dengy02
/
PointManifold_dy

 
			
							#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
@Author: Yue Wang
@Contact: yuewangx@mit.edu
@File: util
@Time: 4/5/19 3:47 PM

Modified by 
@Author: Yu Deng
@Contact: dengy02@pcl.ac.cn
@Time: 2022/7/6 16:30 PM
"""

import os
import sys
import errno
import os.path as osp

import numpy as np
import tensorflow as tf


def cal_loss(labels=None, logits=None, smoothing=True):
    ''' Calculate cross entropy loss, apply label smoothing if needed. '''
    labels = tf.reshape(labels, (-1,))
    # print("labels :", labels.shape)

    if smoothing:
        eps = 0.2
        n_class = logits.shape[1]
        # one_hot = torch.zeros_like(logits).scatter(1, labels.view(-1, 1), 1)        
        one_hot = tf.one_hot(labels, n_class) # dtype=tf.float32
        # print("one_hot", one_hot)
        # print("one_hot shape", one_hot.shape)

        one_hot = one_hot * (1 - eps) + (1 - one_hot) * eps / (n_class - 1)
        
        # log_prb = F.log_softmax(logits, dim=1)
        logits = tf.convert_to_tensor(logits, dtype=tf.float32)
        # logits = tf.convert_to_tensor(logits)
        log_prb = tf.nn.log_softmax(logits=logits, axis=1)
        
        # loss = -(one_hot * log_prb).sum(dim=1).mean()
        mul_hot_log = tf.math.multiply(one_hot, log_prb)
        mul_hot_log_sum = tf.reduce_sum(input_tensor=mul_hot_log, axis=1, keepdims=False, name=None)
        mul_hot_log_mean = tf.reduce_mean(mul_hot_log_sum)
        loss = -mul_hot_log_mean
    else:
        labels = tf.convert_to_tensor(labels, dtype=tf.int32)
        # labels = tf.convert_to_tensor(labels)
        # reduction=tf.keras.losses.Reduction.SUM ，default: reduction=losses_utils.ReductionV2.AUTO
        cce = tf.keras.losses.SparseCategoricalCrossentropy()
        loss = cce(labels, logits)

    return loss


class IOStream():
    def __init__(self, path):
        self.f = open(path, 'a')

    def cprint(self, text):
        print(text)
        self.f.write(text+'\n')
        self.f.flush()

    def close(self):
        self.f.close()


def mkdir_if_missing(directory):
    if not osp.exists(directory):
        try:
            os.makedirs(directory)
        except OSError as e:
            if e.errno != errno.EEXIST:
                raise


class Logger(object):
    """
    Write console output to external text file.
    Code imported from https://github.com/Cysu/open-reid/blob/master/reid/utils/logging.py.
    """
    def __init__(self, fpath=None):
        self.console = sys.stdout
        self.file = None
        if fpath is not None:
            mkdir_if_missing(os.path.dirname(fpath))
            self.file = open(fpath, 'w')

    def __del__(self):
        self.close()

    def __enter__(self):
        pass

    def __exit__(self, *args):
        self.close()

    def write(self, msg):
        self.console.write(msg)
        if self.file is not None:
            self.file.write(msg)

    def flush(self):
        self.console.flush()
        if self.file is not None:
            self.file.flush()
            os.fsync(self.file.fileno())

    def close(self):
        self.console.close()
        if self.file is not None:
            self.file.close()

if __name__ == "__main__":
    # [32, 40]， [32]
    import numpy as np
    import sklearn.metrics as metrics
    import datetime

    logits = np.array([[ 0.4826,  1.8397,  0.1550, -0.2594,  0.3050,  0.8015, -0.2805,  0.0954],
        [ 1.3998, -1.1612,  1.7943,  0.2963,  0.0122, -1.7097,  0.0919,  1.1606],
        [-0.9231, -0.2899,  1.2428,  0.9531, -0.3307,  1.8807,  1.2747,  2.5915],
        [-0.3514,  0.3604, -0.3423, -1.0050, -0.3290,  0.0732,  1.8311,  0.8099]], dtype=np.float32)
    labels = np.array([7, 4, 6, 1], dtype=np.int32)
    logits = tf.convert_to_tensor(logits)
    labels = tf.convert_to_tensor(labels)
    
    print()
    print(logits.shape)
    print(labels.shape)
    print(logits[0])
    print()

