A pytorch implementation of torch_gather_nd with multiple batch dim and multiple channel dim support.

LICENSE

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at 
 
https://www.apache.org/licenses/LICENSE-2.0
 
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and limitations under the License.

test_torch_gather_nd.py

import unittest

class TestTorchGatherND(unittest.TestCase):
    def test_torch_gather_nd(self) -> None:
        # gather 1 dim
        params = torch.arange(10)
        indices = torch.tensor([[0], [2], [3]])
        out = torch_gather_nd(params, indices, batch_dim=0)
        self.assertEqual(out.tolist(), [0, 2, 3])

        # gather 2 dims
        params = torch.tensor([[1,2], [3, 4]])
        indices = torch.tensor([[1, 1], [0, 0], [0, 1]])
        out = torch_gather_nd(params, indices, batch_dim=0)
        self.assertEqual(out.tolist(), [4, 1, 2])

        # gather 1 dim, 1 ch dim
        params = torch.tensor([[1,2], [2,3], [3, 4]])
        indices = torch.tensor([[2], [0]])
        out = torch_gather_nd(params, indices, batch_dim=0)
        self.assertEqual(out.tolist(), [[3, 4], [1, 2]])

        # gather 1 batch dim, 2 dims, 1 ch dim
        params = torch.tensor([
            [[1,2], [2,3], [3, 4]],
            [[5,6], [7,8], [9, 10]],
        ])
        indices = torch.tensor([[[2], [0]]]*2)
        out = torch_gather_nd(params, indices, batch_dim=1)
        self.assertEqual(out.tolist(), [
            [[3, 4], [1,2]],
            [[9, 10], [5, 6]],
        ])

        # gather 2 batch dims, 2 dims, 1 ch dims
        params = torch.rand(1, 2, 10, 6, 3)
        indices = torch.zeros(1, 2, 5, 2, dtype=torch.long)
        out = torch_gather_nd(params, indices, batch_dim=2)
        self.assertEqual(out.shape, (1, 2, 5, 3))

        # gather 2 dims, 2 ch dims
        params = torch.rand(10, 6, 3, 4)
        indices = torch.zeros(5, 2, dtype=torch.long)
        out = torch_gather_nd(params, indices, batch_dim=0)
        self.assertEqual(out.shape, (5, 3, 4))

torch_gather_nd.py

def torch_gather_nd(params: torch.Tensor, indices: torch.Tensor, batch_dim: int = 0) -> torch.Tensor:
    """
    torch_gather_nd implements tf.gather_nd in PyTorch.
    
    This supports multiple batch dimensions as well as multiple channel dimensions.
    """
    index_shape = indices.shape[:-1]
    num_dim = indices.size(-1)
    tail_sizes = params.shape[batch_dim+num_dim:]

    # flatten extra dimensions
    for s in tail_sizes:
        row_indices = torch.arange(s, device=params.device)
        indices = indices.unsqueeze(-2)
        indices = indices.repeat(*[1 for _ in range(indices.dim()-2)], s, 1)
        row_indices = row_indices.expand(*indices.shape[:-2], -1).unsqueeze(-1)
        indices = torch.cat((indices, row_indices), dim=-1)
        num_dim += 1

    # flatten indices and params to batch specific ones instead of channel specific
    for i in range(num_dim):
        size = prod(params.shape[batch_dim+i+1:batch_dim+num_dim])
        indices[..., i] *= size

    indices = indices.sum(dim=-1)
    params = params.flatten(batch_dim, -1)
    indices = indices.flatten(batch_dim, -1)

    out = torch.gather(params, dim=batch_dim, index=indices)
    return out.reshape(*index_shape,*tail_sizes)

What is prod() in 'size = prod(params.shape[batch_dim+i+1:batch_dim+num_dim])' ?