tomer196 / pilot Goto Github PK

Hello,
i'm trying to run your code but there are many errors for the complex number tensors, e.g.
RuntimeError: "batch_norm_stats_cuda" not implemented for 'ComplexFloat'

I'm curious about how you run the code. Could you share please, how shall I run it, in order to avoid these messages?

hello,
I was wondering if you could help me with the low SNR experiments. I couldn't really find how to change this value.
Saw that in the sampling_model.py you have this piece of code:
if self.SNR:
noise_amp=0.001
noise = noise_amp * torch.randn(sub_ksp.shape)
sub_ksp = sub_ksp + noise.to(sub_ksp.device)

How should I change the values here (or somewhere else) in order to get results as mentioned in the paper in Figure 8 (for 22.20 dB, 14.24dB, 8.22 dB)?

Hi Tomer,

Thanks for sharing your code! I noticed you implemented gridding

May I ask can this function backpropogate gradient?

I have some parameters that controls a non-Cartesian trajectory, and I hope to grid the non-Cartesian trajectory onto Cartesian grid, so I could have a rough density in different regions of k-space and then derive a cost function based on the value in Cartesian grid to optimise the control parameters.

I'm wondering if I could use your _gridding2() function to achieve that?

Yours,
Qijia