Home of the Squeezebox™ & Transporter® network music players.
Page 5 of 6 FirstFirst ... 3456 LastLast
Results 41 to 50 of 53
  1. #41
    Senior Member
    Join Date
    Mar 2017
    Posts
    2,953
    Quote Originally Posted by Roland0 View Post
    Not sure what you mean by "album style tracks".
    I suspect this "confusion" is my fault. As you know, Musly requires you to add a random selection of tracks to set the music 'style'. Usually these are just random tracks from the library. I thought it might help to try and take at least 1 track (at random) from each album - hence the term "album style tracks". Not sure it this really makes much difference, it was just something I was playing with...
    Material debug: 1. Launch via http: //SERVER:9000/material/?debug=json (Use http: //SERVER:9000/material/?debug=json,cometd to also see update messages, e.g. play queue) 2. Open browser's developer tools 3. Open console tab in developer tools 4. REQ/RESP messages sent to/from LMS will be logged here.

  2. #42
    Senior Member
    Join Date
    Mar 2017
    Posts
    2,953
    Quote Originally Posted by afriend View Post
    Anyway, I'll probably need a couple of months before I can say anything on how much I like the results musly provides. Thank you for your help and scripts/plugins.
    Just to let you know, Musly has a bug where sometimes all songs get the same similarity score. I noticed this last night after I did a complete rescan. My mixer produced odd results, and they were all from the first artist/albums (ordered A-Z). I looked at the logging and all tracks had the same similarity. This was with using 2000 for 'styletracks', I also tried 1500 and had similar results. Using 1000 (which is the default anyway) seemed to work OK. Changing 'styletracks' requires the jukebox file to be removed, but this only takes a minute or so to be recreated.

    Thought you might like to know in case you are getting really weird results.
    Material debug: 1. Launch via http: //SERVER:9000/material/?debug=json (Use http: //SERVER:9000/material/?debug=json,cometd to also see update messages, e.g. play queue) 2. Open browser's developer tools 3. Open console tab in developer tools 4. REQ/RESP messages sent to/from LMS will be logged here.

  3. #43
    Senior Member
    Join Date
    Jul 2010
    Location
    Oz
    Posts
    272
    Quote Originally Posted by cpd73 View Post
    Just to let you know, Musly has a bug where sometimes all songs get the same similarity score. I noticed this last night after I did a complete rescan. My mixer produced odd results, and they were all from the first artist/albums (ordered A-Z). I looked at the logging and all tracks had the same similarity. This was with using 2000 for 'styletracks', I also tried 1500 and had similar results. Using 1000 (which is the default anyway) seemed to work OK. Changing 'styletracks' requires the jukebox file to be removed, but this only takes a minute or so to be recreated.

    Thought you might like to know in case you are getting really weird results.
    Thank you for letting me know (was using the default 1000). Do you know how many tracks that bug affects? Anyway, I've been quite pleased with Musly so far. Even if there's an odd one every now it's not completely off. I really appreciate this new way to listen to my local music.

    Just one question:
    I remember Roland0 saying that starting the Musly server (without a jukebox file) can take quite some time. And it certainly does on my old hardware!
    So in order to avoid this long startup could I delete the jukebox while Musly is still running? Would the jukebox be recreated on demand in this case or is it necessary to restart Musly server without a jukebox file for it to be rcecreated?
    Thank you.

  4. #44
    Senior Member
    Join Date
    Mar 2017
    Posts
    2,953
    Quote Originally Posted by afriend View Post
    Thank you for letting me know (was using the default 1000). Do you know how many tracks that bug affects?
    It only affects the similarity function, not the analysis. When broken all similarity checks return the same value. To fix you delete the jukbox and recreate with a different styletracks value.

    Quote Originally Posted by afriend View Post
    I remember Roland0 saying that starting the Musly server (without a jukebox file) can take quite some time. And it certainly does on my old hardware!
    So in order to avoid this long startup could I delete the jukebox while Musly is still running? Would the jukebox be recreated on demand in this case or is it necessary to restart Musly server without a jukebox file for it to be rcecreated?
    If you delete the jukebox it needs to be recreated - which is what takes the time. Using an exisiting jukebox file is much faster. Note that my code also does lots of SQL queries to try and choose 1 track for each album for this jukebox, the Musly docs jsut state to choose ~1000 random tracks. So, not sure if my using 1 per album helps or not, but it is slightly slower - as it performs SQL searches to choose a random track per album. I could make this optional - e.g. toggle between this semi-random or just choosing 1000 random tracks. But its musly's subsequent handling of these to create the jukebox that can be slow. I create my jukebox on an i7 laptop, and that takes about 1 min for 1000 style tracks. I then copy the musly DB and jukebox to my Pi for the mixer to use.

    Jukebox is not created whists it is running, it is (AFAIK) read a startup (and created then if it does not exist) and then held in memory.

