From c8f84face87c06c6520292591b6ac8fb7278772c Mon Sep 17 00:00:00 2001
From: HU <franck.hu@ecl20.ec-lyon.fr>
Date: Fri, 1 Dec 2023 17:22:37 +0100
Subject: [PATCH] maj
---
TD2 Deep Learning.ipynb | 102 ++++++++++++++++++++++++++++++++++++++--
1 file changed, 99 insertions(+), 3 deletions(-)
diff --git a/TD2 Deep Learning.ipynb b/TD2 Deep Learning.ipynb
index e581032..ec4ce7d 100644
--- a/TD2 Deep Learning.ipynb
+++ b/TD2 Deep Learning.ipynb
@@ -1871,6 +1871,18 @@
"Apply ther quantization (post and quantization aware) and evaluate impact on model size and accuracy."
]
},
+ {
+ "cell_type": "markdown",
+ "id": "bb08304c",
+ "metadata": {},
+ "source": [
+ "_________\n",
+ "\n",
+ "Dans cette partie, nous allons ajouter un dataset de test. On configure le data_transform pour qu'il prennent en compte les données test contenu dans le dossier test.\n",
+ "\n",
+ "On va aussi configurer la sortie avec la nouvelle structure de neurones."
+ ]
+ },
{
"cell_type": "code",
"execution_count": 42,
@@ -1988,6 +2000,8 @@
"\n",
"# Data augmentation and normalization for training\n",
"# Just normalization for validation\n",
+ "\n",
+ "# on ajoute les data test\n",
"data_transforms = {\n",
" \"train\": transforms.Compose(\n",
" [\n",
@@ -2150,6 +2164,8 @@
"# Replace the final fully connected layer\n",
"# Parameters of newly constructed modules have requires_grad=True by default\n",
"num_ftrs = model.fc.in_features\n",
+ "\n",
+ "# on ajoute les nouvelles couches\n",
"model.fc = nn.Sequential(\n",
" nn.Linear(num_ftrs, 256), # Add a layer with 256 neurons\n",
" nn.ReLU(), # Apply ReLU activation function\n",
@@ -2173,6 +2189,14 @@
"\n"
]
},
+ {
+ "cell_type": "markdown",
+ "id": "ec105406",
+ "metadata": {},
+ "source": [
+ "On a entraîné ce nouveau modèle. on définit maintenant la fonction eval_model(model), qui va évaluer les performances du modèle sur le data test."
+ ]
+ },
{
"cell_type": "markdown",
"id": "8e2bd8fd",
@@ -2211,16 +2235,16 @@
}
],
"source": [
- "\n",
"def eval_model(model):\n",
- "\n",
+ " # initialisation des variables\n",
" running_loss = 0.0\n",
" running_corrects = 0\n",
+ " \n",
+ " #on teste sur les données test\n",
" for inputs, labels in dataloaders[\"test\"]:\n",
" inputs = inputs.to(device)\n",
" labels = labels.to(device)\n",
"\n",
- "\n",
" outputs = model(inputs)\n",
" _, preds = torch.max(outputs, 1)\n",
" loss = criterion(outputs, labels)\n",
@@ -2229,6 +2253,7 @@
" running_loss += loss.item() * inputs.size(0)\n",
" running_corrects += torch.sum(preds == labels.data)\n",
"\n",
+ " #Calcul du loss et de l'accuracy\n",
" loss = running_loss / dataset_sizes[\"test\"]\n",
" acc = running_corrects.double() / dataset_sizes[\"test\"]\n",
" \n",
@@ -2238,6 +2263,77 @@
"print(\"Loss: {:.4f} Acc: {:.4f}\".format(loss, acc))"
]
},
+ {
+ "cell_type": "markdown",
+ "id": "c01bb10c",
+ "metadata": {},
+ "source": [
+ "On obtient une précision de ??????.\n",
+ "\n",
+ "_________\n",
+ "\n",
+ "On va maintenant quantizer le modèle"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 47,
+ "id": "0d8116cf",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "model: int8 \t Size (KB): 45304.25\n",
+ "model: int8 \t Size (KB): 44911.014\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020\n",
+ " warnings.warn(\n",
+ "/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020\n",
+ " warnings.warn(\n",
+ "/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020\n",
+ " warnings.warn(\n",
+ "/Users/franck/Library/Python/3.8/lib/python/site-packages/urllib3/__init__.py:34: NotOpenSSLWarning: urllib3 v2.0 only supports OpenSSL 1.1.1+, currently the 'ssl' module is compiled with 'LibreSSL 2.8.3'. See: https://github.com/urllib3/urllib3/issues/3020\n",
+ " warnings.warn(\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Loss: 0.2149 Acc: 1.0000\n"
+ ]
+ }
+ ],
+ "source": [
+ "quantized_model = torch.quantization.quantize_dynamic(model, dtype=torch.qint8)\n",
+ "print_size_of_model(model, \"int8\")\n",
+ "\n",
+ "print_size_of_model(quantized_model, \"int8\")\n",
+ "\n",
+ "\n",
+ "\n",
+ "loss,acc=eval_model(quantized_model)\n",
+ "print(\"Loss: {:.4f} Acc: {:.4f}\".format(loss, acc))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "913c3dae",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "loss,acc=eval_model(quantized_model)\n",
+ "print(\"Loss: {:.4f} Acc: {:.4f}\".format(loss, acc))"
+ ]
+ },
{
"cell_type": "markdown",
"id": "04a263f0",
--
GitLab