import torch
from collections import OrderedDict

def parse_cfg(cfgfile):
    def erase_comment(line):
        line = line.split('#')[0]
        return line
    blocks = []
    fp = open(cfgfile, 'r')
    block =  None
    line = fp.readline()
    while line != '':
        line = line.rstrip()
        if line == '' or line[0] == '#':
            line = fp.readline()
            continue        
        elif line[0] == '[':
            if block:
                blocks.append(block)
            block = OrderedDict()
            block['type'] = line.lstrip('[').rstrip(']')
            # set default value
            if block['type'] == 'convolutional':
                block['batch_normalize'] = 0
        else:
            line = erase_comment(line)
            key,value = line.split('=')
            key = key.strip()
            if key == 'type':
                key = '_type'
            value = value.strip()
            block[key] = value
        line = fp.readline()

    if block:
        blocks.append(block)
    fp.close()
    return blocks

def print_cfg(blocks):
    for block in blocks:
        print('[%s]' % (block['type']))
        for key,value in block.items():
            if key != 'type':
                print('%s=%s' % (key, value))
        print('')
def save_cfg(blocks, cfgfile):
    with open(cfgfile, 'w') as fp:
        for block in blocks:
            fp.write('[%s]\n' % (block['type']))
            for key,value in block.items():
                if key != 'type':
                    fp.write('%s=%s\n' % (key, value))
            fp.write('\n')

def print_cfg_nicely(blocks):
    print('layer     filters    size              input                output');
    prev_width = 416
    prev_height = 416
    prev_filters = 3
    out_filters =[]
    out_widths =[]
    out_heights =[]
    ind = -2
    for block in blocks:
        ind = ind + 1
        if block['type'] == 'net':
            prev_width = int(block['width'])
            prev_height = int(block['height'])
            continue
        elif block['type'] == 'convolutional':
            filters = int(block['filters'])
            kernel_size = int(block['size'])
            stride = int(block['stride'])
            is_pad = int(block['pad'])
            pad = (kernel_size-1)/2 if is_pad else 0
            width = (prev_width + 2*pad - kernel_size)/stride + 1
            height = (prev_height + 2*pad - kernel_size)/stride + 1
            print('%5d %-6s %4d  %d x %d / %d   %3d x %3d x%4d   ->   %3d x %3d x%4d' % (ind, 'conv', filters, kernel_size, kernel_size, stride, prev_width, prev_height, prev_filters, width, height, filters))
            prev_width = width
            prev_height = height
            prev_filters = filters
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'maxpool':
            pool_size = int(block['size'])
            stride = int(block['stride'])
            width = prev_width/stride
            height = prev_height/stride
            print('%5d %-6s       %d x %d / %d   %3d x %3d x%4d   ->   %3d x %3d x%4d' % (ind, 'max', pool_size, pool_size, stride, prev_width, prev_height, prev_filters, width, height, filters))
            prev_width = width
            prev_height = height
            prev_filters = filters
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'avgpool':
            width = 1
            height = 1
            print('%5d %-6s                   %3d x %3d x%4d   ->      %3d' % (ind, 'avg', prev_width, prev_height, prev_filters,  prev_filters))
            prev_width = 1
            prev_height = 1
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'softmax':
            print('%5d %-6s                                    ->      %3d' % (ind, 'softmax', prev_filters))
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'cost':
            print('%5d %-6s                                     ->      %3d' % (ind, 'cost', prev_filters))
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'reorg':
            stride = int(block['stride'])
            filters = stride * stride * prev_filters
            width = prev_width/stride
            height = prev_height/stride
            print('%5d %-6s             / %d   %3d x %3d x%4d   ->   %3d x %3d x%4d' % (ind, 'reorg', stride, prev_width, prev_height, prev_filters, width, height, filters))
            prev_width = width
            prev_height = height
            prev_filters = filters
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'route':
            layers = block['layers'].split(',')
            layers = [int(i) if int(i) > 0 else int(i)+ind for i in layers]
            if len(layers) == 1:
                print('%5d %-6s %d' % (ind, 'route', layers[0]))
                prev_width = out_widths[layers[0]]
                prev_height = out_heights[layers[0]]
                prev_filters = out_filters[layers[0]]
            elif len(layers) == 2:
                print('%5d %-6s %d %d' % (ind, 'route', layers[0], layers[1]))
                prev_width = out_widths[layers[0]]
                prev_height = out_heights[layers[0]]
                assert(prev_width == out_widths[layers[1]])
                assert(prev_height == out_heights[layers[1]])
                prev_filters = out_filters[layers[0]] + out_filters[layers[1]]
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'region':
            print('%5d %-6s' % (ind, 'detection'))
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'shortcut':
            from_id = int(block['from'])
            from_id = from_id if from_id > 0 else from_id+ind
            print('%5d %-6s %d' % (ind, 'shortcut', from_id))
            prev_width = out_widths[from_id]
            prev_height = out_heights[from_id]
            prev_filters = out_filters[from_id]
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'softmax':
            print('%5d %-6s' % (ind, 'softmax'))
            out_widths.append(prev_width)
            out_heights.append(prev_height)
            out_filters.append(prev_filters)
        elif block['type'] == 'connected':
            filters = int(block['output'])
            print('%5d %-6s                            %d  ->      %3d' % (ind, 'connected', prev_filters,  filters))
            prev_filters = filters
            out_widths.append(1)
            out_heights.append(1)
            out_filters.append(prev_filters)
        else:
            print('unknown type %s' % (block['type']))

