#!/usr/bin/env python3
# -*- coding: utf-8 -*-
#
# Copyright (C) IBM Corporation 2018
#
# 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
#
# http://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.
"""memory.py: Class for editing memory """
__author__ = "Younes Bouhadjar"
import torch
from miprometheus.models.dwm.tensor_utils import sim, outer_prod
[docs]class Memory:
"""
Implementation of the memory of the DWM.
"""
[docs] def __init__(self, mem_t):
"""
Initializes the memory.
:param mem_t: the memory at time t [batch_size, memory_content_size, memory_addresses_size]
"""
self._memory = mem_t
[docs] def attention_read(self, wt):
"""
Returns the data read from memory.
:param wt: head's weights [batch_size, num_heads, memory_addresses_size]
:return: the read data [batch_size, num_heads, memory_content_size]
"""
return sim(wt, self._memory)
[docs] def add_weighted(self, add, wt):
"""
Writes data to memory.
:param wt: head's weights [batch_size, num_heads, memory_addresses_size]
:param add: the data to be added to memory [batch_size, num_heads, memory_content_size]
:return the updated memory [batch_size, memory_addresses_size, memory_content_size]
"""
# memory = memory + sum_{head h} weighted add(h)
self._memory = self._memory + torch.sum(outer_prod(add, wt), dim=-3)
[docs] def erase_weighted(self, erase, wt):
"""
Erases elements from memory.
:param wt: head's weights [batch_size, num_heads, memory_addresses_size]
:param erase: data to be erased from memory [batch_size, num_heads, memory_content_size]
:return the updated memory [batch_size, memory_addresses_size, memory_content_size]
"""
# memory = memory * product_{head h} (1 - weighted erase(h))
self._memory = self._memory * \
torch.prod(1 - outer_prod(erase, wt), dim=-3)
[docs] def content_similarity(self, k):
"""
Calculates the dot product for Content aware addressing.
:param k: the keys emitted by the controller [batch_size, num_heads, memory_content_size]
:return: the dot product between the keys and query [batch_size, num_heads, memory_addresses_size]
"""
return sim(k, self._memory, l2_normalize=True, aligned=False)
@property
def size(self):
"""
Returns the size of the memory.
:return: Int size of the memory
"""
return self._memory.size()
@property
def content(self):
"""
Returns the entire memory.
:return: the memory []
"""
return self._memory