apavlo89
commented
1 month ago I used this guide https://www.kaggle.com/code/seyidcemkarakas/kan-tabular-data-binary-classification for my dataset and got this error:
checkpoint directory created: ./model
saving model version 0.0
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
[<ipython-input-16-bff360054870>](https://localhost:8080/#) in <cell line: 41>()
39
40 # KAN model training
---> 41 results = model.train({'train_input': train_input, 'train_label': train_label, 'test_input': val_input, 'test_label': val_label},
42 metrics=(train_acc, test_acc),
43 opt="LBFGS", steps=100, loss_fn=torch.nn.CrossEntropyLoss())
TypeError: Module.train() got an unexpected keyword argument 'metrics'`
`import torch
from kan import KAN
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
# Converting data to Torch tensor
train_input = torch.tensor(X_train.to_numpy(), dtype=torch.float32)
train_label = torch.tensor(y_train.to_numpy(), dtype=torch.long)
val_input = torch.tensor(X_val.to_numpy(), dtype=torch.float32)
val_label = torch.tensor(y_val.to_numpy(), dtype=torch.long)
test_input = torch.tensor(X_test.to_numpy(), dtype=torch.float32)
test_label = torch.tensor(y_test.to_numpy(), dtype=torch.long)
dataset = {
'train_input': train_input,
'train_label': train_label,
'val_input': val_input,
'val_label': val_label,
'test_input': test_input,
'test_label': test_label
}
# Create model
model = KAN(width=[11, 2], grid=10, k=2)
# Fuctions for getting accuracy scores while training
def train_acc():
preds = torch.argmax(model(dataset['train_input']), dim=1)
return torch.mean((preds == dataset['train_label']).float())
def test_acc():
preds = torch.argmax(model(dataset['test_input']), dim=1)
return torch.mean((preds == dataset['test_label']).float())
# KAN model training
results = model.train({'train_input': train_input, 'train_label': train_label, 'test_input': val_input, 'test_label': val_label},
metrics=(train_acc, test_acc),
opt="LBFGS", steps=100, loss_fn=torch.nn.CrossEntropyLoss())
# Predictions of train val and test datasets
test_preds = torch.argmax(model.forward(test_input).detach(),dim=1)
test_labels = test_label
train_preds = torch.argmax(model.forward(train_input).detach(),dim=1)
train_labels = train_label
val_preds = torch.argmax(model.forward(val_input).detach(),dim=1)
val_labels = val_label
# Evaluate metrics
print("Train ACC:", accuracy_score(train_labels.numpy(), train_preds.numpy()))
print("Val ACC:", accuracy_score(val_labels.numpy(), val_preds.numpy()))
print("Test ACC:", accuracy_score(test_labels.numpy(), test_preds.numpy()))
# Plotting KAN network
model.plot(scale=10)
# Learning curve based on ACC and LOSS
plt.figure(figsize=(10, 5))
plt.plot(results["train_acc"], label='Training Accuracy')
plt.plot(results["test_acc"], label='Val Accuracy')
plt.plot(results["train_loss"], label='Training Loss')
plt.plot(results["test_loss"], label='Val Loss')
plt.title('Training and Val Accuracy over Iterations')
plt.xlabel('Iteration')
plt.ylabel('Accuracy & Loss')
plt.legend()
plt.grid(True)
plt.show()`
seyidcemkarakas
zozi0406
KindXiaoming
Are there any step-by-step guides on how to apply KAN on tabular binary classification datasets??