How to tune Pytorch model in GPU

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
1

I can see the example here and, I use the parameter ‘use_gpu = True’, But Still It is not using the GPU.

Please help

2

Hi @Animesh_Kumar_Paul,

since the example is using skorch, you’ll also have to pass the device to the NeuralNetClassifier. Full example:

"""
An example training a PyTorch NeuralNetClassifier, performing
grid search using TuneGridSearchCV.
The NeuralNetClassifier is derived from a scikit-learn compatible
neural network library that wraps PyTorch. See more at
https://skorch.readthedocs.io/en/stable/index.html
"""

import numpy as np
from sklearn.datasets import make_classification
from torch import nn
import torch.nn.functional as F
from skorch import NeuralNetClassifier
from tune_sklearn import TuneGridSearchCV

X, y = make_classification(10000, 20, n_informative=10, random_state=0)
X = X.astype(np.float32)
y = y.astype(np.int64)


class MyModule(nn.Module):
    def __init__(self, num_units=10, nonlin=F.relu):
        super(MyModule, self).__init__()
        self.dense0 = nn.Linear(20, num_units)
        self.nonlin = nonlin
        self.dropout = nn.Dropout(0.5)
        self.dense1 = nn.Linear(num_units, 10)
        self.output = nn.Linear(10, 2)

    def forward(self, X, **kwargs):
        X = self.nonlin(self.dense0(X))
        X = self.dropout(X)
        X = F.relu(self.dense1(X))
        X = F.softmax(self.output(X))
        return X


net = NeuralNetClassifier(
    MyModule,
    max_epochs=10,
    lr=0.1,
    # Shuffle training data on each epoch
    iterator_train__shuffle=True,
    device="cuda:0"
)

params = {
    "lr": [0.01, 0.02],
    "module__num_units": [10, 20],
}

gs = TuneGridSearchCV(net, params, scoring="accuracy", use_gpu=True)
gs.fit(X, y)
print(gs.best_score_, gs.best_params_)