However, no error is thrown when multiplying the batch of sparse matrices by a dense tensor of shape (B, N, D), with D>1. Code to reproduce the first case: import torch n_nodes = 10 n_edges = 20 batch_size = 2 sparse_list = [] for i in range(batch_size): indices = torch...
python3.11/site-packages/torch/_functorch/vmap.py", line 480, in _flat_vmap batched_outputs = func(*batched_inputs, **kwargs) File "/home/cwtan/anaconda3/envs/florch/lib/python3.11/site-packages/torch/_functorch/eager_transforms.py", line 1301, in push_jvp output = _jvp_with_arg...