"
],
"image/png": "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\n"
},
"metadata": {}
}
],
"source": [
"y_preds_pt = np.argmax(preds_pt.predictions, axis = 1)\n",
"\n",
"# plotting confusion matrix\n",
"plot_confusion_matrix(y_preds_pt, y_valid, labels)"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "JmonaWOOGzLC"
},
"source": [
"1. the model got confused with love and joy. It predicted joy when some of them is love which seems natural\n",
"2. also the model got confused with surprise. It predicted some of them as joy and fear instead of surprise"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "paasM8SyEFFY"
},
"source": [
"## Fine-Tunning with Keras\n",
"while using tensorflow we can fine-tune models using the KERAS API. It has built in fit() method and no trainer class like the one in PyTorch API"
]
},
{
"cell_type": "code",
"execution_count": 64,
"metadata": {
"id": "rE3rlKyXFW2O"
},
"outputs": [],
"source": [
"from transformers import TFAutoModelForSequenceClassification\n",
"\n",
"tf_model = (TFAutoModelForSequenceClassification\n",
" .from_pretrained(model_ckpt, num_labels = num_labels))"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "2hy9--QTIjXw"
},
"source": [
"converting the dataset to tf dataset using to_tf_dataset() to meet the requirements for using keras api"
]
},
{
"cell_type": "code",
"execution_count": 65,
"metadata": {
"id": "HPu_NwmFFk1e",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "42a1328b-f1f5-4e6c-b3cc-52683cb9a7d4"
},
"outputs": [
{
"output_type": "stream",
"name": "stderr",
"text": [
"/usr/local/lib/python3.10/dist-packages/datasets/arrow_dataset.py:400: FutureWarning: The output of `to_tf_dataset` will change when a passing single element list for `labels` or `columns` in the next datasets version. To return a tuple structure rather than dict, pass a single string.\n",
"Old behaviour: columns=['a'], labels=['labels'] -> (tf.Tensor, tf.Tensor) \n",
" : columns='a', labels='labels' -> (tf.Tensor, tf.Tensor) \n",
"New behaviour: columns=['a'],labels=['labels'] -> ({'a': tf.Tensor}, {'labels': tf.Tensor}) \n",
" : columns='a', labels='labels' -> (tf.Tensor, tf.Tensor) \n",
" warnings.warn(\n"
]
}
],
"source": [
"# converting the tokenized dataset to tf dataset\n",
"tokenizer_columns = tokenizer.model_input_names\n",
"\n",
"tf_train_dataset = emotions_encoded[\"train\"].to_tf_dataset(\n",
" columns = tokenizer_columns, label_cols = [\"label\"],\n",
" shuffle = True,\n",
" batch_size = batch_size\n",
")\n",
"\n",
"tf_eval_dataset = emotions_encoded[\"validation\"].to_tf_dataset(\n",
" columns = tokenizer_columns,\n",
" label_cols = [\"label\"],\n",
" shuffle = False,\n",
" batch_size = batch_size\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 66,
"metadata": {
"id": "FihUeuNyHTo0",
"colab": {
"base_uri": "https://localhost:8080/"
},
"outputId": "e64d55d0-b0d9-441b-dc79-2d1f257103c0"
},
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"Epoch 1/2\n",
"250/250 [==============================] - 187s 548ms/step - loss: 0.5285 - sparse_categorical_accuracy: 0.8153 - val_loss: 0.2016 - val_sparse_categorical_accuracy: 0.9240\n",
"Epoch 2/2\n",
