import matplotlib
from torch.nn import functional as F

import torch
matplotlib.use('Agg')
import time

class Timer():
    ''' Timer for recording training time distribution. '''
    def __init__(self):
        self.prev_t = time.time()
        self.clear()

    def set(self):
        self.prev_t = time.time()

    def cnt(self, mode):
        self.time_table[mode] += time.time()-self.prev_t
        self.set()
        if mode == 'bw':
            self.click += 1

    def show(self):
        total_time = sum(self.time_table.values())
        self.time_table['avg'] = total_time/self.click
        self.time_table['rd'] = 100*self.time_table['rd']/total_time
        self.time_table['fw'] = 100*self.time_table['fw']/total_time
        self.time_table['bw'] = 100*self.time_table['bw']/total_time
        msg = '{avg:.3f} sec/step (rd {rd:.1f}% | fw {fw:.1f}% | bw {bw:.1f}%)'.format(
            **self.time_table)
        self.clear()
        return msg

    def clear(self):
        self.time_table = {'rd': 0, 'fw': 0, 'bw': 0}
        self.click = 0

# Reference : https://github.com/espnet/espnet/blob/master/espnet/nets/pytorch_backend/e2e_asr.py#L168

def human_format(num):
    magnitude = 0
    while num >= 1000:
        magnitude += 1
        num /= 1000.0
    # add more suffixes if you need them
    return '{:3.1f}{}'.format(num, [' ', 'K', 'M', 'G', 'T', 'P'][magnitude])


# provide easy access of attribute from dict, such abc.key 
class AttrDict(dict):
    def __init__(self, *args, **kwargs):
        super(AttrDict, self).__init__(*args, **kwargs)
        self.__dict__ = self

def init_weights(m, mean=0.0, std=0.01):
    classname = m.__class__.__name__
    if classname.find("Conv") != -1:
        m.weight.data.normal_(mean, std)

def get_padding(kernel_size, dilation=1):
    return int((kernel_size*dilation - dilation)/2)


def sequence_mask(length, max_length=None):
  if max_length is None:
    max_length = length.max()
  x = torch.arange(max_length, dtype=length.dtype, device=length.device)
  return x.unsqueeze(0) < length.unsqueeze(1)

def slice_segments(x, ids_str, segment_size=4):
  ret = torch.zeros_like(x[:, :, :segment_size])
  for i in range(x.size(0)):
    idx_str = ids_str[i]
    idx_end = idx_str + segment_size
    ret[i] = x[i, :, idx_str:idx_end]
  return ret

def rand_slice_segments(x, x_lengths=None, segment_size=4):
  b, d, t = x.size()
  if x_lengths is None:
    x_lengths = t
  ids_str_max = x_lengths - segment_size + 1
  ids_str = (torch.rand([b]).to(device=x.device) * ids_str_max).to(dtype=torch.long)
  ret = slice_segments(x, ids_str, segment_size)
  return ret, ids_str


def convert_pad_shape(pad_shape):
  l = pad_shape[::-1]
  pad_shape = [item for sublist in l for item in sublist]
  return pad_shape

def generate_path(duration, mask):
  """
  duration: [b, 1, t_x]
  mask: [b, 1, t_y, t_x]
  """
  device = duration.device
  
  b, _, t_y, t_x = mask.shape
  cum_duration = torch.cumsum(duration, -1)
  
  cum_duration_flat = cum_duration.view(b * t_x)
  path = sequence_mask(cum_duration_flat, t_y).to(mask.dtype)
  path = path.view(b, t_x, t_y)
  path = path - F.pad(path, convert_pad_shape([[0, 0], [1, 0], [0, 0]]))[:, :-1]
  path = path.unsqueeze(1).transpose(2,3) * mask
  return path

@torch.jit.script
def fused_add_tanh_sigmoid_multiply(input_a, input_b, n_channels):
  n_channels_int = n_channels[0]
  in_act = input_a + input_b
  t_act = torch.tanh(in_act[:, :n_channels_int, :])
  s_act = torch.sigmoid(in_act[:, n_channels_int:, :])
  acts = t_act * s_act
  return acts


def subsequent_mask(length):
  mask = torch.tril(torch.ones(length, length)).unsqueeze(0).unsqueeze(0)
  return mask


def intersperse(lst, item):
  result = [item] * (len(lst) * 2 + 1)
  result[1::2] = lst
  return result


def clip_grad_value_(parameters, clip_value, norm_type=2):
  if isinstance(parameters, torch.Tensor):
    parameters = [parameters]
  parameters = list(filter(lambda p: p.grad is not None, parameters))
  norm_type = float(norm_type)
  if clip_value is not None:
    clip_value = float(clip_value)

  total_norm = 0
  for p in parameters:
    param_norm = p.grad.data.norm(norm_type)
    total_norm += param_norm.item() ** norm_type
    if clip_value is not None:
      p.grad.data.clamp_(min=-clip_value, max=clip_value)
  total_norm = total_norm ** (1. / norm_type)
  return total_norm