PureFrameworkTensorFoundError, Runtime error -FedeartedLearning

Question

I am trying a Linear Regression algorithm with Federated learning using Pytorch and I face the following error. I am implementing it on Colab. According to me this error might be due to some code line in the train() function. Kindly help is you have worked with Pysyft and have faced such error before.

RuntimeError: invalid argument 8: lda should be at least max(1, 0), but have 0 at /pytorch/aten/src/TH/generic/THBlas.cpp:363

And the following is the code:

#import the necessasry packages
import torch
from torch.autograd import Variable
import torch.nn as nn
import torch.nn.functional as F
import syft as sy

#create target and data variables as tensors
x_data=Variable(torch.Tensor([[1.0],[0.0],[1.0],[0.0]]))
y_data=Variable(torch.Tensor([[0.0],[0.0],[1.0],[1.0]]))

#Create virtual Workers
hook = sy.TorchHook(torch)
bob = sy.VirtualWorker(hook, id="bob")
alice = sy.VirtualWorker(hook, id="alice")

data_bob = x_data[0:2]
target_bob = y_data[0:2]
data_alice = x_data[2:0]
target_alice = y_data[2:0]

#creating a class that does Linear Regression
class LinearRegression (nn.Module):

  def __init__(self):
    super(LinearRegression,self). __init__ ()
    self.linear = torch.nn.Linear(1,1)

  def forward(self, x):
    y_pred = self.linear(x)
    return y_pred

#assign the function to the variable name 'Model'
model=LinearRegression()

#send the data to the virtual worker pointers
data_bob = data_bob.send(bob)
data_alice = data_alice.send(alice)

target_bob = target_bob.send(bob)
target_alice = target_alice.send(alice)

# organize pointers into a list
datasets = [(data_bob,target_bob),(data_alice,target_alice)]

#create optimizer and calculate the loss
opt = torch.optim.SGD(params=model.parameters(),lr=0.1)
criterion = torch.nn.MSELoss(size_average=False)

def train():
  opt = torch.optim.SGD(params=model.parameters(),lr=0.1)
  for epoch in range (20):
    model.train()
    print("Training started..")

    for x_data,y_data in datasets:

      model.send(x_data.location) 

      opt.zero_grad()

       #forwardpass
       #the model here is the linear regression model
      y_pred = model(x_data)

      #ComputeLoss
      loss=criterion(y_pred,y_data)

      #BackwardPass
      loss.backward()

      opt.step()

      model.get() 

      print(loss.get())

train()

Answer 1

you have a typo here:

data_alice = x_data[2:0]
target_alice = y_data[2:0]

Should be [2:]

Because data_alice is failing, you had this error.