    ## one_hot like:
    # [[0., 1., 0., 0., 0., 0., 0., 0.],
    # [0., 0., 0., 1., 0., 0., 0., 0.],
    # [0., 1., 0., 0., 0., 0., 0., 0.],
    # [0., 0., 0., 0., 1., 0., 0., 0.]]  
    loss = cal_loss(logits=logits, labels=labels, smoothing=True)
    print("loss smoothing", loss)

    loss = cal_loss(logits=logits, labels=labels, smoothing=False)
    print("loss", loss)
    print("============")

    global_batch_size = 8
    strategy = tf.distribute.MirroredStrategy(["GPU:0", "GPU:1"])
    # dataset = tf.data.Dataset.from_tensor_slices([1, 2, 3, 4, 5, 6, 7, 8]).repeat().batch(global_batch_size)
    dataset = tf.data.Dataset.from_tensor_slices([1, 2, 3, 4, 5, 6, 7, 8]).repeat(8).batch(global_batch_size)
    distributed_iterator = iter(strategy.experimental_distribute_dataset(dataset))
    print(distributed_iterator)
    print("iter out \n:", next(distributed_iterator))

    # x = [np.float32(0.83), np.float32(1.12), np.float32(1.10), np.float32(1.0), np.float32(1.53)]
    # l = [[29, 21, 15, 2, 2, 10, 34, 10], 
    #     [30, 16, 1, 8, 4, 30, 30, 30],
    #     [4, 0, 26, 24, 30, 30, 39, 33],
    #     [21, 8, 0, 30, 15, 36, 0, 0],
    #     [10, 0, 9, 30, 28, 36, 15, 8]]
    # labels = np.array(l)
    # preds = np.array(x)
    # preds = np.stack(preds)
    # print(labels.shape)
    # print(preds.shape)
    # print()
    # accuracy_score = metrics.accuracy_score(labels, preds)
    # balanced_accuracy_score = metrics.balanced_accuracy_score(labels, preds)
    # print(accuracy_score)
    # print(balanced_accuracy_score)

