Dilation greater than zero

[Yurnerosk]

I have a time series dataframe containing 80 features, 29922 rows, and the idea was to first use MINIROCKET for feature creation, and then use a linear regressor to reconstruct a target column. I'm doing:

import pandas as pd
import torch
from tsai.models.MINIROCKET_Pytorch import *
from sklearn.linear_model import LinearRegression  # You can choose another regressor
from sklearn.metrics import mean_squared_error
from sklearn.model_selection import train_test_split

...


X = df.drop(columns=[target_column]).values  # Feature matrix (80 columns - target)
y = df[target_column].values  # Target vector

# Step 2: Split the data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Step 3: Initialize and fit MiniRocket
num_samples, num_features = X_train.shape
time_steps = 1  # Assuming you have 1 time step

# Reshape to 3D format
X_train_reshaped = X_train.reshape(num_samples, time_steps, num_features)
X_test_reshaped = X_test.reshape(X_test.shape[0], time_steps, num_features)

To finally:

# Initialize MiniRocket
mrf = MiniRocketFeatures(num_features, time_steps).to(torch.device('cpu'))
mrf.fit(X_train_reshaped)

However, it seems that the dilation cannot be greater than 1.

---> [74](file:///C:/Users.../.venv/Lib/site-packages/tsai/models/MINIROCKET_Pytorch.py:74) C = F.conv1d(x, self.kernels, padding=padding, dilation=dilation, groups=self.c_in)
     [75](file:///C:/Users.../.venv/Lib/site-packages/tsai/models/MINIROCKET_Pytorch.py:75) if self.c_in > 1: # multivariate
     [76](file:///C:/Users.../.venv/Lib/site-packages/tsai/models/MINIROCKET_Pytorch.py:76)     C = C.reshape(x.shape[0], self.c_in, self.num_kernels, -1)

RuntimeError: dilation should be greater than zero

What should I do?

Python 3.12.2
Windows 10
tsai 0.3.9
torch 2.2.2

[Yurnerosk]

I think I wasn't too specific in my question, but what i'm trying to do is:

print(X_train_reshaped.shape)
print(y_train.shape)
# (23937, 1, 79) num samples, timesteps, num_features
# (23937,)

# Initialize MiniRocket
mrf = MiniRocketFeatures(c_in=num_features, seq_len=time_steps).to(torch.device('cpu'))
mrf.dilations
# array([-2147483648,           0])

mrf.fit(X_train_reshaped)

File c:\Users\wilso\OneDrive\Documentos\GitHub\Colhedora\Dados_season_2024\.venv\Lib\site-packages\torch\nn\modules\module.py:1520, in Module._call_impl(self, *args, **kwargs)
   [1515](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/torch/nn/modules/module.py:1515) # If we don't have any hooks, we want to skip the rest of the logic in
   [1516](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/torch/nn/modules/module.py:1516) # this function, and just call forward.
   [1517](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/torch/nn/modules/module.py:1517) if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   [1518](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/torch/nn/modules/module.py:1518)         or _global_backward_pre_hooks or _global_backward_hooks
   [1519](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/torch/nn/modules/module.py:1519)         or _global_forward_hooks or _global_forward_pre_hooks):
-> [1520](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/torch/nn/modules/module.py:1520)     return forward_call(*args, **kwargs)
   [1522](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/torch/nn/modules/module.py:1522) try:
...
---> [74](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/tsai/models/MINIROCKET_Pytorch.py:74) C = F.conv1d(x, self.kernels, padding=padding, dilation=dilation, groups=self.c_in)
     [75](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/tsai/models/MINIROCKET_Pytorch.py:75) if self.c_in > 1: # multivariate
     [76](file:///C:/Users/wilso/OneDrive/Documentos/GitHub/Colhedora/Dados_season_2024/.venv/Lib/site-packages/tsai/models/MINIROCKET_Pytorch.py:76)     C = C.reshape(x.shape[0], self.c_in, self.num_kernels, -1)

RuntimeError: dilation should be greater than zero

I noted the dilations seems to be a bit off, what can I do?

[Yurnerosk]

RuntimeError: Given groups=9, weight of size [756, 1, 9], expected input[168, 1, 9] to have 9 channels, but got 1 channels instead.

I'm getting crazy of how Rockt is not being able to recognize the channels, even when I did my array3d [sample_size x n_features x 1].