"250/250 [==============================] - 130s 520ms/step - loss: 0.1430 - sparse_categorical_accuracy: 0.9387 - val_loss: 0.1268 - val_sparse_categorical_accuracy: 0.9425\n"
]
},
{
"output_type": "execute_result",
"data": {
"text/plain": [
""
]
},
"metadata": {},
"execution_count": 66
}
],
"source": [
"# training the model using keras api\n",
"import tensorflow as tf\n",
"\n",
"tf_model.compile(\n",
" optimizer = tf.keras.optimizers.Adam(learning_rate = 5e-5),\n",
" loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits = True),\n",
" metrics = tf.metrics.SparseCategoricalAccuracy()\n",
")\n",
"\n",
"tf_model.fit(tf_train_dataset,\n",
" validation_data = tf_eval_dataset,\n",
" epochs = 2)"
]
},
{
"cell_type": "markdown",
"source": [
"Sparse categorical accuracy is a metric commonly used in machine learning, especially in the context of neural networks for classification tasks. It is calculated as the number of correctly predicted instances for the target class divided by the total number of instances.\n",
"\n",
"To be more specific:\n",
"\n",
"Sparse Categorical Accuracy = Number of Correctly Predicted Instances / Total Number of Instances\n",
"\n",
"In the case of sparse categorical accuracy, it is often used when the labels are integers (e.g., class indices) rather than one-hot encoded vectors. The \"sparse\" in its name refers to the fact that the true labels are provided as integers (sparse representation) rather than one-hot encoded vectors (dense representation).\n",
"\n",
"So, in summary, it is indeed the count of correctly predicted instances divided by the total number of predictions, but it's specifically applied when dealing with integer-encoded class labels."
],
"metadata": {
"id": "aZhNe9LjTKgF"
}
},
{
"cell_type": "code",
"execution_count": 67,
"metadata": {
"id": "tbVpYC9BLbo9",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 581
},
"outputId": "ef8a1a47-86cd-46f7-aca3-e911450fd84d"
},
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"32/32 [==============================] - 6s 139ms/step\n"
]
},
{
"output_type": "display_data",
"data": {
"text/plain": [
"
"
],
"image/png": "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\n"
},
"metadata": {}
}
],
"source": [
"# plotting confusion matrix for both pytorch and tensorflow model and both are trained for 2 epochs\n",
"def plot_confusion_matrix(y_preds_tf, y_preds_pt, y_true, labels):\n",
" fig, axes = plt.subplots(1, 2, figsize = (12, 5))\n",
" cm_tf = confusion_matrix(y_true, y_preds_tf, normalize = 'true')\n",
" cm_pt = confusion_matrix(y_true, y_preds_pt, normalize = 'true')\n",
" disp_tf = ConfusionMatrixDisplay(confusion_matrix = cm_tf, display_labels = labels)\n",
" disp_pt = ConfusionMatrixDisplay(confusion_matrix = cm_pt, display_labels = labels)\n",
" disp_tf.plot(cmap = \"Blues\", values_format = \".2f\", ax = axes[0], colorbar = False)\n",
" disp_pt.plot(cmap = \"Blues\", values_format = \".2f\", ax = axes[1], colorbar = False)\n",
" axes[0].set_title(\"Tensorflow\")\n",
" axes[1].set_title(\"PyTorch\")\n",
" plt.show()\n",
"\n",
"plot_confusion_matrix(y_preds_tf, y_preds_pt, y_valid, labels)"
]
},
{
"cell_type": "markdown",
"source": [
"## Error analysis"
],
"metadata": {
"id": "PdKPJV2YQ15a"
}
},
{
"cell_type": "code",
"source": [
"import numpy as np"
