \u0627\u0632 \u0648\u0627\u0628\u0633\u062a\u06af\u06cc\u200c\u0647\u0627 \u0631\u062f \u0634\u0648\u0645:<\/p>\n
\n
pip install<\/span> -r<\/span> requirements-mac-arm64.txt\npip install<\/span> -r<\/span> requirements.txt\npip install<\/span> -r<\/span> requirements.txt --no-deps<\/span>\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0633\u067e\u0633 \u06cc\u06a9 \u0634\u06a9\u0633\u062a \u062c\u0627\u0644\u0628 \u067e\u0631\u0648\u062a\u0648\u0628\u0627\u0641 \u06af\u0631\u0641\u062a\u0645.<\/p>\n
\n
If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.\nIf you cannot immediately regenerate your protos, some other possible workarounds are:\n 1. Downgrade the protobuf package to 3.20.x or lower.\n 2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).\n\nMore information: https:\/\/developers.google.com\/protocol-buffers\/docs\/news\/2022-05-06#python-updates\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0631\u0641\u0639 \u0633\u0631\u06cc\u0639: \u062c\u062f\u06cc\u062f\u062a\u0631\u06cc\u0646 \u0646\u0633\u062e\u0647 3.20.x protobuf\u060c 3.20.3 \u0631\u0627 \u0628\u06af\u06cc\u0631\u06cc\u062f.<\/p>\n
\u0627\u067e\u0644 \u06cc\u06a9 \u0627\u0633\u06a9\u0631\u06cc\u067e\u062a \u0622\u0632\u0645\u0627\u06cc\u0634\u06cc \u0627\u0631\u0627\u0626\u0647 \u0645\u06cc \u062f\u0647\u062f:<\/p>\n
\n
import tensorflow as tf\n\ncifar = tf.keras.datasets.cifar100\n(x_train, y_train), (x_test, y_test) = cifar.load_data()\nmodel = tf.keras.applications.ResNet50(\n    include_top=True,\n    weights=None,\n    input_shape=(32, 32, 3),\n    classes=100,)\n\nloss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False)\nmodel.compile(optimizer=\"adam\", loss=loss_fn, metrics=[\"accuracy\"])\nmodel.fit(x_train, y_train, epochs=5, batch_size=64)\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u06cc\u06a9 \u0645\u062f\u0644 180 \u0645\u06af\u0627\u0628\u0627\u06cc\u062a\u06cc \u062f\u0627\u0646\u0644\u0648\u062f \u0628\u0639\u062f\u0627\u2026 \u0645\u0627 \u0637\u0644\u0627\u06cc\u06cc \u0634\u062f\u06cc\u0645.<\/p>\n
\n
2024-06-05 23:10:30.794862: I tensorflow\/core\/grappler\/optimizers\/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled.\n<\/code><\/pre>\n
\n
\n    \u0648\u0627\u0631\u062f \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n
    \u0627\u0632 \u062d\u0627\u0644\u062a \u062a\u0645\u0627\u0645 \u0635\u0641\u062d\u0647 \u062e\u0627\u0631\u062c \u0634\u0648\u06cc\u062f<\/title>\n    \n<\/svg><\/p>\n<\/div>\n<\/div>\n<\/div>\n
\u0641\u0642\u0637 \u0628\u0631\u0627\u06cc \u062a\u0623\u06cc\u06cc\u062f\u060c \u0627\u062c\u0627\u0632\u0647 \u062f\u0647\u06cc\u062f Activity Monitor \u0631\u0627 \u0628\u0631\u0631\u0633\u06cc \u06a9\u0646\u06cc\u0645 – \u0648 \u0628\u0644\u0647\u060c \u0627\u0632 gpu \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0645\u06cc \u06a9\u0646\u062f.  \ud83c\udf89 \ud83d\ude24<\/p>\n
<\/p>\n
