Questions about some numbers regarding dataset and code comment

[IvanWeiYR]

Hi, I have some questions about the numbers that you mentioned in the comment and in the blog. https://keunwoochoi.wordpress.com/2016/02/23/lstmetallica/.

In the code segment in main_lstM_etallica.py:

	path = 'metallica_drums_text.txt' # Corpus file
	text = open(path).read()
	print('corpus length:', len(text))

	if character_mode:
		chars = set(text)
	else:
		chord_seq = text.split(' ')
		chars = set(chord_seq)
		text = chord_seq

	char_indices = dict((c, i) for i, c in enumerate(chars))
	indices_char = dict((i, c) for i, c in enumerate(chars))
	num_chars = len(char_indices)
	print('total chars:', num_chars)

The line:

print('corpus length:', len(text))

Shouldn't it be printed after text.split(' ')? Because you mentioned in the blog that this is the number of words(including bars) in the corpus, which is 2,141,692 if we print it here. But if we print it after the split, the number will be 185760.

Also, for the "total number of words", you mentioned in the blog that it is 119, but I found that 119 contains one empty string (''), is that one really needed?

And for the line:

if __name__=='__main__':
	for maxlen in [256]: # for wrod, 256 is about 32 bars.

Since you quantised the note into 16th note, shouldn't 256 be 16 bars?