],
"metadata": {
"id": "unFtcZHDURTE"
},
"execution_count": 69,
"outputs": []
},
{
"cell_type": "code",
"execution_count": 70,
"metadata": {
"id": "WxPolEXYWxed"
},
"outputs": [],
"source": [
"from torch.nn.functional import cross_entropy\n",
"\n",
"def forward_pass_with_label(batch):\n",
" # place all input tensorss on the same device as the model\n",
" inputs = {k : v.to(device) for k, v in batch.items()\n",
" if k in tokenizer.model_input_names}\n",
"\n",
" with torch.inference_mode():\n",
" output = model(**inputs)\n",
" pred_label = torch.argmax(output.logits, axis = -1),\n",
" loss = cross_entropy(output.logits,\n",
" batch[\"label\"].to(device),\n",
" reduction = \"none\")\n",
"\n",
" # place outputs on cpu for compatibility with other dataset columns\n",
" return {\"loss\" : loss,\n",
" \"output_logits\" : output.logits}"
]
},
{
"cell_type": "code",
"execution_count": 71,
"metadata": {
"id": "n93cWj4SX1j-",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 77,
"referenced_widgets": [
"5ed007695d9f48c2980d47b6190a477d",
"473785d41f4044289c5ed46eddc0351e",
"9bf8f7fb95fc421eb413d83aee8bd60f",
"57ef3ac27ffa41c581376434822d6d7d",
"80a26df4496a4809b201d0269c32c3bd",
"22ff0cfc94ef4cfea0ccc189e8039e1f",
"f0ce428a653e46678175ab3e0ebcf063",
"f5609260b73d4aa3b58629eea02e6e47",
"21418dc66c3c465f950666e72b6fba7f",
"7383ca96858a48528e1e2a085339c07e",
"274f528f440547668c0b09d780522f98"
]
},
"outputId": "337f6596-8c05-482d-ce1d-fa22965654e5"
},
"outputs": [
{
"output_type": "display_data",
"data": {
"text/plain": [
"Map: 0%| | 0/2000 [00:00, ? examples/s]"
],
"application/vnd.jupyter.widget-view+json": {
"version_major": 2,
"version_minor": 0,
"model_id": "5ed007695d9f48c2980d47b6190a477d"
}
},
"metadata": {}
}
],
"source": [
"# using map method to compute the losses for all the samples\n",
"\n",
"# convert our dataset back to pytorch tensors\n",
"emotions_encoded.set_format(\"torch\",\n",
" columns = [\"input_ids\", \"attention_mask\", \"label\"])\n",
"\n",
"# compute loss values\n",
"emotions_encoded[\"validation\"] = emotions_encoded[\"validation\"].map(forward_pass_with_label,\n",
" batched = True,\n",
" batch_size = 16)"
]
},
{
"cell_type": "code",
"source": [
"emotions_encoded.set_format(\"pandas\")\n",
"cols = [\"text\", \"label\", \"output_logits\", \"loss\"]\n",
"df_test = emotions_encoded[\"validation\"][:][cols]"
],
"metadata": {
"id": "ALw1RBBPZHnF"
},
"execution_count": 72,
"outputs": []
},
{
"cell_type": "code",
"source": [
"df_test[\"predicted_label\"] = df_test[\"output_logits\"].apply(lambda x : np.argmax(x))"
],
"metadata": {
"id": "gaAhmcUqeJjM"
},
"execution_count": 73,
"outputs": []
},
{
"cell_type": "code",
"source": [
"label_dict = {i : labels[i] for i in range(len(labels))}\n",
"label_dict"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "ChICVMW0e4rM",
"outputId": "90879c9d-f3e0-4e0b-8a5e-afc2706e7ea9"
},
"execution_count": 74,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"{0: 'sadness', 1: 'joy', 2: 'love', 3: 'anger', 4: 'fear', 5: 'surprise'}"
]
},
"metadata": {},
"execution_count": 74
}
]
},
{
"cell_type": "code",
"source": [
"df_test[[\"predicted_label\", \"label\"]] = df_test[[\"predicted_label\", \"label\"]].replace(label_dict)"
