arthur/handson-ml3
1
0
Fork 0
mirror of https://github.com/ArthurDanjou/handson-ml3.git synced 2026-01-14 12:14:36 +01:00
Code Issues Packages Projects Releases Wiki Activity

Update notebook to latest Colab library versions

Browse Source
This commit is contained in:
Aurélien Geron
2024-10-06 11:13:18 +13:00
parent 779848d82e
commit c02e03e14f
1 changed files with 13 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
14_deep_computer_vision_with_cnns.ipynb
View File

@@ -1147,6 +1147,13 @@
"model = tf.keras.applications.ResNet50(weights=\"imagenet\")" "model = tf.keras.applications.ResNet50(weights=\"imagenet\")"
] ]
}, },
{
"cell_type": "markdown",
"metadata": {},
"source": [
"**Warning**: The expression `load_sample_images()[\"images\"]` returns a Python list of images. However, in the latest versions, Keras does not accept Python lists anymore, so we must convert this list to a tensor. We can do this using `tf.constant()`, but here I have used `K.constant()` instead: it simply calls `tf.constant()` if you are using TensorFlow as the backend (as is the case here), but if you ever decide to use JAX or PyTorch as the backend instead, `K.constant()` will call the appropriate function from the chosen backend."
]
},
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 35, "execution_count": 35,
@@ -1155,7 +1162,8 @@
}, },
"outputs": [], "outputs": [],
"source": [ "source": [
"images = load_sample_images()[\"images\"]\n", "K = tf.keras.backend\n",
"images = K.constant(load_sample_images()[\"images\"])\n",
"images_resized = tf.keras.layers.Resizing(height=224, width=224,\n", "images_resized = tf.keras.layers.Resizing(height=224, width=224,\n",
" crop_to_aspect_ratio=True)(images)" " crop_to_aspect_ratio=True)(images)"
] ]
@@ -1852,7 +1860,7 @@
"cell_type": "markdown", "cell_type": "markdown",
"metadata": {}, "metadata": {},
"source": [ "source": [
"**Note**: the code below used to reuse the optimizer from the previous model. This was fine in earlier versions of TensorFlow, but in more recent versions it can cause some issues, so I added a line to create a new optimizer here." "**Note**: the code below used to reuse the optimizer from the previous model. This was fine in earlier versions of TensorFlow, but in more recent versions it can cause some issues, so I added a line to create a new optimizer here. Also, recent versions of Keras expect one metric per output, so I added the `\"mse\"` metric to the list."
] ]
}, },
{ {
@@ -1874,7 +1882,7 @@
"optimizer = tf.keras.optimizers.SGD(learning_rate=0.01, momentum=0.9) # added this line\n", "optimizer = tf.keras.optimizers.SGD(learning_rate=0.01, momentum=0.9) # added this line\n",
"model.compile(loss=[\"sparse_categorical_crossentropy\", \"mse\"],\n", "model.compile(loss=[\"sparse_categorical_crossentropy\", \"mse\"],\n",
" loss_weights=[0.8, 0.2], # depends on what you care most about\n", " loss_weights=[0.8, 0.2], # depends on what you care most about\n",
" optimizer=optimizer, metrics=[\"accuracy\"])" " optimizer=optimizer, metrics=[\"accuracy\", \"mse\"])"
] ]
}, },
{ {
@@ -2215,7 +2223,7 @@
"provenance": [] "provenance": []
}, },
"kernelspec": { "kernelspec": {
"display_name": "Python 3 (ipykernel)", "display_name": "Python 3",
"language": "python", "language": "python",
"name": "python3" "name": "python3"
}, },
@@ -2229,7 +2237,7 @@
"name": "python", "name": "python",
"nbconvert_exporter": "python", "nbconvert_exporter": "python",
"pygments_lexer": "ipython3", "pygments_lexer": "ipython3",
"version": "3.10.13" "version": "3.9.10"
}, },
"nav_menu": {}, "nav_menu": {},
"toc": { "toc": {