I am not familiar with the sequence format presented here to know if BioPerl modules such as Bio::IOSeq can handle it as it is, hence its a good idea to clean up the file first and extract only the needed data, unfortunately after having done just that , I got stuck, I feel I may need to use a data structure like hashes of arrays but I can not land on what exactly I have to do to realize this. Also, I have manually removed the part from your sample file that goes like:

# ----- prediction on sequence number 3 (length = 713, name = seq_03) 
+-----
#
# Constraints/Hints:
# (none)
# Predicted genes for sequence number 3 on both strands
# start gene g4 ....
[as same as above]......so on and on...
[download]

However, following is my initial take at it, I hope a wiser monk than myself can land it at its destination so we can learn something new:

#!/usr/local/bin/perl

#Title "extraction of sequences"
#saved the sample file in bioinfo.txt
use strict;
use warnings;
use IO::File;

my $handle = new IO::File;
$handle->autoflush(1);
$handle->open("<bioinfo.txt")  or die("$!");

my @input_array;
my @new_array;

@input_array=<$handle>;
@input_array = grep {s/#//g} @input_array;

for (my $i=0; $i<$#input_array; $i++){
        chomp $input_array[$i];
        delete $input_array[$i] if $input_array[$i]=~ /((none)|checked
+|constraints|predicted)/i; #shedding extras
        next unless $input_array[$i];           #ignoring empty lines.
        push  @new_array, $input_array[$i];     #capturing the element
+s that I need
             }


for(my $i=0;$i<$#new_array;$i++){
        print "$i-$new_array[$i]\n";         #preparing for further pr
+ocessing
        }
[download]

Here is the output from the snippet above:

0- ----- prediction on sequence number 1 (length = 105, name = seq_01)
+ --
1- start gene g1
2- coding sequence = [atgtcgtccctccccactctcatctttctccaccc
3- atcgctgcggtcctcgccgacccttttgtgccggaagtagggaccgg]
4- protein sequence = [MTASAFVLGTVAFLHNRLRRSRPRQASTAHR
5- GTETPLLRSDKENLTTVLDATILVHSLGQKTNLALGATSSSLDLQKTNLAL
6- VAALTPGIVFPLPSPFVATGLCLQKTNLALGATSSSLDL]
7- end gene g1
8- start gene g2
9- coding sequence = [atgccgtcctcgtcaaagcagctggcgatgcc
10- tcggcccctccttctgcaaaccgccctgccgcccgcctcggctcctccgaa
11- gccgagcagcctacgcaggggccgcagatgctcgcgggagggaatatcgg]
12- protein sequence =[MPLDSSSTPTSNPAPSHSSTAYLLFERLHIAEQ
13- CCPGQGIRHGKWSPGSSEAPT]
14- end gene g2
15- ----- prediction on sequence number 2 (length = 710, name = seq_02
+) -----
16- start gene g3
17- coding sequence = [agctgccctcctcggggccagccttctcttaactc
18- tttgagaccttcaatcctgaggcgtgagacgcagtctggaggagcagctc]
19- protein sequence = [LRRETQSGGAALCSLFDPPPTPTACAHANSP]
[download]

Out of curiosity I have translated the gene sequence and also backtranslated the proteins on ExPasy but they're giving out different results than what is in the sample file you provided and are unrelated. Any clues ?