File size: 9,602 Bytes
6f1e670 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 | {
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Introduction\n",
"\n",
"In the previous notebook, we saw how to use the ```splitters```, ```featurizers```, ```predictors```, and ```proposers``` modules to train and deploy models to perform property prediction. In this notebook, we will see how to test multiple methods of data splits, featurizations, and models at once. We will initialize several different ```splitters```, ```featurizers```, and ```predictors``` and test them all at once."
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"from model.splitters import *\n",
"from model.featurizers import *\n",
"from model.predictors import *\n",
"from model.proposers import *"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Setting up\n",
"\n",
"First, define the following variables as before:\n",
"- ```experiment_name```: the name of the experiment\n",
"\n",
"- ```protein_name```: the name of the protein\n",
"\n",
"- ```wt_file```: the path to the wildtype sequence\n",
"\n",
"- ```training_dataset_fname```: the path to the training dataset"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"experiment_name = \"example_experiment\"\n",
"protein_name = \"example_protein\"\n",
"wt_file = \"../../../data/example_protein/apex.fasta\"\n",
"training_dataset_fname = '../../../data/example_protein/example_dataset.csv'"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Non-neural network models\n",
"\n",
"We will show an example for non-neural network models first."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"First, we initialize our desired splitters and save them in a list."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"random_splitter = RandomProteinSplitter(protein_name, training_dataset_fname, wt_file, csv_has_header=True, use_cache=True, y_scaling=False, val_split=None)\n",
"round_splitter = RoundProteinSplitter(protein_name, training_dataset_fname, wt_file, csv_has_header=True, use_cache=True, y_scaling=False, val_split=None)\n",
"\n",
"random_splitter.split_data(test_size=0.2)\n",
"round_splitter.split_data(min_test_round=1, max_train_round=0)\n",
"\n",
"splitters = [random_splitter, round_splitter]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Next, we initialize our desired featurizers."
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"onehot = OneHotFeaturizer(protein=protein_name, use_cache=True)\n",
"georgiev = GeorgievFeaturizer(protein=protein_name, use_cache=True)\n",
"\n",
"featurizers = [onehot, georgiev]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Next, we initialize our desired predictors in the form of a list."
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"predictors = [RidgeRegressor, RandomForestRegressor]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Finally, we run the following function ```run_model_experiments``` to train and test all the models. The function will return a pandas dataframe with the results, which are also saved in the directory where the training dataset is located in the folder ```model_cache/dataset_name/results```\n",
"\n",
"```run_model_experiments``` takes the following arguments:\n",
"\n",
"- ```splitters```: A list of splitters to use for the experiment.\n",
"\n",
"- ```featurizers```: A list of featurizers to use for the experiment.\n",
"\n",
"- ```predictors```: A list of predictors to use for the experiment.\n",
"\n",
"- ```experiment_name```: The name of the experiment.\n",
"\n",
"- ```use_cache```: Whether to use the cache for the splitters, featurizers, and predictors.\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"results = run_model_experiments(splitters, featurizers, predictors, experiment_name,use_cache=False)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"results"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Neural network models\n",
"\n",
"Now, we will show an example for neural network models. You will need to have a wandb account to run this example.\n",
"\n",
"For neural network models, we used wandb to perform hyperparameter sweeps, which include comparing different methods of splitting and featurizing.\n",
"\n",
"We will initialize the same splitters and featurizers as before, and train various fully connected neural networks with different architectures and hyperparameters."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"random_splitter = RandomProteinSplitter(protein_name, training_dataset_fname, wt_file, csv_has_header=True, use_cache=True, y_scaling=False, val_split=0.15)\n",
"round_splitter = RoundProteinSplitter(protein_name, training_dataset_fname, wt_file, csv_has_header=True, use_cache=True, y_scaling=False, val_split=0.15)\n",
"\n",
"random_splitter.split_data(test_size=0.2)\n",
"round_splitter.split_data(min_test_round=1, max_train_round=0)\n",
"\n",
"splitters = [random_splitter, round_splitter]"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"onehot = OneHotFeaturizer(protein=protein_name, use_cache=True) # just as an example, we will just use onehot featurizer\n",
"\n",
"featurizers = [onehot]"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"models = [Fcn]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"For neural network models, we use the function ```run_nn_model_experiments``` to train and test all the models. Unlike the previous function, the results of the model training are saved on the wandb server.\n",
"\n",
"```run_nn_model_experiments``` takes the following arguments:\n",
"\n",
"- ```splitters```: A list of splitters to use for the experiment.\n",
"\n",
"- ```featurizers```: A list of featurizers to use for the experiment.\n",
"\n",
"- ```models```: A list of models to use for the experiment.\n",
"\n",
"- ```experiment_name```: The name of the experiment.\n",
"\n",
"- ```use_cache```: Whether to use the cache for the splitters, featurizers, and predictors.\n",
"\n",
"- ```sweep_depth```: The depth of the hyperparameter sweep. Options are 'test', 'standard', and 'custom'. 'test' is to test the training on a single hyperparameter configuration. 'standard' will test a standard set of hyperparameter configurations. 'custom' is a modified version of 'standard' that can be modified to test a subset of hyperparameter configurations.\n",
"\n",
"\n",
"- ```search_method```: The method of hyperparameter search. Options are 'test', 'grid', and 'bayes'. 'test' is to test the training on a single hyperparameter configuration. 'grid' is to test the training on all possible hyperparameter configurations. 'bayes' is to test the training on a bayesian search of hyperparameter configurations.\n",
"\n",
"- ```count```: The number of different hyperparameter configurations to test. This is only used if ```search_method``` is 'bayes'.\n",
"\n",
"If the user wants to modify the hyperparameter sweeps to test other hyperparameters, then the configuration files for the sweep configs can be found in ```model/predictors/sweep_configs```."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"tags": []
},
"outputs": [],
"source": [
"run_nn_model_experiments(splitters, \n",
" featurizers, \n",
" models, \n",
" experiment_name=experiment_name,\n",
" use_cache=False,\n",
" sweep_depth='test', \n",
" search_method='test',\n",
" count=1\n",
" )"
]
}
],
"metadata": {
"environment": {
"kernel": "multievolve",
"name": "workbench-notebooks.m128",
"type": "gcloud",
"uri": "us-docker.pkg.dev/deeplearning-platform-release/gcr.io/workbench-notebooks:m128"
},
"kernelspec": {
"display_name": "multievolve",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.11.11"
}
},
"nbformat": 4,
"nbformat_minor": 4
}
|