OneLLM

OneLLM copied to clipboard

Hi, Any idea or reference of the input fmri format or how to process the data?

Hi @fredfyyang ,we use the pre-processed dataset at: https://huggingface.co/datasets/pscotti/naturalscenesdataset

Each fmri data is converted into a 1-D tensor such as a tensor with shape [1, 1536].

For details, you may refer to: https://github.com/MedARC-AI/fMRI-reconstruction-NSD

Do we need to preprocess the original dataset from https://github.com/MedARC-AI/fMRI-reconstruction-NSD (tar files) to something like test/test_subj01_1/sample000007482.nsdgeneral.npy ourselves?

and for batch inference, should it be [batch_size, n_voxels] or [batch_size, 1, n_voxels]?