schuetoliver
/
handson-ml
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Use cloned model when reusing a pretrained model, fixes #454

main
Aurélien Geron 2021-10-07 17:41:46 +13:00
parent c2a729d62d
commit b3081ceab6
1 changed files with 12 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
11_training_deep_neural_networks.ipynb
View File

@ -970,6 +970,13 @@
"model_B_on_A.add(keras.layers.Dense(1, activation=\"sigmoid\"))" "model_B_on_A.add(keras.layers.Dense(1, activation=\"sigmoid\"))"
] ]
}, },
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Note that `model_B_on_A` and `model_A` actually share layers now, so when we train one, it will update both models. If we want to avoid that, we need to build `model_B_on_A` on top of a *clone* of `model_A`:"
]
},
{ {
"cell_type": "code", "cell_type": "code",
"execution_count": 61, "execution_count": 61,
@ -977,7 +984,9 @@
"outputs": [], "outputs": [],
"source": [ "source": [
"model_A_clone = keras.models.clone_model(model_A)\n", "model_A_clone = keras.models.clone_model(model_A)\n",
"model_A_clone.set_weights(model_A.get_weights())" "model_A_clone.set_weights(model_A.get_weights())\n",
"model_B_on_A = keras.models.Sequential(model_A_clone.layers[:-1])\n",
"model_B_on_A.add(keras.layers.Dense(1, activation=\"sigmoid\"))"
] ]
}, },
{ {
@ -1042,7 +1051,7 @@
"cell_type": "markdown", "cell_type": "markdown",
"metadata": {}, "metadata": {},
"source": [ "source": [
"Great! We got quite a bit of transfer: the error rate dropped by a factor of 4.5!" "Great! We got quite a bit of transfer: the error rate dropped by a factor of 4.9!"
] ]
}, },
{ {
@ -1051,7 +1060,7 @@
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
"(100 - 97.05) / (100 - 99.35)" "(100 - 97.05) / (100 - 99.40)"
] ]
}, },
{ {