From 76a2cbe57025c96c47b64df3dc317065873278cb Mon Sep 17 00:00:00 2001 From: leavelet Date: Wed, 6 Jul 2022 12:21:09 +0800 Subject: [PATCH] add requirements readme --- README.md | 25 +- requirements/align_environment_cpu.yml | 27 ++ requirements/align_environment_gpu.yml | 22 ++ requirements/melody_extraction.txt | 4 +- .../vocal-extraction/environment-m1.yml | 2 +- requirements/vocal-extraction/environment.yml | 4 +- utils/english-align/prepare_dict.sh | 306 ------------------ 7 files changed, 74 insertions(+), 316 deletions(-) create mode 100644 requirements/align_environment_cpu.yml create mode 100644 requirements/align_environment_gpu.yml delete mode 100755 utils/english-align/prepare_dict.sh diff --git a/README.md b/README.md index 79d9e53..6e65af7 100644 --- a/README.md +++ b/README.md @@ -94,14 +94,29 @@ When dealing with mutiple AI projects, it will make your life much easier to set We use recommand conda, and we take this as an example -2. crate environment for song-cutter and those non-AI program +2. crate environments ```shell - conda create -n music-dealer + cd requirements + + # misc + conda create -n singing-dealer + conda activate singing-dealer + pip install -r make_midi.txt + + # vocal extraction + conda env create -f vocal-extraction/environment.yml + #if you use arm mac, use environment-m1.yml + conda activate vocal-extraction + pip install -r vocal-extraction/requirements.txt + + # alignment & melody extraction + conda env create -f vocal-extraction/maker_ai_cpu.yml + # if you have gpu, use maker_ai_gpu.yml + conda activate maker_ai + pip install -r melody_extraction.txt ``` - - ### not finished, next update on 6.28 - + ### Seperate the song diff --git a/requirements/align_environment_cpu.yml b/requirements/align_environment_cpu.yml new file mode 100644 index 0000000..5f48e07 --- /dev/null +++ b/requirements/align_environment_cpu.yml @@ -0,0 +1,27 @@ +name: maker_ai +channels: + - pytorch + - conda-forge + - defaults +dependencies: + - audioread=2.1.8 + - decorator=4.4.1 + - ffmpeg=4.2 + - libflac=1.3.1 + - libiconv=1.15 + - libllvm8=8.0.1 + - libogg=1.3.2 + - librosa=0.7.2 + - libsndfile=1.0.28 + - libuuid=2.32.1 + - libvorbis=1.3.5 + - libxcb=1.13 + - numba=0.47.0 + - pyqt=5.6.0 + - pysoundfile=0.10.2 + - resampy=0.2.2 + - numpy=1.18.1 + - numpy-base=1.18.1 + - python=3.6.10 + - pytorch=1.5.0 + diff --git a/requirements/align_environment_gpu.yml b/requirements/align_environment_gpu.yml new file mode 100644 index 0000000..8073b7e --- /dev/null +++ b/requirements/align_environment_gpu.yml @@ -0,0 +1,22 @@ +name: maker_ai +channels: + - pytorch + - conda-forge + - defaults +dependencies: + - audioread=2.1.8 + - decorator=4.4.1 + - libsndfile=1.0.28 + - libuuid=2.32.1 + - libvorbis=1.3.5 + - libxcb=1.13 + - numba=0.47.0 + - pyqt=5.6.0 + - pysoundfile=0.10.2 + - resampy=0.2.2 + - numpy=1.18.1 + - numpy-base=1.18.1 + - python=3.6.10 + - pytorch=1.5.0 + - cudatoolkit=9.2 + diff --git a/requirements/melody_extraction.txt b/requirements/melody_extraction.txt index 43b3803..d04a619 100644 --- a/requirements/melody_extraction.txt +++ b/requirements/melody_extraction.txt @@ -5,4 +5,6 @@ madmom==0.16.1 numpy==1.18.5 SciPy==1.1.0 pysndfx==0.3.6 -tensorflow-gpu==1.15.5 \ No newline at end of file +tensorflow-gpu==1.15.5 +decorator==4.4.1 +cudatoolkit=9.2 diff --git a/requirements/vocal-extraction/environment-m1.yml b/requirements/vocal-extraction/environment-m1.yml index f29d59e..462d0aa 100644 --- a/requirements/vocal-extraction/environment-m1.yml +++ b/requirements/vocal-extraction/environment-m1.yml @@ -1,4 +1,4 @@ -name: pytorch +name: vocal-extract channels: - conda-forge - anaconda diff --git a/requirements/vocal-extraction/environment.yml b/requirements/vocal-extraction/environment.yml index 5b9b613..cd59a4d 100644 --- a/requirements/vocal-extraction/environment.yml +++ b/requirements/vocal-extraction/environment.yml @@ -1,4 +1,4 @@ -name: extract +name: vocal-extract channels: - conda-forge - anaconda @@ -22,5 +22,3 @@ dependencies: - asteroid>=0.5.0 - demucs==2.0.3 -prefix: /home/ielab/miniconda3/envs/pytorch - diff --git a/utils/english-align/prepare_dict.sh b/utils/english-align/prepare_dict.sh deleted file mode 100755 index a8221ba..0000000 --- a/utils/english-align/prepare_dict.sh +++ /dev/null @@ -1,306 +0,0 @@ -#!/usr/bin/env bash -#Copyright 2016 LeSpeech (Author: Xingyu Na) - -# prepare dictionary for aidatatang -# it is done for English and Chinese separately, -# For English, we use CMU dictionary, and Sequitur G2P -# for OOVs, while all englist phone set will concert to Chinese -# phone set at the end. For Chinese, we use an online dictionary, -# for OOV, we just produce pronunciation using Charactrt Mapping. - -dict_dir=./dict -mkdir -p $dict_dir -mkdir -p $dict_dir/lexicon-{en,ch} - -# extract full vocabulary -cat dict/text.txt | awk '{for (i = 2; i <= NF; i++) print $i}' |\ - perl -ape 's/ /\n/g;' | sort -u | grep -v '\[LAUGHTER\]' | grep -v '\[NOISE\]' |\ - grep -v '\[VOCALIZED-NOISE\]' > $dict_dir/words.txt || exit 1; - -# split into English and Chinese -cat $dict_dir/words.txt > $dict_dir/lexicon-en/words-en.txt || exit 1; - - -##### produce pronunciations for english -if [ ! -f $dict_dir/cmudict/cmudict.0.7a ]; then - echo "--- Downloading CMU dictionary ..." - svn co -r 13068 https://svn.code.sf.net/p/cmusphinx/code/trunk/cmudict \ - $dict_dir/cmudict || exit 1; -fi - -# format cmudict -echo "--- Striping stress and pronunciation variant markers from cmudict ..." -perl $dict_dir/cmudict/scripts/make_baseform.pl \ - $dict_dir/cmudict/cmudict.0.7a /dev/stdout |\ - sed -e 's:^\([^\s(]\+\)([0-9]\+)\(\s\+\)\(.*\):\1\2\3:' > $dict_dir/cmudict/cmudict-plain.txt || exit 1; - -# extract in-vocab lexicon and oov words -echo "--- Searching for English OOV words ..." -echo 1 -awk 'NR==FNR{words[$1]; next;} !