def load_conv(buf, start, conv_model):
    num_w = conv_model.weight.numel()
    num_b = conv_model.bias.numel()
    conv_model.bias.data.copy_(torch.from_numpy(buf[start:start+num_b]));   start = start + num_b
    conv_model.weight.data.copy_(torch.from_numpy(buf[start:start+num_w])); start = start + num_w
    return start

def save_conv(fp, conv_model):
    if conv_model.bias.is_cuda:
        convert2cpu(conv_model.bias.data).numpy().tofile(fp)
        convert2cpu(conv_model.weight.data).numpy().tofile(fp)
    else:
        conv_model.bias.data.numpy().tofile(fp)
        conv_model.weight.data.numpy().tofile(fp)

def load_conv_bn(buf, start, conv_model, bn_model):
    num_w = conv_model.weight.numel()
    num_b = bn_model.bias.numel()
    bn_model.bias.data.copy_(torch.from_numpy(buf[start:start+num_b]));     start = start + num_b
    bn_model.weight.data.copy_(torch.from_numpy(buf[start:start+num_b]));   start = start + num_b
    bn_model.running_mean.copy_(torch.from_numpy(buf[start:start+num_b]));  start = start + num_b
    bn_model.running_var.copy_(torch.from_numpy(buf[start:start+num_b]));   start = start + num_b
    conv_model.weight.data.copy_(torch.from_numpy(buf[start:start+num_w])); start = start + num_w 
    return start

def save_conv_bn(fp, conv_model, bn_model):
    if bn_model.bias.is_cuda:
        convert2cpu(bn_model.bias.data).numpy().tofile(fp)
        convert2cpu(bn_model.weight.data).numpy().tofile(fp)
        convert2cpu(bn_model.running_mean).numpy().tofile(fp)
        convert2cpu(bn_model.running_var).numpy().tofile(fp)
        convert2cpu(conv_model.weight.data).numpy().tofile(fp)
    else:
        bn_model.bias.data.numpy().tofile(fp)
        bn_model.weight.data.numpy().tofile(fp)
        bn_model.running_mean.numpy().tofile(fp)
        bn_model.running_var.numpy().tofile(fp)
        conv_model.weight.data.numpy().tofile(fp)

def save_conv_shrink_bn(fp, conv_model, bn_model, eps=1e-5):
    if bn_model.bias.is_cuda:
        bias = bn_model.bias.data - bn_model.running_mean * bn_model.weight.data / torch.sqrt(bn_model.running_var + eps)
        convert2cpu(bias).numpy().tofile(fp)
        s = conv_model.weight.data.size()
        weight = conv_model.weight.data * (bn_model.weight.data / torch.sqrt(bn_model.running_var + eps)).view(-1,1,1,1).repeat(1, s[1], s[2], s[3])
        convert2cpu(weight).numpy().tofile(fp)
    else:
        bias = bn_model.bias.data - bn_model.running_mean * bn_model.weight.data / torch.sqrt(bn_model.running_var + eps)
        bias.numpy().tofile(fp)
        s = conv_model.weight.data.size()
        weight = conv_model.weight.data * (bn_model.weight.data / torch.sqrt(bn_model.running_var + eps)).view(-1,1,1,1).repeat(1, s[1], s[2], s[3])
        weight.numpy().tofile(fp)

def load_fc(buf, start, fc_model):
    num_w = fc_model.weight.numel()
    num_b = fc_model.bias.numel()
    fc_model.bias.data.copy_(torch.from_numpy(buf[start:start+num_b]));     start = start + num_b
    fc_model.weight.data.copy_(torch.from_numpy(buf[start:start+num_w]));   start = start + num_w 
    return start

def save_fc(fp, fc_model):
    fc_model.bias.data.numpy().tofile(fp)
    fc_model.weight.data.numpy().tofile(fp)


if __name__ == '__main__':
    import sys
    if len(sys.argv) != 2:
        print('Usage: python cfg.py model.cfg')
        exit()

    blocks = parse_cfg(sys.argv[1])
    print_cfg_nicely(blocks)