2026a8939d9ccc3e26ac776db5b4788846fd166c,dnc/sparse_memory.py,SparseMemory,read_from_sparse_memory,#SparseMemory#Any#Any#Any#Any#,181

Before Change


    // TODO: for now we assume infinite memory, hence least used is the first free cell
    read_vectors = []
    pos = read_positions.squeeze().data.cpu().numpy()
    for b in range(read_positions.size(0)):
      read_vectors.append(T.stack([memory[b, p] for p in list(pos[b])], 0))
    read_vectors = T.stack(read_vectors, 0)

    return read_vectors, read_positions, read_weights

After Change


    read_positions = var(read_positions)
    read_positions = T.cat([read_positions, last_used_mem.unsqueeze(1)], 2)

    (b, m, w) = memory.size()
    read_vectors = memory.gather(1, read_positions.squeeze().unsqueeze(2).expand(b, self.K+1, w))

    return read_vectors, read_positions, read_weights
Italian Trulli
In pattern: SUPERPATTERN

Frequency: 3

Non-data size: 7

Instances


Project Name: ixaxaar/pytorch-dnc
Commit Name: 2026a8939d9ccc3e26ac776db5b4788846fd166c
Time: 2017-12-03
Author: root@ixaxaar.in
File Name: dnc/sparse_memory.py
Class Name: SparseMemory
Method Name: read_from_sparse_memory


Project Name: onnx/onnx-tensorflow
Commit Name: a39084b995d9d6d88300632c54895488985e5cfd
Time: 2017-10-24
Author: gbercea@gheorghodorsmbp.watson.ibm.com
File Name: onnx_tf/backend.py
Class Name: TensorflowBackend
Method Name: handle_flatten


Project Name: maciejkula/spotlight
Commit Name: bc51dbc0c56f68ed30857755026633f78eef1ae8
Time: 2017-08-20
Author: maciej.kula@gmail.com
File Name: spotlight/layers.py
Class Name: BloomEmbedding
Method Name: forward