\u0641\u0648\u0648  \u062e\u0648\u0628\u060c \u0627\u0645\u06cc\u062f\u0648\u0627\u0631\u0645 \u0627\u06cc\u0646 \u06cc\u06a9 \u0645\u0648\u0631\u062f \u06cc\u06a9 \u0628\u0627\u0631 \u0628\u0627\u0634\u062f.  \u0628\u06cc\u0634\u062a\u0631 \u062a\u0648\u0633\u0639\u0647 \u0645\u0627 \u0627\u0628\u0631 \u0627\u0633\u062a \u06a9\u0647 x86 \u0627\u0633\u062a\u060c \u0641\u0631\u0645\u062a \u0628\u0627\u06cc\u0646\u0631\u06cc \u0631\u0627\u06cc\u062c \u062a\u0631.<\/p>\n
\u062a\u0627 \u0645\u0627\u062c\u0631\u0627\u062c\u0648\u06cc\u06cc \u0628\u0639\u062f\u06cc \u0645\u0627 \u0628\u0627 \u0628\u0627\u06cc\u0646\u0631\u06cc \u0647\u0627 \u270c<\/p>\n
\u062a\u0635\u0648\u06cc\u0631 \u062c\u0644\u062f \u062a\u0648\u0633\u0637 Kari Shea \u062f\u0631 Unsplash<\/p>\n<\/p><\/div>\n","protected":false},"excerpt":{"rendered":"
\u0631\u0627\u06cc\u0627\u0646\u0647 \u0645\u0646 \u0627\u062e\u06cc\u0631\u0627\u064b \u06cc\u06a9 \u0631\u0627\u0628\u0637 \u0646\u0627\u062e\u0648\u0634\u0627\u06cc\u0646\u062f \u0628\u0627 \u062f\u06cc \u0647\u06cc\u062f\u0631\u0648\u0698\u0646 \u0645\u0648\u0646\u0648\u06a9\u0633\u06cc\u062f \u062f\u0627\u0634\u062a. \u0628\u0631\u0627\u06cc \u062a\u0639\u06cc\u06cc\u0646 \u0627\u06cc\u0646\u06a9\u0647 \u0622\u06cc\u0627 \u062f\u0648\u0628\u0627\u0631\u0647 \u0628\u0647 \u0632\u0646\u062f\u06af\u06cc \u0628\u0627\u0632 \u0645\u06cc \u06af\u0631\u062f\u062f \u0627\u0645\u0627 \u0638\u0627\u0647\u0631 \u062e\u0648\u0628\u06cc \u0646\u062f\u0627\u0631\u062f. \u0628\u0646\u0627\u0628\u0631\u0627\u06cc\u0646\u060c \u0645\u0646 \u06cc\u06a9 \u0645\u06a9 \u0628\u0648\u06a9 \u062c\u062f\u06cc\u062f \u062e\u0631\u06cc\u062f\u0645\u060c \u06cc\u0639\u0646\u06cc \u0686\u06cc\u067e M3 (arm64). \u0645\u0646 \u062a\u062c\u0631\u0628\u0647 \u062e\u0648\u0628\u06cc \u0628\u0627 \u0627\u0633\u062a\u0641\u0627\u062f\u0647 \u0627\u0632 M1 \u062f\u0631 \u06a9\u0627\u0631 \u0642\u0628\u0644\u06cc \u062e\u0648\u062f \u062f\u0627\u0634\u062a\u0645\u060c \u0628\u0646\u0627\u0628\u0631\u0627\u06cc\u0646 \u0645\u0634\u062a\u0627\u0642\u0627\u0646\u0647 \u0645\u0646\u062a\u0638\u0631 \u0622\u0646 \u0628\u0648\u062f\u0645. \u0627\u0644\u0628\u062a\u0647 \ud83d\ude29 …<\/p>\n","protected":false},"author":2,"featured_media":65386,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"","fifu_image_alt":"","footnotes":""},"categories":[339],"tags":[],"class_list":["post-65385","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-dev"],"_links":{"self":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts\/65385","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/comments?post=65385"}],"version-history":[{"count":0,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/posts\/65385\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/media\/65386"}],"wp:attachment":[{"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/media?parent=65385"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/categories?post=65385"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nabfollower.com\/blog\/wp-json\/wp\/v2\/tags?post=65385"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}