($1 in words)' \ - $dict_dir/cmudict/cmudict-plain.txt $dict_dir/lexicon-en/words-en.txt |\ - egrep -v '<.?s>' > $dict_dir/lexicon-en/words-en-oov.txt; -echo 2 -awk 'NR==FNR{words[$1]; next;} ($1 in words)' \ - $dict_dir/lexicon-en/words-en.txt $dict_dir/cmudict/cmudict-plain.txt |\ - egrep -v '<.?s>' > $dict_dir/lexicon-en/lexicon-en-iv.txt; - -echo 1 - -wc -l $dict_dir/lexicon-en/words-en-oov.txt -wc -l $dict_dir/lexicon-en/lexicon-en-iv.txt - - -echo "--- Preparing pronunciations for OOV words ..." -g2p=`which g2p.py` -if [ ! -x $g2p ]; then - echo "g2p.py is not found. Checkout tools/extras/install_sequitur.sh." - exit 1 -fi -g2p.py --model=conf/g2p_model --apply $dict_dir/lexicon-en/words-en-oov.txt \ - > $dict_dir/lexicon-en/lexicon-en-oov.txt || exit 1; - -# merge in-vocab and oov lexicon -cat $dict_dir/lexicon-en/lexicon-en-oov.txt $dict_dir/lexicon-en/lexicon-en-iv.txt |\ - sort > $dict_dir/lexicon-en/lexicon-en-phn.txt || exit 1; - -# convert cmu phoneme to pinyin phonenme -mkdir -p $dict_dir/map -cat conf/cmu2pinyin | awk '{print $1g;}' | sort -u > $dict_dir/map/cmu || exit 1; -cat conf/pinyin2cmu | awk -v cmu=$dict_dir/map/cmu \ - 'BEGIN{while((getline $dict_dir/map/cmu-used || exit 1; -cat $dict_dir/map/cmu | awk -v cmu=$dict_dir/map/cmu-used \ - 'BEGIN{while((getline $dict_dir/map/cmu-not-used || exit 1; - -awk 'NR==FNR{words[$1]; next;} ($1 in words)' \ - $dict_dir/map/cmu-not-used conf/cmu2pinyin |\ - egrep -v '<.?s>' > $dict_dir/map/cmu-py || exit 1; - -cat $dict_dir/map/cmu-py | \ - perl -e ' - open(MAPS, $ARGV[0]) or die("could not open map file"); - my %py2ph; - foreach $line () { - @A = split(" ", $line); - $py = shift(@A); - $py2ph{$py} = [@A]; - } - my @entry; - while () { - @A = split(" ", $_); - @entry = (); - $W = shift(@A); - push(@entry, $W); - for($i = 0; $i < @A; $i++) { push(@entry, @{$py2ph{$A[$i]}}); } - print "@entry"; - print "\n"; - } -' conf/pinyin2cmu > $dict_dir/map/cmu-cmu || exit 1; - -cat $dict_dir/lexicon-en/lexicon-en-phn.txt | \ - perl -e ' - open(MAPS, $ARGV[0]) or die("could not open map file"); - my %py2ph; - foreach $line () { - @A = split(" ", $line); - $py = shift(@A); - $py2ph{$py} = [@A]; - } - my @entry; - while () { - @A = split(" ", $_); - @entry = (); - $W = shift(@A); - push(@entry, $W); - for($i = 0; $i < @A; $i++) { - if (exists $py2ph{$A[$i]}) { push(@entry, @{$py2ph{$A[$i]}}); } - else {push(@entry, $A[$i])}; - } - print "@entry"; - print "\n"; - } -' $dict_dir/map/cmu-cmu > $dict_dir/lexicon-en/lexicon-en.txt || exit 1; - - -##### produce pronunciations for chinese -if [ ! -f $dict_dir/cedict/cedict_1_0_ts_utf-8_mdbg.txt ]; then - echo "------------- Downloading cedit dictionary ---------------" - mkdir -p $dict_dir/cedict - wget -P $dict_dir/cedict https://www.mdbg.net/chinese/export/cedict/cedict_1_0_ts_utf-8_mdbg.txt.gz - gzip -d $dict_dir/cedict/cedict_1_0_ts_utf-8_mdbg.txt.gz -fi - -cat $dict_dir/cedict/cedict_1_0_ts_utf-8_mdbg.txt | grep -v '#' | awk -F '/' '{print $1}' |\ - perl -e ' - while () { - @A = split(" ", $_); - print $A[1]; - for($n = 2; $n < @A; $n++) { - $A[$n] =~ s:\[?