张大千 zhpmatrix

🎯

Focusing

NLP/Question Answering/Knowledge Graph/Large Language Model

738 followers · 18 following

View GitHub Profile

Recently created

Least recently created

Recently updated

Least recently updated

zhpmatrix / check_convex.py

Created December 23, 2019 03:07 — forked from mblondel/check_convex.py

A small script to get numerical evidence that a function is convex

	# Authors: Mathieu Blondel, Vlad Niculae
	# License: BSD 3 clause

	import numpy as np


	def _gen_pairs(gen, max_iter, max_inner, random_state, verbose):
	rng = np.random.RandomState(random_state)

	# if tuple, interpret as randn

zhpmatrix / top_k_viterbi.py

Last active November 22, 2019 12:24 — forked from PetrochukM/top_k_viterbi.py

Implemented a Top K Viterbi Decoder algorithm in PyTorch. Useful for Conditional Random Fields (CRFs)-based probabilistic graphical modelling. Learn more here: https://nlp.stanford.edu/joberant/esslli_2016/kbest-ict.pdf

	import torch

	# Credits to AllenNLP for the base implementation and base tests:
	# https://github.com/allenai/allennlp/blob/master/allennlp/nn/util.py#L174

	# Modified AllenNLP `viterbi_decode` to support `top_k` sequences efficiently.
	def viterbi_decode(tag_sequence: torch.Tensor, transition_matrix: torch.Tensor, top_k: int=5):
	"""
	Perform Viterbi decoding in log space over a sequence given a transition matrix
	specifying pairwise (transition) potentials between tags and a matrix of shape

zhpmatrix / weighted_loss.py

Created December 17, 2018 03:12 — forked from zedom1/weighted_loss.py

	import tensorflow as tf
	from numpy import *
	from random import randint

	max_length = 2
	batch_size = 3

	label = array([[[0,0] for _ in range(max_length)] for _ in range(batch_size)])
	for i in range(batch_size):
	for j in range(max_length):

zhpmatrix / sequence_mask.py

Created December 17, 2018 03:12 — forked from zedom1/sequence_mask.py

	import tensorflow as tf
	from numpy import *
	from random import randint

	max_length = 3
	batch_size = 5

	targets = array([[1 for _ in range(max_length)] for _ in range(batch_size)])
	logits = array([[[randint(0,10)/10,randint(0,10)/10] for _ in range(max_length)] for _ in range(batch_size)])
	sequence_length = array([randint(1,max_length) for _ in range(batch_size)])

zhpmatrix / nms_fast.m

Created May 22, 2018 08:42 — forked from quantombone/nms_fast.m

blazing fast nms (non-maximum suppression)

	function top = nms(boxes, overlap)
	% top = nms_fast(boxes, overlap)
	% Non-maximum suppression. (FAST VERSION)
	% Greedily select high-scoring detections and skip detections
	% that are significantly covered by a previously selected
	% detection.
	% NOTE: This is adapted from Pedro Felzenszwalb's version (nms.m),
	% but an inner loop has been eliminated to significantly speed it
	% up in the case of a large number of boxes
	% Tomasz Malisiewicz ([email protected])

zhpmatrix / keras_logistic_regression.py

Created May 3, 2017 13:48 — forked from fchollet/keras_logistic_regression.py

	from keras.models import Sequential
	from keras.layers import Dense

	x, y = ...
	x_val, y_val = ...

	# 1-dimensional MSE linear regression in Keras
	model = Sequential()
	model.add(Dense(1, input_dim=x.shape[1]))
	model.compile(optimizer='rmsprop', loss='mse')