HELP!CUDA cannot be used with my own dataset

[AnBydAm]

My dataset was built based on a.csv file of my own. In the previous version, I could use model.train to train the model and could use cuda normally. But now I'm using the latest pykan0.2.4, where model.train is replaced by model.fit, but I can't use cuda in model.fit,here's my code： 1、I first took the following code to create the dataset from a.csv file： import pandas as pd import torch import numpy as np from sklearn.preprocessing import StandardScaler import os os.environ['CUBLAS_WORKSPACE_CONFIG'] = ':4096:8' torch.set_default_dtype(torch.float64) device = 'cuda' data = pd.read_csv('0°冬夏半年辐射能量12000.csv') bp_year_range = (0, 12000) filtered_data = data[(data['BP Year'] >= bp_year_range[0]) & (data['BP Year'] <= bp_year_range[1])] split_index = int(len(filtered_data) * 0.8) duration = filtered_data['Winter year Days'][:split_index].values.reshape(-1, 1) intensity = filtered_data['Winter year Intensity'][:split_index].values.reshape(-1, 1) train_input = np.hstack((duration, intensity)) train_label = filtered_data['Winter year energy '][:split_index].values.reshape(-1, 1) duration_test = filtered_data['Winter year Days'][split_index:].values.reshape(-1, 1) intensity_test = filtered_data['Winter year Intensity'][split_index:].values.reshape(-1, 1) test_input = np.hstack((duration_test, intensity_test)) test_label = filtered_data['Winter year energy '][split_index:].values.reshape(-1, 1) scaler_input = StandardScaler() scaler_label = StandardScaler() scaler_input.fit(train_input) scaler_label.fit(train_label) train_input_scaled = scaler_input.transform(train_input) test_input_scaled = scaler_input.transform(test_input) train_label_scaled = scaler_label.transform(train_label) test_label_scaled = scaler_label.transform(test_label) train_input_tensor = torch.tensor(train_input_scaled, dtype=torch.float64).to(device) train_label_tensor = torch.tensor(train_label_scaled, dtype=torch.float64).to(device) test_input_tensor = torch.tensor(test_input_scaled, dtype=torch.float64).to(device) test_label_tensor = torch.tensor(test_label_scaled, dtype=torch.float64).to(device) dataset = { 'train_input': train_input_tensor, 'test_input': test_input_tensor, 'train_label': train_label_tensor, 'test_label': test_label_tensor } 2、Both model creation and initialization go smoothly： from kan import * model = KAN(width=[2,3,1], grid=5, k=3).to(device) model(dataset['train_input']); model.plot() 3、An error occurred when using model.fit： model.fit(dataset, opt="LBFGS", steps=20, lamb=0.01, lamb_entropy=10.);

RuntimeError Traceback (most recent call last) Cell In[4], line 9 1 # 对模型进行训练 2 # 使用 LBFGS 优化器 3 # 进行 20 步训练 (...) 7 # model.train(dataset, opt="LBFGS", steps=20,device=device); 8 # model.to(device) ----> 9 model.fit(dataset, opt="LBFGS", steps=20, lamb=0.01, lamb_entropy=10.)

File ~\pykan-master0.2.4\pykan-master0.24\kan\MultKAN.py:946, in MultKAN.fit(self, dataset, opt, steps, log, lamb, lamb_l1, lamb_entropy, lamb_coef, lamb_coefdiff, update_grid, grid_update_num, loss_fn, lr, start_grid_update_step, stop_grid_update_step, batch, metrics, save_fig, in_vars, out_vars, beta, save_fig_freq, img_folder, singularity_avoiding, y_th, reg_metric, display_metrics) 943 test_id = np.random.choice(dataset['test_input'].shape[0], batch_size_test, replace=False) 945 if _ % grid_update_freq == 0 and _ < stop_grid_update_step and update_grid and _ >= start_grid_update_step: --> 946 self.update_grid(dataset['train_input'][train_id]) 948 if opt == "LBFGS": 949 optimizer.step(closure)

File ~\pykan-master0.2.4\pykan-master0.24\kan\MultKAN.py:358, in MultKAN.update_grid(self, x) 357 def update_grid(self, x): --> 358 self.update_grid_from_samples(x)

File ~\pykan-master0.2.4\pykan-master0.24\kan\MultKAN.py:355, in MultKAN.update_grid_from_samples(self, x) 353 for l in range(self.depth): 354 self.get_act(x) --> 355 self.act_fun[l].update_grid_from_samples(self.acts[l])

File ~\pykan-master0.2.4\pykan-master0.24\kan\KANLayer.py:241, in KANLayer.update_grid_from_samples(self, x, mode) 238 y_eval = coef2curve(x_pos, self.grid, self.coef, self.k) 240 self.grid.data = extend_grid(grid, k_extend=self.k) --> 241 self.coef.data = curve2coef(x_pos, y_eval, self.grid, self.k)

File ~\pykan-master0.2.4\pykan-master0.24\kan\spline.py:176, in curve2coef(x_eval, y_eval, grid, k, lamb) 174 n1, n2, n = XtX.shape[0], XtX.shape[1], XtX.shape[2] 175 identity = torch.eye(n,n)[None, None, :, :].expand(n1, n2, n, n) --> 176 A = XtX + lamb * identity 177 B = Xty 178 coef = (A.pinverse() @ B)[:,:,:,0]

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! The fact is, I made several attempts at debugging, including the utilization of the author's API_10_device.ipynb; however, regrettably, I was unable to resolve the issue. If you have any valuable suggestions, please do not hesitate to share them.