],
"metadata": {
"id": "LYvAlU5ZfMIO"
},
"execution_count": 75,
"outputs": []
},
{
"cell_type": "code",
"source": [
"df_test.sort_values(\"loss\", ascending = False).head(10)"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 883
},
"id": "kMH5KeVPnP2c",
"outputId": "f7da6af3-3a2d-4cd7-a631-e61c35976479"
},
"execution_count": 90,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
" text label \\\n",
"1950 i as representative of everything thats wrong ... surprise \n",
"1963 i called myself pro life and voted for perry w... joy \n",
"882 i feel badly about reneging on my commitment t... love \n",
"1801 i feel that he was being overshadowed by the s... love \n",
"1509 i guess this is a memoir so it feels like that... joy \n",
"1870 i guess i feel betrayed because i admired him ... joy \n",
"1500 i guess we would naturally feel a sense of lon... anger \n",
"1111 im lazy my characters fall into categories of ... joy \n",
"405 i have been feeling extraordinarily indecisive... fear \n",
"1683 i had applied for a job and they had assured m... anger \n",
"\n",
" output_logits loss \\\n",
"1950 [3.4680815, -1.1708108, -1.8415996, 1.1100867,... 5.556944 \n",
"1963 [3.8983426, -1.3576488, -1.1388916, -0.1126643... 5.305911 \n",
"882 [3.4326234, -0.59110284, -1.6783965, 0.7716352... 5.220197 \n",
"1801 [4.071471, -0.75424206, -0.87233037, -0.929799... 4.979969 \n",
"1509 [0.13551404, -1.288383, -1.3879347, -0.9505784... 4.970778 \n",
"1870 [3.5769184, -1.1778387, -0.46349433, -0.818754... 4.823200 \n",
"1500 [4.2843747, -0.8608792, -1.1037171, -0.4985274... 4.809717 \n",
"1111 [0.23640306, -1.6873248, -2.0475214, 1.9557683... 4.565818 \n",
"405 [-2.1552153, 3.273236, 0.4525951, -2.016532, -... 4.296227 \n",
"1683 [-0.907729, 3.4793534, -1.1827734, -0.7358772,... 4.273014 \n",
"\n",
" predicted_label \n",
"1950 sadness \n",
"1963 sadness \n",
"882 sadness \n",
"1801 sadness \n",
"1509 fear \n",
"1870 sadness \n",
"1500 sadness \n",
"1111 fear \n",
"405 joy \n",
"1683 joy "
],
"text/html": [
"\n",
"
\n",
"
\n",
"\n",
"
\n",
" \n",
"
\n",
"
\n",
"
text
\n",
"
label
\n",
"
output_logits
\n",
"
loss
\n",
"
predicted_label
\n",
"
\n",
" \n",
" \n",
"
\n",
"
1950
\n",
"
i as representative of everything thats wrong ...
\n",
"
surprise
\n",
"
[3.4680815, -1.1708108, -1.8415996, 1.1100867,...
\n",
"
5.556944
\n",
"
sadness
\n",
"
\n",
"
\n",
"
1963
\n",
"
i called myself pro life and voted for perry w...
\n",
"
joy
\n",
"
[3.8983426, -1.3576488, -1.1388916, -0.1126643...
\n",
"
5.305911
\n",
"
sadness
\n",
"
\n",
"
\n",
"
882
\n",
"
i feel badly about reneging on my commitment t...
\n",
"
love
\n",
"
[3.4326234, -0.59110284, -1.6783965, 0.7716352...
\n",
"
5.220197
\n",
"
sadness
\n",
"
\n",
"
\n",
"
1801
\n",
"
i feel that he was being overshadowed by the s...
\n",
"
love
\n",
"
[4.071471, -0.75424206, -0.87233037, -0.929799...
\n",
"
4.979969
\n",
"
sadness
\n",
"
\n",
"
\n",
"
1509
\n",
"
i guess this is a memoir so it feels like that...
\n",
"
joy
\n",
"
[0.13551404, -1.288383, -1.3879347, -0.9505784...
\n",
"
4.970778
\n",
"
fear
\n",
"
\n",
"
\n",
"
1870
\n",
"
i guess i feel betrayed because i admired him ...
\n",
"
joy
\n",
"
[3.5769184, -1.1778387, -0.46349433, -0.818754...