    logits_np = np.array([[ 2.0835e+00,  2.8684e-01, -3.8064e-01, -6.6219e-03, -4.9030e-02,
         -7.0756e-01, -1.6587e-01,  3.7841e-02,  8.4555e-01, -2.8922e-01,
         -7.8332e-01,  1.3700e+00,  1.6362e-01, -5.7401e-01,  1.6290e+00,
          4.1743e-01, -5.2905e-01,  2.2805e-01, -2.6672e+00, -1.6487e-01,
          2.8328e-01, -3.1602e-01,  1.8948e+00,  8.3975e-01,  3.1507e-01,
          1.8589e+00, -5.6998e-01, -9.6480e-01, -5.6747e-01, -3.3958e-01,
         -1.6440e+00, -3.5754e-01, -7.9882e-01,  2.3381e-01,  6.8907e-02,
         -4.4294e-01,  3.3727e-02, -3.4425e-01, -4.4010e-01, -5.5271e-01],
        [-4.0442e-01, -3.7830e-01,  7.4986e-01, -1.9608e-01, -5.2080e-01,
         -2.9841e-01, -1.6777e+00, -1.8840e-01, -8.2758e-01, -5.0043e-01,
          6.0243e-01, -4.9775e-01, -3.9553e-01,  2.3184e-01,  1.5759e+00,
         -1.3985e+00,  4.8364e-01,  7.5944e-02, -9.1163e-01, -1.0120e+00,
         -1.0840e+00, -1.1333e+00,  7.3563e-01,  7.2417e-01, -3.8258e-01,
          2.1450e+00, -2.9394e-01,  1.6343e+00,  1.5634e-01, -7.5331e-01,
          1.8599e+00,  2.9857e-01,  2.1150e-01,  4.3066e-01, -6.0429e-01,
          9.0129e-01,  2.6864e-01,  2.7161e+00, -1.0155e+00, -8.1855e-01],
        [-2.7429e+00, -2.9181e+00,  3.0671e+00,  1.3629e+00, -1.5668e+00,
         -8.3793e-01, -3.7353e+00,  4.0808e-01, -1.5371e+00, -1.9909e+00,
          3.9708e-01, -1.5419e+00, -1.9746e+00, -1.3238e-01, -1.5096e+00,
          4.9266e-01, -1.7744e-01,  1.1247e-01,  6.9266e-01, -8.2123e-01,
         -3.9770e+00, -2.1140e+00,  9.9297e-01, -1.1321e+00, -8.7195e-01,
          5.5473e+00,  1.0612e+00,  2.4748e+00,  3.0698e-01, -2.0269e+00,
          1.7185e+00,  5.8352e+00,  3.6988e+00, -9.0987e-01,  8.1961e-01,
         -5.3954e-01, -1.4290e+00,  8.5451e+00, -2.6355e+00, -1.2917e+00],
        [ 1.1162e+01,  6.6738e-02, -1.8976e+00, -4.3443e+00, -1.9987e+00,
         -1.7699e+00,  4.5539e+00,  5.1470e+00,  3.7383e+00, -8.5479e-02,
         -3.3659e+00,  6.0107e+00,  2.9951e-02, -1.3219e+00,  1.0165e+00,
          1.7946e+00, -1.4292e+00, -1.0467e+00, -3.3188e+00, -2.4000e+00,
          4.3941e+00,  9.8264e-02, -1.6303e+00, -2.3844e+00, -6.7374e-01,
         -4.3825e+00,  1.5664e+00, -9.1431e-01, -2.6670e+00, -1.7077e+00,
          1.3887e+00, -2.2931e+00, -2.9852e+00, -3.0248e+00, -2.5534e+00,
          1.0055e+00, -1.4695e+00, -2.5426e+00, -9.7849e-01, -1.9550e+00],
        [ 3.5921e+00,  8.4306e-01, -1.3569e+00, -2.9547e+00,  4.2032e-01,
         -1.3029e+00,  1.9426e+00,  1.3834e+00,  1.7537e+00,  9.6460e-03,
         -1.5346e+00,  2.0150e+00,  2.5273e-01, -5.5676e-01,  1.9191e+00,
         -1.0179e-01,  2.7503e-01, -1.4862e+00, -4.0593e+00, -1.3711e+00,
          1.6630e+00,  3.8593e-01,  1.3875e+00,  3.4905e-01,  5.8096e-01,
         -1.4988e+00,  1.0648e+00, -1.1916e+00, -4.3314e-01, -1.2875e+00,
          8.7663e-01, -7.4405e-01, -1.5142e+00, -2.8483e-02, -2.9628e-01,
          2.5531e+00,  6.4484e-01, -2.0996e+00, -5.0988e-01, -8.8208e-01],
        [-9.0217e-01, -2.3921e-01,  7.0113e-01, -6.9945e-01, -7.7149e-01,
         -1.5625e+00, -1.4836e+00,  2.3214e-01,  2.5339e-01, -1.5055e+00,
          1.1614e+00, -9.1424e-01, -6.2381e-01,  4.1322e-02, -2.8845e-01,
         -1.1310e-01,  2.6380e+00, -9.1250e-02,  1.5196e+00, -8.0327e-01,
         -7.2858e-01, -2.9471e+00,  1.9594e+00,  1.6616e+00, -2.4765e-01,
          3.1741e+00, -6.2680e-01,  4.8777e-01,  2.8064e-01, -1.4813e+00,
          1.0913e+00,  7.4820e-02,  2.6674e-01,  4.9987e-01, -1.2711e+00,
          2.7519e+00,  2.3401e+00,  3.1536e+00, -1.6224e+00, -8.7516e-01],
        [-1.6432e+00, -6.1273e-01,  7.9820e-01,  1.0426e+00, -1.2958e+00,
         -4.0188e-01, -1.3746e+00, -8.7540e-01, -7.6528e-01, -1.2126e+00,
          1.3798e+00, -3.9156e-02, -1.2898e+00,  6.4550e-01,  6.8631e-01,
         -1.5323e+00,  5.7061e-01,  9.7129e-01, -2.0090e+00, -8.3243e-01,
         -1.9216e+00, -9.5328e-01,  1.8287e+00, -7.1166e-01, -7.5810e-01,
          2.4084e+00, -4.9811e-02,  4.6670e-01, -1.6608e-01, -2.6434e-01,
          2.1043e+00,  2.0021e+00,  7.8653e-01,  8.5328e-01, -1.0869e-01,
         -7.9721e-01,  2.4434e+00,  6.0053e+00, -1.5854e+00, -4.2325e-01],
        [ 1.3624e+00,  1.2876e+00, -1.1838e+00, -1.1431e+00,  1.5006e-01,
          4.2363e-01,  2.3242e+00,  1.2580e+00,  1.9115e+00,  6.0025e-01,
         -1.3256e+00,  3.6321e+00, -2.3834e-01, -1.7159e+00,  7.0819e-01,
          8.8843e-01, -4.8113e-02, -2.3913e+00, -9.3553e-01, -1.6055e+00,
          2.3139e+00, -5.0882e-01,  5.5813e-01, -6.4976e-01,  5.2835e-01,
         -2.4074e+00,  1.4382e+00, -1.7691e+00, -1.6073e+00, -7.5976e-01,
          4.4626e-01, -2.1728e+00, -1.8500e+00, -3.3810e-01,  1.0280e+00,
          8.1578e-01,  3.8114e-01, -9.4520e-01, -2.5134e-01, -5.3149e-01]])

    logits_torch = torch.from_numpy(logits_np)
    preds_torch = logits_torch.max(dim=1)[1]
    print("torch preds", preds_torch)

    logits_tf = tf.convert_to_tensor(logits_np)
    preds_tf = tf.argmax(logits_tf, axis=1)
    print("tf preds", preds_tf)
    