([a-zA-Z0-9\:]+)\]?:$1:; - $tmp = uc($A[$n]); - print " $tmp"; - } - print "\n"; - } - ' | sort -k1 > $dict_dir/cedict/ch-dict.txt || exit 1; - -echo "--- Searching for Chinese OOV words ..." -awk 'NR==FNR{words[$1]; next;} !($1 in words)' \ - $dict_dir/cedict/ch-dict.txt $dict_dir/lexicon-ch/words-ch.txt |\ - egrep -v '<.?s>' > $dict_dir/lexicon-ch/words-ch-oov.txt || exit 1; - -awk 'NR==FNR{words[$1]; next;} ($1 in words)' \ - $dict_dir/lexicon-ch/words-ch.txt $dict_dir/cedict/ch-dict.txt |\ - egrep -v '<.?s>' > $dict_dir/lexicon-ch/lexicon-ch-iv.txt || exit 1; - -wc -l $dict_dir/lexicon-ch/words-ch-oov.txt -wc -l $dict_dir/lexicon-ch/lexicon-ch-iv.txt - - -# validate Chinese dictionary and compose a char-based -# dictionary in order to get OOV pronunciations -cat $dict_dir/cedict/ch-dict.txt |\ - perl -e ' - use utf8; - binmode(STDIN,":encoding(utf8)"); - binmode(STDOUT,":encoding(utf8)"); - while () { - @A = split(" ", $_); - $word_len = length($A[0]); - $proun_len = @A - 1 ; - if ($word_len == $proun_len) {print $_;} - } - ' > $dict_dir/cedict/ch-dict-1.txt || exit 1; - -# extract chars -cat $dict_dir/cedict/ch-dict-1.txt | awk '{print $1}' |\ - perl -e ' - use utf8; - binmode(STDIN,":encoding(utf8)"); - binmode(STDOUT,":encoding(utf8)"); - while () { - @A = split(" ", $_); - @chars = split("", $A[0]); - foreach (@chars) { - print "$_\n"; - } - } - ' | grep -v '^$' > $dict_dir/lexicon-ch/ch-char.txt || exit 1; - -# extract individual pinyins -cat $dict_dir/cedict/ch-dict-1.txt |\ - awk '{for(i=2; i<=NF; i++) print $i}' |\ - perl -ape 's/ /\n/g;' > $dict_dir/lexicon-ch/ch-char-pinyin.txt || exit 1; - -# first make sure number of characters and pinyins -# are equal, so that a char-based dictionary can -# be composed. -nchars=`wc -l < $dict_dir/lexicon-ch/ch-char.txt` -npinyin=`wc -l < $dict_dir/lexicon-ch/ch-char-pinyin.txt` -if [ $nchars -ne $npinyin ]; then - echo "Found $nchars chars and $npinyin pinyin. Please check!" - exit 1 -fi - -paste $dict_dir/lexicon-ch/ch-char.txt $dict_dir/lexicon-ch/ch-char-pinyin.txt |\ - sort -u > $dict_dir/lexicon-ch/ch-char-dict.txt || exit 1; - -# create a multiple pronunciation dictionary -cat $dict_dir/lexicon-ch/ch-char-dict.txt |\ - perl -e ' - my $prev = ""; - my $out_line = ""; - while () { - @A = split(" ", $_); - $cur = $A[0]; - $cur_py = $A[1]; - #print length($prev); - if (length($prev) == 0) { $out_line = $_; chomp($out_line);} - if (length($prev)>0 && $cur ne $prev) { print $out_line; print "\n"; $out_line = $_; chomp($out_line);} - if (length($prev)>0 && $cur eq $prev) { $out_line = $out_line."/"."$cur_py";} - $prev = $cur; - } - print $out_line; - ' > $dict_dir/lexicon-ch/ch-char-dict-mp.txt || exit 1; - -# get lexicon for Chinese OOV words -local/create_oov_char_lexicon.pl $dict_dir/lexicon-ch/ch-char-dict-mp.txt \ - $dict_dir/lexicon-ch/words-ch-oov.txt > $dict_dir/lexicon-ch/lexicon-ch-oov.txt || exit 1; - -# seperate multiple prons for Chinese OOV lexicon -cat $dict_dir/lexicon-ch/lexicon-ch-oov.txt |\ - perl -e ' - my @entry; - my @entry1; - while () { - @A = split(" ", $_); - @entry = (); - push(@entry, $A[0]); - for($i = 1; $i < @A; $i++ ) { - @py = split("/", $A[$i]); - @entry1 = @entry; - @entry = (); - for ($j = 0; $j < @entry1; $j++) { - for ($k = 0; $k < @py; $k++) { - $tmp = $entry1[$j]." ".$py[$k]; - push(@entry, $tmp); - } - } - } - for ($i = 0; $i < @entry; $i++) { - print $entry[$i]; - print "\n"; - } - } - ' > $dict_dir/lexicon-ch/lexicon-ch-oov-mp.txt || exit 1; - -# compose IV and OOV lexicons for Chinese -cat $dict_dir/lexicon-ch/lexicon-ch-oov-mp.txt $dict_dir/lexicon-ch/lexicon-ch-iv.txt |\ - awk '{if (NF > 1 && $2 ~ /[A-Za-z0-9]+/) print $0;}' > $dict_dir/lexicon-ch/lexicon-ch.txt || exit 1; - -# convert Chinese pinyin to CMU format -cat $dict_dir/lexicon-ch/lexicon-ch.txt | sed -e 's/U:/V/g' | sed -e 's/ R\([0-9]\)/ ER\1/g'|\ - utils/pinyin_map.pl conf/pinyin2cmu > $dict_dir/lexicon-ch/lexicon-ch-cmu.txt || exit 1; - -# combine English and Chinese lexicons -cat $dict_dir/lexicon-en/lexicon-en.txt $dict_dir/lexicon-ch/lexicon-ch-cmu.txt |\ - sort -u > $dict_dir/lexicon1.txt || exit 1; - -cat $dict_dir/lexicon1.txt | awk '{ for(n=2;n<=NF;n++){ phones[$n] = 1; }} END{for (p in phones) print p;}'| \ - sort -u |\ - perl -e ' - my %ph_cl; - while () { - $phone = $_; - chomp($phone); - chomp($_); - $phone =~ s:([A-Z]+)[0-9]:$1:; - if (exists $ph_cl{$phone}) { push(@{$ph_cl{$phone}}, $_) } - else { $ph_cl{$phone} = [$_]; } - } - foreach $key ( keys %ph_cl ) { - print "@{ $ph_cl{$key} }\n" - } - ' | sort -k1 > $dict_dir/nonsilence_phones.txt || exit 1; - -( echo SIL; echo SPN; echo NSN; echo LAU ) > $dict_dir/silence_phones.txt - -echo SIL > $dict_dir/optional_silence.txt - -# No "extra questions" in the input to this setup, as we don't -# have stress or tone - -cat $dict_dir/silence_phones.txt| awk '{printf("%s ", $1);} END{printf "\n";}' > $dict_dir/extra_questions.txt || exit 1; -cat $dict_dir/nonsilence_phones.txt | perl -e 'while(<>){ foreach $p (split(" ", $_)) { - $p =~ m:^([^\d]+)(\d*)$: || die "Bad phone $_"; $q{$2} .= "$p "; } } foreach $l (values %q) {print "$l\n";}' \ - >> $dict_dir/extra_questions.txt || exit 1; - -# Add to the lexicon the silences, noises etc. -(echo '!SIL SIL'; echo '[VOCALIZED-NOISE] SPN'; echo '[NOISE] NSN'; echo '[LAUGHTER] LAU'; - echo ' SPN' ) | \ - cat - $dict_dir/lexicon1.txt > $dict_dir/lexicon.txt || exit 1; - -echo "$0: aidatatang_200zh dict preparation succeeded" -exit 0;