\n",
"
4.823200
\n",
"
sadness
\n",
"
\n",
"
\n",
"
1500
\n",
"
i guess we would naturally feel a sense of lon...
\n",
"
anger
\n",
"
[4.2843747, -0.8608792, -1.1037171, -0.4985274...
\n",
"
4.809717
\n",
"
sadness
\n",
"
\n",
"
\n",
"
1111
\n",
"
im lazy my characters fall into categories of ...
\n",
"
joy
\n",
"
[0.23640306, -1.6873248, -2.0475214, 1.9557683...
\n",
"
4.565818
\n",
"
fear
\n",
"
\n",
"
\n",
"
405
\n",
"
i have been feeling extraordinarily indecisive...
\n",
"
fear
\n",
"
[-2.1552153, 3.273236, 0.4525951, -2.016532, -...
\n",
"
4.296227
\n",
"
joy
\n",
"
\n",
"
\n",
"
1683
\n",
"
i had applied for a job and they had assured m...
\n",
"
anger
\n",
"
[-0.907729, 3.4793534, -1.1827734, -0.7358772,...
\n",
"
4.273014
\n",
"
joy
\n",
"
\n",
" \n",
"
\n",
"
\n",
"
\n",
"\n",
"
\n",
" \n",
"\n",
" \n",
"\n",
" \n",
"
\n",
"\n",
"\n",
"
\n",
" \n",
"\n",
"\n",
"\n",
" \n",
"
\n",
"\n",
"
\n",
"
\n"
]
},
"metadata": {},
"execution_count": 90
}
]
},
{
"cell_type": "code",
"source": [
"df_test.sort_values(\"loss\", ascending = True).head(10)"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 883
},
"id": "unXbtuXznabR",
"outputId": "a149e79c-d05c-4707-d8d7-c03c684ac6e7"
},
"execution_count": 92,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
" text label \\\n",
"1873 i feel practically virtuous this month i have ... joy \n",
"1505 im feeling hopeful about a great deal of thing... joy \n",
"1205 i log on feeling vaguely sociable and after a ... joy \n",
"578 i got to christmas feeling positive about the ... joy \n",
"941 i expected but it did feel hopeful and it defi... joy \n",
"1265 im feeling more hopeful today than i did yeste... joy \n",
"1263 i feel this way about blake lively joy \n",
"530 i feel pretty safe but i do realize that we do... joy \n",
"1421 i feel undeservingly lucky to be surrounded by... joy \n",
"1449 im also feeling brave enough to publish my tho... joy \n",
"\n",
" output_logits loss \\\n",
"1873 [-1.4237986, 4.5504737, -0.7130102, -1.2840729... 0.016220 \n",
"1505 [-1.3845941, 4.5399165, -0.666625, -1.4946276,... 0.016611 \n",
"1205 [-1.3749591, 4.5414786, -0.58795214, -1.436796... 0.016634 \n",
"578 [-1.3239833, 4.533265, -0.67640454, -1.3442787... 0.016641 \n",
"941 [-1.3034065, 4.5182724, -0.6394909, -1.4151322... 0.016916 \n",
"1265 [-1.2545577, 4.517784, -0.73379093, -1.4223977... 0.016923 \n",
"1263 [-1.5053567, 4.519288, -0.5932072, -1.4356557,... 0.017054 \n",
"530 [-1.103302, 4.513775, -0.7744781, -1.2526917, ... 0.017084 \n",
"1421 [-1.333859, 4.5334177, -0.5009244, -1.4169365,... 0.017085 \n",
"1449 [-1.4296385, 4.545243, -0.4256305, -1.3768531,... 0.017189 \n",
"\n",
" predicted_label \n",
"1873 joy \n",
"1505 joy \n",
"1205 joy \n",
"578 joy \n",
"941 joy \n",
"1265 joy \n",
"1263 joy \n",
"530 joy \n",
"1421 joy \n",
"1449 joy "
],
"text/html": [
"\n",
"