ValueError: Using a target size (torch.Size([32, 1])) that is different to the input size (torch.Size([7200, 1])) is deprecated. Please ensure they have the same size

ptrblck · May 18, 2023, 9:17pm

It seems the change wasn’t done properly as the error message is still the same and the target tensor still has a single dimension.

Yes, 46 features sound correct assuming all 46 columns contain feature values.

zakcay · May 11, 2024, 3:54pm

Hi, sorry for the hundredth version of the same question but I see that this is the go to place for it

I’m using a code shared on github to model electronic health records. The data structure is:

[patient id, label, [ visit1, …,visitn]] where each visit is of this form: [time difference between previous visit, list of medical codes]

And this is an example from my training set where medical codes are turned into integers (after preprocessing).

[111, 0, [[[0], [1313]], [[12], [1313]], [[79], [1929]], [[29], [1007, 1930, 1931, 554, 1932, 1779]], [[6], [554]], [[20], [1933, 1934]], [[2], [1933, 1935, 1934, 1936]], [[27], [1267, 414]], [[6], [1935, 1929, 1937]], [[20], [1267]], [[1], [557, 477]]…]]

I get this error:
“…
le “/home/jupyter/.local/lib/python3.10/site-packages/torch/nn/functional.py”, line 3113, in binary_cross_entropy
raise ValueError(
ValueError: Using a target size (torch.Size([128, 1, 1])) that is different to the input size (torch.Size([128])) is deprecated. Please ensure they have the same size.”

I’ve read the previous comments but need a little more guidance or a translation about what this error has to do in the context of my problem as I have lists of lists.

This is the repository I’m using:

github.com

ZhiGroup/pytorch_ehr/blob/master/README.md

# ![image](https://github.com/ZhiGroup/pytorch_ehr/assets/25290490/c2ac41bd-1e13-4243-8664-e9b2cf556bd6) : Predictive Modeling on Electronic Health Records(EHR) using Pytorch 

**Overview**

Although there are plenty of repos on vision and NLP models, there are very limited repos on EHR using deep learning that we can find. Here we open source our repo, implementing data preprocessing, data loading, and a zoo of common RNN models. The main goal is to lower the bar of entering this field for researchers. We are not claiming any state-of-the-art performance, though our models are quite competitive (a paper describing our work will be available soon).  

Based on existing works (e.g., Dr. AI and RETAIN), we represent electronic health records (EHRs) using the pickled list of list of list, which contain histories of patients' diagnoses, medications, and other various events. We integrated all relevant information of a patient's history, allowing easy subsetting.

Currently, this repo includes the following predictive models: Vanilla RNN, GRU, LSTM, Bidirectional RNN, Bidirectional GRU, Bidirectional LSTM, Dilated RNN, Dilated GRU, Dilated LSTM, QRNN,and T-LSTM to analyze and predict clinical performaces. Additionally we have tutorials comparing perfomance to plain LR, Random Forest. 

**Pipeline**

![pipeline](tutorials/Pipeline%20for%20data%20flow.png)


**Primary Results**


![Results Summary](tutorials/image.png)

This file has been truncated. show original

Any guidance is appreciated.

ptrblck · May 11, 2024, 6:25pm

squeeze the target in dim1 and dim2 to remove the additional dimensions with a size of 1 creating a matching shape to the model output.

zakcay · May 12, 2024, 12:37am

Thank you for your quick response.
Originally, the code was like this:
loss = criterion(output, label_tensor)

I changed it to this:
output2=torch.squeeze(output,1)
loss = criterion(output, label_tensor)

which gave the error

[RuntimeError: dimension out of range (expected to be in range of [-1, 0], but got 1)]

Not sure if I used it in the correct place.

ptrblck · May 12, 2024, 1:26am

Use label_tensor = label_tensor.squeeze(1).squeeze(1) and it should work.

zakcay · May 12, 2024, 2:14am

Yes it worked! Thank you so much!