    Why do you want to remove this file?
    Material debug: 1. Launch via http: //SERVER:9000/material/?debug=json (Use http: //SERVER:9000/material/?debug=json,cometd to also see update messages, e.g. play queue) 2. Open browser's developer tools 3. Open console tab in developer tools 4. REQ/RESP messages sent to/from LMS will be logged here.

  5. #45
    Senior Member
    Join Date
    Jul 2010
    Location
    Oz
    Posts
    272
    Quote Originally Posted by cpd73 View Post
    Changing 'styletracks' requires the jukebox file to be removed, but this only takes a minute or so to be recreated.
    Sorry, misread that one. Wasn't necessary. And I did it anyway. :-)

    If I could ask one thing in some future update it would be debug log markers that tell me (maybe every 1000 tracks or so) about the progress of those steps that take time / loop through the library.

    Some day I might test musly analysis on a more recent machine. For now, I'm stuck with this old one.

    Thanks for your help.

  6. #46
    Senior Member
    Join Date
    Jul 2010
    Location
    Oz
    Posts
    272
    I've been using Similarity a bit more. I like it. Far better than random mixes.

    Can I just ask about genre filtering: if I create only one genre filter group for Classical music genres does this mean that if I start a mix with no classical music seed tracks does it mean that this mix will cover all genres except classical music? Just asking because while I'd like a classical music mix to feature only classical music I don't want classical music tracks to appear in any other mix.

    Do you think you could make the number of tracks DSTM adds (default=5) a config option? The DSTM similarity mix uses the similarity score of the last 5 seed tracks (the ones it just added) and therefore it would a kind of 'evolving' mix changing the character of the playlist as it 'evolves'? Compared to always using the similarity score of the original/first seed tracks. If it's evolving then larger batches would mean that the 'character' of the playlist would change more slowly (which can be nice). Or maybe another option to lock the similarity scores to the original seed tracks to keep the character of the playlist?

    I'm very glad you added the "Create Similarity Mix" context menu item and that it adds 50 tracks at once because the DSTM mixer is not really working for me right now.

    Plugins::MusicSimilarity::Plugin::__ANON__ (176) Failed to fetch URL: Timed out waiting for data
    Plugins::MusicSimilarity::Plugin::_mixFailed (202) Return empty list

    The first batch is usually added without problems but when the second batch is supposed to be added (last but one song started) it seems DSTM or Similarity times out before musly server has returned the results and DSTM adds the 'song mix'. The muslyserver python process is still busy when DSTM adds the song mix.
    At first I thought it was related to my setup: I often use a touchscreen piCorePlayer (without Squeezelite running) to control other clients - like a Boom in this case.

    Do you know how to fix this? Or this someting that can only be addressed in the DSTM plugin? (didn't find any timeout option in DSTM though)

    Thank you.

  7. #47
    Senior Member
    Join Date
    Mar 2017
    Posts
    2,953
    Quote Originally Posted by afriend View Post
    I've been using Similarity a bit more. I like it. Far better than random mixes.
    Glad you like it

    Quote Originally Posted by afriend View Post
    Can I just ask about genre filtering: if I create only one genre filter group for Classical music genres does this mean that if I start a mix with no classical music seed tracks does it mean that this mix will cover all genres except classical music? Just asking because while I'd like a classical music mix to feature only classical music I don't want classical music tracks to appear in any other mix.
    A track that is in a a genre group will only match others in the group. A track not in a genre group will only match others not in a genre group. So, there's a kind of "all others" genre group.

    Quote Originally Posted by afriend View Post
    Do you think you could make the number of tracks DSTM adds (default=5) a config option? The DSTM similarity mix uses the similarity score of the last 5 seed tracks (the ones it just added) and therefore it would a kind of 'evolving' mix changing the character of the playlist as it 'evolves'?
    Can do, but not sure I see the use. DSTM, AFAIK, picks seed tracks at random each time, its not based upon the 1st/last tracks - so should be different tracks each time.

    Quote Originally Posted by afriend View Post
    the DSTM mixer is not really working for me right now.
    How so? If you run the musly-server locally you should see what seed tracks are being used, along with the filtering, etc. I only ever use the DSTM side, the 'Create Mix' was just added as an experiment - I've never really used it in anger.

    Quote Originally Posted by afriend View Post
    The muslyserver python process is still busy when DSTM adds the song mix.
    What hardware are you running this on? I run the musly-server on a Pi4 with LMS, with ~25k tracks, and never seen such issues.
    Material debug: 1. Launch via http: //SERVER:9000/material/?debug=json (Use http: //SERVER:9000/material/?debug=json,cometd to also see update messages, e.g. play queue) 2. Open browser's developer tools 3. Open console tab in developer tools 4. REQ/RESP messages sent to/from LMS will be logged here.

