Using pytorch to run simulations?

1

Hello.

Is it possible to use pytorch to run simulations? I would like to utilize my GPU if possible to make this faster.

Here is an example of one of my simulations using a student T distribution:

starting_cash = 38927.72

sim = pd.DataFrame()
iterations = 50000
time_horizon = 504
annual_investment = 0


for x in range(iterations):
    
    #expected average return
    expected_return = expected_return
    
    #expected volatility based on mean absolute deviation
    expected_volatility = expected_volatility
    
    #collect the data
    stream = []
    
    for i in range(time_horizon):

        #random weights for starting allocation
        weights = np.asarray(sharpe_portfolio.iloc[:,3:].values)
#         print(type(weights))

        starting_allocation = starting_cash * weights
        
        return_change = np.random.standard_t(df = df_return.shape[0] - 1)
        
        end = np.round(starting_allocation.T * (1 + return_change) + annual_investment, 2)
        
        stream.append(end.sum())
        
        starting_allocation = end
#         print(x, i)
    sim[x] = stream

How could I turn that into a pytorch-ran module utilizing the gpu?