    # preds1 = logits.max(dim=1)[1]
    # preds2 = logits.argmax(dim=1)
    # print("logits.max(dim=1)", logits.max(dim=1))
    # print("preds1", preds1)
    # print("preds2", preds2)
    # print(preds1.all() == preds2.all())

    # logits = tf.convert_to_tensor(np.array([[ 2.0835e+00,  2.8684e-01, -3.8064e-01, -6.6219e-03, -4.9030e-02,
    #      -7.0756e-01, -1.6587e-01,  3.7841e-02,  8.4555e-01, -2.8922e-01,
    #      -7.8332e-01,  1.3700e+00,  1.6362e-01, -5.7401e-01,  1.6290e+00,
    #       4.1743e-01, -5.2905e-01,  2.2805e-01, -2.6672e+00, -1.6487e-01,
    #       2.8328e-01, -3.1602e-01,  1.8948e+00,  8.3975e-01,  3.1507e-01,
    #       1.8589e+00, -5.6998e-01, -9.6480e-01, -5.6747e-01, -3.3958e-01,
    #      -1.6440e+00, -3.5754e-01, -7.9882e-01,  2.3381e-01,  6.8907e-02,
    #      -4.4294e-01,  3.3727e-02, -3.4425e-01, -4.4010e-01, -5.5271e-01],
    #     [-4.0442e-01, -3.7830e-01,  7.4986e-01, -1.9608e-01, -5.2080e-01,
    #      -2.9841e-01, -1.6777e+00, -1.8840e-01, -8.2758e-01, -5.0043e-01,
    #       6.0243e-01, -4.9775e-01, -3.9553e-01,  2.3184e-01,  1.5759e+00,
    #      -1.3985e+00,  4.8364e-01,  7.5944e-02, -9.1163e-01, -1.0120e+00,
    #      -1.0840e+00, -1.1333e+00,  7.3563e-01,  7.2417e-01, -3.8258e-01,
    #       2.1450e+00, -2.9394e-01,  1.6343e+00,  1.5634e-01, -7.5331e-01,
    #       1.8599e+00,  2.9857e-01,  2.1150e-01,  4.3066e-01, -6.0429e-01,
    #       9.0129e-01,  2.6864e-01,  2.7161e+00, -1.0155e+00, -8.1855e-01],
    #     [-2.7429e+00, -2.9181e+00,  3.0671e+00,  1.3629e+00, -1.5668e+00,
    #      -8.3793e-01, -3.7353e+00,  4.0808e-01, -1.5371e+00, -1.9909e+00,
    #       3.9708e-01, -1.5419e+00, -1.9746e+00, -1.3238e-01, -1.5096e+00,
    #       4.9266e-01, -1.7744e-01,  1.1247e-01,  6.9266e-01, -8.2123e-01,
    #      -3.9770e+00, -2.1140e+00,  9.9297e-01, -1.1321e+00, -8.7195e-01,
    #       5.5473e+00,  1.0612e+00,  2.4748e+00,  3.0698e-01, -2.0269e+00,
    #       1.7185e+00,  5.8352e+00,  3.6988e+00, -9.0987e-01,  8.1961e-01,
    #      -5.3954e-01, -1.4290e+00,  8.5451e+00, -2.6355e+00, -1.2917e+00],
    #     [ 1.1162e+01,  6.6738e-02, -1.8976e+00, -4.3443e+00, -1.9987e+00,
    #      -1.7699e+00,  4.5539e+00,  5.1470e+00,  3.7383e+00, -8.5479e-02,
    #      -3.3659e+00,  6.0107e+00,  2.9951e-02, -1.3219e+00,  1.0165e+00,
    #       1.7946e+00, -1.4292e+00, -1.0467e+00, -3.3188e+00, -2.4000e+00,
    #       4.3941e+00,  9.8264e-02, -1.6303e+00, -2.3844e+00, -6.7374e-01,
    #      -4.3825e+00,  1.5664e+00, -9.1431e-01, -2.6670e+00, -1.7077e+00,
    #       1.3887e+00, -2.2931e+00, -2.9852e+00, -3.0248e+00, -2.5534e+00,
    #       1.0055e+00, -1.4695e+00, -2.5426e+00, -9.7849e-01, -1.9550e+00],
    #     [ 3.5921e+00,  8.4306e-01, -1.3569e+00, -2.9547e+00,  4.2032e-01,
    #      -1.3029e+00,  1.9426e+00,  1.3834e+00,  1.7537e+00,  9.6460e-03,
