NeuronBlocks/block_zoo/CRF.py

245 строки
11 KiB
Python

# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT license.
from block_zoo.BaseLayer import BaseLayer, BaseConf
from utils.DocInherit import DocInherit
import torch
import torch.nn as nn
from copy import deepcopy
import torch.autograd as autograd
def argmax(vec):
# return the argmax as a python int
_, idx = torch.max(vec, 1)
return idx.item()
def log_sum_exp(vec, m_size):
"""
calculate log of exp sum
args:
vec (batch_size, vanishing_dim, hidden_dim) : input tensor
m_size : hidden_dim
return:
batch_size, hidden_dim
"""
_, idx = torch.max(vec, 1) # B * 1 * M
max_score = torch.gather(vec, 1, idx.view(-1, 1, m_size)).view(-1, 1, m_size) # B * M
return max_score.view(-1, m_size) + torch.log(torch.sum(torch.exp(vec - max_score.expand_as(vec)), 1)).view(-1, m_size) # B * M
class CRFConf(BaseConf):
"""
Configuration of CRF layer
Args:
"""
def __init__(self, **kwargs):
super(CRFConf, self).__init__(**kwargs)
@DocInherit
def default(self):
self.START_TAG = "<start>"
self.STOP_TAG = "<eos>"
@DocInherit
def declare(self):
self.num_of_inputs = 1
self.input_ranks = [3]
@DocInherit
def inference(self):
self.output_dim = [1]
# add target dict judgement start or end
self.target_dict = deepcopy(self.target_dict.cell_id_map)
if not self.target_dict.get(self.START_TAG):
self.target_dict[self.START_TAG] = len(self.target_dict)
if not self.target_dict.get(self.STOP_TAG):
self.target_dict[self.STOP_TAG] = len(self.target_dict)
super(CRFConf, self).inference()
@DocInherit
def verify(self):
super(CRFConf, self).verify()
class CRF(BaseLayer):
""" Conditional Random Field layer
Args:
layer_conf(CRFConf): configuration of CRF layer
"""
def __init__(self, layer_conf):
super(CRF, self).__init__(layer_conf)
self.target_size = len(self.layer_conf.target_dict)
init_transitions = torch.zeros(self.target_size, self.target_size)
init_transitions[:, self.layer_conf.target_dict[self.layer_conf.START_TAG]] = -10000.0
init_transitions[self.layer_conf.target_dict[self.layer_conf.STOP_TAG], :] = -10000.0
init_transitions[:, 0] = -10000.0
init_transitions[0, :] = -10000.0
if self.layer_conf.use_gpu:
init_transitions = init_transitions.cuda()
self.transitions = nn.Parameter(init_transitions)
def _calculate_forward(self, feats, mask):
"""
input:
feats: (batch, seq_len, self.tag_size)
masks: (batch, seq_len)
"""
batch_size = feats.size(0)
seq_len = feats.size(1)
tag_size = feats.size(2)
mask = mask.transpose(1, 0).contiguous()
ins_num = seq_len * batch_size
# be careful the view shape, it is .view(ins_num, 1, tag_size) but not .view(ins_num, tag_size, 1)
feats = feats.transpose(1, 0).contiguous().view(ins_num, 1, tag_size).expand(ins_num, tag_size, tag_size)
# need to consider start
scores = feats + self.transitions.view(1, tag_size, tag_size).expand(ins_num, tag_size, tag_size)
scores = scores.view(seq_len, batch_size, tag_size, tag_size)
# build iter
seq_iter = enumerate(scores)
_, inivalues = next(seq_iter) # bat_size * from_target_size * to_target_size
# only need start from start_tag
partition = inivalues[:, self.layer_conf.target_dict[self.layer_conf.START_TAG], :].clone().view(batch_size, tag_size, 1) # bat_size * to_target_size
for idx, cur_values in seq_iter:
# previous to_target is current from_target
# partition: previous results log(exp(from_target)), #(batch_size * from_target)
# cur_values: bat_size * from_target * to_target
cur_values = cur_values + partition.contiguous().view(batch_size, tag_size, 1).expand(batch_size, tag_size, tag_size)
cur_partition = log_sum_exp(cur_values, tag_size)
# (bat_size * from_target * to_target) -> (bat_size * to_target)
# partition = utils.switch(partition, cur_partition, mask[idx].view(bat_size, 1).expand(bat_size, self.tagset_size)).view(bat_size, -1)
mask_idx = mask[idx, :].view(batch_size, 1).expand(batch_size, tag_size)
# effective updated partition part, only keep the partition value of mask value = 1
masked_cur_partition = cur_partition.masked_select(mask_idx)
# let mask_idx broadcastable, to disable warning
mask_idx = mask_idx.contiguous().view(batch_size, tag_size, 1)
# replace the partition where the maskvalue=1, other partition value keeps the same
partition.masked_scatter_(mask_idx, masked_cur_partition)
# until the last state, add transition score for all partition (and do log_sum_exp) then select the value in STOP_TAG
cur_values = self.transitions.view(1, tag_size, tag_size).expand(batch_size, tag_size, tag_size) + partition.contiguous().view(batch_size, tag_size, 1).expand(batch_size, tag_size, tag_size)
cur_partition = log_sum_exp(cur_values, tag_size)
final_partition = cur_partition[:, self.layer_conf.target_dict[self.layer_conf.STOP_TAG]]
return final_partition.sum(), scores
def _viterbi_decode(self, feats, mask):
"""
input:
feats: (batch, seq_len, self.tag_size)
mask: (batch, seq_len)
output:
decode_idx: (batch, seq_len) decoded sequence
path_score: (batch, 1) corresponding score for each sequence
"""
batch_size = feats.size(0)
seq_len = feats.size(1)
tag_size = feats.size(2)
# calculate sentence length for each sentence
length_mask = torch.sum(mask.long(), dim=1).view(batch_size, 1).long()
# mask to (seq_len, batch_size)
mask = mask.transpose(1, 0).contiguous()
ins_num = seq_len * batch_size
# be careful the view shape, it is .view(ins_num, 1, tag_size) but not .view(ins_num, tag_size, 1)
feats = feats.transpose(1, 0).contiguous().view(ins_num, 1, tag_size).expand(ins_num, tag_size, tag_size)
# need to consider start
scores = feats + self.transitions.view(1, tag_size, tag_size).expand(ins_num, tag_size, tag_size)
scores = scores.view(seq_len, batch_size, tag_size, tag_size)
# build iter
seq_iter = enumerate(scores)
# record the position of best score
back_points = list()
partition_history = list()
# reverse mask (bug for mask = 1- mask, use this as alternative choice)
mask = (1 - mask.long()).byte()
_, inivalues = next(seq_iter) # bat_size * from_target_size * to_target_size
# only need start from start_tag
partition = inivalues[:, self.layer_conf.target_dict[self.layer_conf.START_TAG], :].clone().view(batch_size, tag_size) # bat_size * to_target_size
# print "init part:",partition.size()
partition_history.append(partition)
# iter over last scores
for idx, cur_values in seq_iter:
# previous to_target is current from_target
# partition: previous results log(exp(from_target)), #(batch_size * from_target)
# cur_values: batch_size * from_target * to_target
cur_values = cur_values + partition.contiguous().view(batch_size, tag_size, 1).expand(batch_size, tag_size, tag_size)
partition, cur_bp = torch.max(cur_values, 1)
partition_history.append(partition)
# cur_bp: (batch_size, tag_size) max source score position in current tag
# set padded label as 0, which will be filtered in post processing
cur_bp.masked_fill_(mask[idx].view(batch_size, 1).expand(batch_size, tag_size), 0)
back_points.append(cur_bp)
# add score to final STOP_TAG
partition_history = torch.cat(partition_history, 0).view(seq_len, batch_size, -1).transpose(1, 0).contiguous() # (batch_size, seq_len. tag_size)
# get the last position for each setences, and select the last partitions using gather()
last_position = length_mask.view(batch_size, 1, 1).expand(batch_size, 1, tag_size) - 1
last_partition = torch.gather(partition_history, 1, last_position).view(batch_size,tag_size,1)
# calculate the score from last partition to end state (and then select the STOP_TAG from it)
last_values = last_partition.expand(batch_size, tag_size, tag_size) + self.transitions.view(1, tag_size, tag_size).expand(batch_size, tag_size, tag_size)
_, last_bp = torch.max(last_values, 1)
pad_zero = autograd.Variable(torch.zeros(batch_size, tag_size)).long()
if self.layer_conf.use_gpu:
pad_zero = pad_zero.cuda()
back_points.append(pad_zero)
back_points = torch.cat(back_points).view(seq_len, batch_size, tag_size)
# select end ids in STOP_TAG
pointer = last_bp[:, self.layer_conf.target_dict[self.layer_conf.STOP_TAG]]
insert_last = pointer.contiguous().view(batch_size, 1, 1).expand(batch_size, 1, tag_size)
back_points = back_points.transpose(1, 0).contiguous()
# move the end ids(expand to tag_size) to the corresponding position of back_points to replace the 0 values
back_points.scatter_(1, last_position, insert_last)
back_points = back_points.transpose(1, 0).contiguous()
# decode from the end, padded position ids are 0, which will be filtered if following evaluation
decode_idx = autograd.Variable(torch.LongTensor(seq_len, batch_size))
if self.layer_conf.use_gpu:
decode_idx = decode_idx.cuda()
decode_idx[-1] = pointer.detach()
for idx in range(len(back_points)-2, -1, -1):
pointer = torch.gather(back_points[idx], 1, pointer.contiguous().view(batch_size, 1))
decode_idx[idx] = pointer.detach().view(batch_size)
path_score = None
decode_idx = decode_idx.transpose(1, 0)
return path_score, decode_idx
def forward(self, string, string_len):
"""
CRF layer process: include use transition matrix compute score and viterbi decode
Args:
string(Tensor): [batch_size, seq_len, target_num]
string_len(Tensor): [batch_size]
Returns:
score: the score by CRF inference
best_path: the best bath of viterbi decode
"""
assert string_len is not None, "CRF layer need string length for mask."
masks = []
string_len_val = string_len.cpu().data.numpy()
for i in range(len(string_len)):
masks.append(
torch.cat([torch.ones(string_len_val[i]), torch.zeros(string.shape[1] - string_len_val[i])]))
masks = torch.stack(masks).view(string.shape[0], string.shape[1]).byte()
if self.layer_conf.use_gpu:
masks = masks.cuda()
forward_score, scores = self._calculate_forward(string, masks)
_, tag_seq = self._viterbi_decode(string, masks)
return (forward_score, scores, masks, tag_seq, self.transitions, self.layer_conf), string_len