search

sidhomj / DeepTCR

Deep Learning Methods for Parsing T-Cell Receptor Sequencing (TCRSeq) Data
https://sidhomj.github.io/DeepTCR/
MIT License
113 stars 40 forks source link

Error when changing max_length #34

Closed emm1R closed 3 years ago

emm1R commented 3 years ago

Hi, In Train_VAE(), a ValueError is raised if a different max_length has been given to DeepTRC_U. The tensor shapes do not match on row 2292 when tf.equal() is called.

Näyttökuva 2020-12-5 kello 2 48 19

sidhomj commented 3 years ago

Just to confirm, you're trying to use a max length of 70 for your analysis?

sidhomj commented 3 years ago

Updated the code to fix this issue. should be able now to use any max_length argument with the VAE. Let me know if this solves the problem! Thanks for using DeepTCR!

emm1R commented 3 years ago

Yes, it works now. Thank you! And yes, I am using a max_length of 70. There were a very small percentage of TCRs that were longer than 40 and I didn’t want them to be filtered out. Will this impact the training or clustering in any negative way?

sidhomj commented 3 years ago

It shouldn't! Just wanted to make sure I was solving for the right use-case. It's actually a feature I was meaning to write but I had not gotten around to handing other input lengths for the VAE. The max_length argument was put in initially for the supervised models but I had to write a bit of code of modify VAE for other length inputs (other than 40). Thank you for highlighting this!