    #      -1.5346e+00,  2.0150e+00,  2.5273e-01, -5.5676e-01,  1.9191e+00,
    #      -1.0179e-01,  2.7503e-01, -1.4862e+00, -4.0593e+00, -1.3711e+00,
    #       1.6630e+00,  3.8593e-01,  1.3875e+00,  3.4905e-01,  5.8096e-01,
    #      -1.4988e+00,  1.0648e+00, -1.1916e+00, -4.3314e-01, -1.2875e+00,
    #       8.7663e-01, -7.4405e-01, -1.5142e+00, -2.8483e-02, -2.9628e-01,
    #       2.5531e+00,  6.4484e-01, -2.0996e+00, -5.0988e-01, -8.8208e-01],
    #     [-9.0217e-01, -2.3921e-01,  7.0113e-01, -6.9945e-01, -7.7149e-01,
    #      -1.5625e+00, -1.4836e+00,  2.3214e-01,  2.5339e-01, -1.5055e+00,
    #       1.1614e+00, -9.1424e-01, -6.2381e-01,  4.1322e-02, -2.8845e-01,
    #      -1.1310e-01,  2.6380e+00, -9.1250e-02,  1.5196e+00, -8.0327e-01,
    #      -7.2858e-01, -2.9471e+00,  1.9594e+00,  1.6616e+00, -2.4765e-01,
    #       3.1741e+00, -6.2680e-01,  4.8777e-01,  2.8064e-01, -1.4813e+00,
    #       1.0913e+00,  7.4820e-02,  2.6674e-01,  4.9987e-01, -1.2711e+00,
    #       2.7519e+00,  2.3401e+00,  3.1536e+00, -1.6224e+00, -8.7516e-01],
    #     [-1.6432e+00, -6.1273e-01,  7.9820e-01,  1.0426e+00, -1.2958e+00,
    #      -4.0188e-01, -1.3746e+00, -8.7540e-01, -7.6528e-01, -1.2126e+00,
    #       1.3798e+00, -3.9156e-02, -1.2898e+00,  6.4550e-01,  6.8631e-01,
    #      -1.5323e+00,  5.7061e-01,  9.7129e-01, -2.0090e+00, -8.3243e-01,
    #      -1.9216e+00, -9.5328e-01,  1.8287e+00, -7.1166e-01, -7.5810e-01,
    #       2.4084e+00, -4.9811e-02,  4.6670e-01, -1.6608e-01, -2.6434e-01,
    #       2.1043e+00,  2.0021e+00,  7.8653e-01,  8.5328e-01, -1.0869e-01,
    #      -7.9721e-01,  2.4434e+00,  6.0053e+00, -1.5854e+00, -4.2325e-01],
    #     [ 1.3624e+00,  1.2876e+00, -1.1838e+00, -1.1431e+00,  1.5006e-01,
    #       4.2363e-01,  2.3242e+00,  1.2580e+00,  1.9115e+00,  6.0025e-01,
    #      -1.3256e+00,  3.6321e+00, -2.3834e-01, -1.7159e+00,  7.0819e-01,
    #       8.8843e-01, -4.8113e-02, -2.3913e+00, -9.3553e-01, -1.6055e+00,
    #       2.3139e+00, -5.0882e-01,  5.5813e-01, -6.4976e-01,  5.2835e-01,
    #      -2.4074e+00,  1.4382e+00, -1.7691e+00, -1.6073e+00, -7.5976e-01,
    #       4.4626e-01, -2.1728e+00, -1.8500e+00, -3.3810e-01,  1.0280e+00,
    #       8.1578e-01,  3.8114e-01, -9.4520e-01, -2.5134e-01, -5.3149e-01]]))

    # print()
    # preds = tf.argmax(logits, axis=1)
    # print("preds", preds)


    # dttime = datetime.datetime(year=2022, month=7, day=15)
    # print(dttime)

    # import time
    # t = time.strftime('%Y-%m-%d %H:%M:%S', time.gmtime())
    # print(t)

    # # 打印本地时间
    # print("SS", time.localtime(time.time()))      # 打印本地时间
    # # 打印格式化时间
    # print(time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))       # 打印按指定格式排版的时间
    # FT = time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time()))
    # print(FT)