|
42 | 42 | "id": "hRTa3Ee15WsJ" |
43 | 43 | }, |
44 | 44 | "source": [ |
45 | | - "# Retrain a classification model for Edge TPU with quant-aware training (TF 1.11)" |
| 45 | + "# Retrain a classification model for Edge TPU with quant-aware training (TF 1.15)" |
46 | 46 | ] |
47 | 47 | }, |
48 | 48 | { |
|
54 | 54 | "source": [ |
55 | 55 | "This notebook uses a set of TensorFlow training scripts to perform transfer-learning on a quantization-aware classification model and then convert it for compatibility with the [Edge TPU](https://coral.ai/products/).\n", |
56 | 56 | "\n", |
57 | | - "Specifically, this tutorial shows you how to perform [fine-tuning](https://github.com/tensorflow/models/blob/master/research/slim/README.md#fine-tuning-a-model-from-an-existing-checkpoint) on the MobileNet V1 model so it can recognize a new set of classes (five types of flowers), using TensorFlow r1.11.\n", |
| 57 | + "Specifically, this tutorial shows you how to perform [fine-tuning](https://github.com/tensorflow/models/blob/master/research/slim/README.md#fine-tuning-a-model-from-an-existing-checkpoint) on the MobileNet V1 model so it can recognize a new set of classes (five types of flowers), using TensorFlow r1.15.\n", |
58 | 58 | "\n", |
59 | 59 | "Beware that, compared to a desktop computer, this training can take much longer in Colab because Colab provides limited resources for long-running operations. So you'll likely see faster training speeds if you [connect this notebook to a local runtime](https://research.google.com/colaboratory/local-runtimes.html), or instead follow the [tutorial to run this training in Docker](https://coral.ai/docs/edgetpu/retrain-classification/) (which includes more documentation about this process).\n", |
60 | 60 | "\n", |
|
82 | 82 | "outputs": [], |
83 | 83 | "source": [ |
84 | 84 | "! pip uninstall tensorflow -y\n", |
85 | | - "! pip install tensorflow==1.11" |
| 85 | + "! pip install tensorflow-gpu==1.15" |
86 | 86 | ] |
87 | 87 | }, |
88 | 88 | { |
|
0 commit comments