Question about few-shot setting without non-binding peptide

[dxl3742]

Hi,
Great work on the package!
I am wondering will the few-shot setting work without defining a 0 non-binding peptide? I only defined the binding and the unknown peptides, and I have got this error.
RuntimeError: The expanded size of the tensor (5) must match the existing size (0) at non-singleton dimension 2. Target sizes: [0, 40, 5]. Tensor sizes: [0].
In most cases it would be difficult to obtain knowledge of non-binding peptides, as most databases only have binding peptides and corresponding TCRs.

[GaoYiChengTJ]

Thanks for your interest. It's a good question. Our model was built on the existing databases, i.e. having binding peptides and their corresponding TCRs. From my perspective, the problem involving binding peptides and non-binding peptides needs to develop a new model to solve it.

[yanpinlu]

Hi, Great work on the package! I am wondering will the few-shot setting work without defining a 0 non-binding peptide? I only defined the binding and the unknown peptides, and I have got this error. RuntimeError: The expanded size of the tensor (5) must match the existing size (0) at non-singleton dimension 2. Target sizes: [0, 40, 5]. Tensor sizes: [0]. In most cases it would be difficult to obtain knowledge of non-binding peptides, as most databases only have binding peptides and corresponding TCRs.

I also encountered RuntimeError when I used my own data set to make predictions: The expanded size of the tensor (5) must match the existing size (0) at non-singleton dimension 2. Target sizes: [0, 40, 5]. Tensor sizes: [0], what causes this? The data set provided has three states, 0,1 and unknown. How can I solve this problem @GaoYiChengTJ @dxl3742