  8. #48
    Senior Member
    Join Date
    Jul 2010
    Location
    Oz
    Posts
    272
    Thank you for adding that option. I'm a bit of a serialskipper sometimes. This setting will also help prevent me looping back to the beginning of the playlist when I skip too many tracks before enough new tracks have been added. Even if it wasn't intended to do that :-)

    And you're right, my hardware is very old (10y+), runs LMS, muslyserver and other stuff, and I have more tracks in my db than you so response times are not snappy, at all. And the machine goes to sleep if not used + the music files are stored on an external HDD (spin-up time). In the end it all works but definitely much more slowly than on modern PCs like yours. But I won't throw it away unless it's really broken.

    Good news is that I may have 'fixed' my song mix/timeout problem for now. Your Similarity code suggests that you'd planned to make the getjsondata timeout value a configurable pref option but maybe forgot to add the UI settings part? A timeout value of 5 secs is probably too low for my paleo setup. So I manually set the timout value to 25 secs in the prefs file. So far it works. Which makes me happy because I'd find it difficult to go back to random mixes once you got used to musly mixes.

    P.S. No matter what, I like the 'create mix' option. Fast way to start a mix from 1 track or to see what musly thinks is similar without starting a mix (kind of like a dry-run).

    EDIT:
    I just saw that similarity added 3 tracks (by different artists) from the same compilation album in a row. Does the 'don't repeat album' setting not apply to compilations? My compilations have album artist='Various Artist‘ and compilation flag=1.
    Last edited by afriend; 2021-06-12 at 05:16.

  9. #49
    Senior Member
    Join Date
    Mar 2017
    Posts
    2,953
    Quote Originally Posted by afriend View Post
    Good news is that I may have 'fixed' my song mix/timeout problem for now. Your Similarity code suggests that you'd planned to make the getjsondata timeout value a configurable pref option but maybe forgot to add the UI settings part? A timeout value of 5 secs is probably too low for my paleo setup. So I manually set the timout value to 25 secs in the prefs file.
    Ah! Perhaps I should just change the default to 30seconds? Strange your machine is so slow, as I run on a Pi4 - and that's hardly the fastest machine...

    [Edit] Changed default to 30 seconds.


    Quote Originally Posted by afriend View Post
    I just saw that similarity added 3 tracks (by different artists) from the same compilation album in a row. Does the 'don't repeat album' setting not apply to compilations? My compilations have album artist='Various Artist‘ and compilation flag=1.
    As you noted 'don't repeat album' does not apply to 'Various Artists' - my thinking was they are already a mix of similar tracks, so no harm if they play one after another. I just did not want repeated tracks by the same artist - which Musly seems to produce very often.
    Last edited by cpd73; 2021-06-14 at 13:28.
    Material debug: 1. Launch via http: //SERVER:9000/material/?debug=json (Use http: //SERVER:9000/material/?debug=json,cometd to also see update messages, e.g. play queue) 2. Open browser's developer tools 3. Open console tab in developer tools 4. REQ/RESP messages sent to/from LMS will be logged here.

  10. #50
    Senior Member
    Join Date
    Jul 2010
    Location
    Oz
    Posts
    272
    Quote Originally Posted by cpd73 View Post
    [Edit] Changed default to 30 seconds.
    I did actually change exactly that value in your code for testing. You probably won't believe this (I wouldn't) but even though logic dictates otherwise in the end I wasn't able to get rid of this timeout problem until I set timeout: 30 in the plugin prefs file. A mystery. :-) Anyway, it works now, so...

    Quote Originally Posted by cpd73 View Post
    As you noted 'don't repeat album' does not apply to 'Various Artists' - my thinking was they are already a mix of similar tracks, so no harm if they play one after another. I just did not want repeated tracks by the same artist - which Musly seems to produce very often
    I see. I'm usually hoping for musly to give me similar tracks but also from different sources/albums/artists/genres. It's an interesting way to (re)discover my library. So for a compilation would mean similar tracks but the same source.

    One suggestion if I may: often it either takes too many clicks (iPeng) or it's impossible (jivelite - pCP, Touch, Radio) to access a plugin's setting page. Therefore I thought maybe you could just register 2 similarity DSTM mixers: 1 without and 1 with enabled genre filtering. The DSTM menu is only 1 click away and also available in jivelite.

    I'm really using Musly/Similarity a lot right now and sometimes I wish there were other ways to use the similarity score/musly results than DSTM.
    Like... choose a track and get a list with similiar tracks by that track's artist, starting with the most similar ones.
    Or... get a list of similar dancable tracks from a certain decade and a certain genre.
    The DynamicPlaylist plugin allows such choices before starting a mix. Maybe musly/similarity could be used with DPL? I don't know. Just throwing around some ideas...
    Last edited by afriend; 2021-06-17 at 08:06.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •