Upload 4 files
Browse files
app.js
CHANGED
|
@@ -44,17 +44,55 @@ const canvasCtx = noteCanvas.getContext("2d");
|
|
| 44 |
|
| 45 |
// Inference & Audio Variables
|
| 46 |
let activeModelName = "";
|
| 47 |
-
let
|
| 48 |
-
let session;
|
| 49 |
-
let stoi;
|
| 50 |
-
let itos;
|
| 51 |
-
let vocabSize = 0;
|
| 52 |
-
let h;
|
| 53 |
-
let c;
|
| 54 |
-
let currentId;
|
| 55 |
let running = false;
|
| 56 |
let pumping = false;
|
| 57 |
let schedulerId = 0;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
let nextNoteTime = 0;
|
| 59 |
let generatedEvents = 0;
|
| 60 |
let queue = [];
|
|
@@ -569,141 +607,6 @@ function updateReadouts(lastToken = null) {
|
|
| 569 |
if (topKValue) topKValue.textContent = topKInput.value;
|
| 570 |
}
|
| 571 |
|
| 572 |
-
function tokenId(token) {
|
| 573 |
-
return stoi?.[token] ?? null;
|
| 574 |
-
}
|
| 575 |
-
|
| 576 |
-
function makeTensorId(id) {
|
| 577 |
-
return new ort.Tensor("int64", BigInt64Array.from([BigInt(id)]), [1, 1]);
|
| 578 |
-
}
|
| 579 |
-
|
| 580 |
-
function zeroState() {
|
| 581 |
-
return new ort.Tensor("float32", new Float32Array(numLayers * hiddenSize), [numLayers, 1, hiddenSize]);
|
| 582 |
-
}
|
| 583 |
-
|
| 584 |
-
// Dimensions are updated explicitly on model load
|
| 585 |
-
|
| 586 |
-
function promptIds() {
|
| 587 |
-
if (midiSeedActive && midiTokens.length > 0) {
|
| 588 |
-
const sliced = getTokensUpToBar(midiTokens, midiStartBar);
|
| 589 |
-
return sliced.map(tokenId).filter((id) => id !== null);
|
| 590 |
-
}
|
| 591 |
-
const eventPrompt = ["BOS", "BPM_120", "GRID_64", "BAR", "POS_0"];
|
| 592 |
-
const abcPrompt = ["X:", "1", "<NL>", "T:", "piece", "<NL>", "M:", "4/4", "<NL>", "L:", "1/8", "<NL>", "Q:", "1/4=120", "<NL>", "K:", "C", "<NL>"];
|
| 593 |
-
const prompt = activeTokenizer === "giantmidi_event" ? eventPrompt : abcPrompt;
|
| 594 |
-
const ids = prompt.map(tokenId).filter((id) => id !== null);
|
| 595 |
-
return ids.length ? ids : [0];
|
| 596 |
-
}
|
| 597 |
-
|
| 598 |
-
async function warmPrompt() {
|
| 599 |
-
h = zeroState();
|
| 600 |
-
c = zeroState();
|
| 601 |
-
|
| 602 |
-
const allTokensStr = midiSeedActive && midiTokens.length > 0
|
| 603 |
-
? getTokensUpToBar(midiTokens, midiStartBar)
|
| 604 |
-
: (activeTokenizer === "giantmidi_event"
|
| 605 |
-
? ["BOS", "BPM_120", "GRID_64", "BAR", "POS_0"]
|
| 606 |
-
: ["X:", "1", "<NL>", "T:", "piece", "<NL>", "M:", "4/4", "<NL>", "L:", "1/8", "<NL>", "Q:", "1/4=120", "<NL>", "K:", "C", "<NL>"]);
|
| 607 |
-
|
| 608 |
-
isWarmingUp = true;
|
| 609 |
-
|
| 610 |
-
// 1. Feed all prompt tokens to parser to advance its clock
|
| 611 |
-
for (const token of allTokensStr) {
|
| 612 |
-
if (parser) {
|
| 613 |
-
parser.feed(token);
|
| 614 |
-
}
|
| 615 |
-
}
|
| 616 |
-
|
| 617 |
-
// 2. Slice model warm-up tokens to last 256 (matching model's context window)
|
| 618 |
-
const modelTokensStr = allTokensStr.slice(-256);
|
| 619 |
-
const ids = modelTokensStr.map(tokenId).filter((id) => id !== null);
|
| 620 |
-
|
| 621 |
-
for (const id of ids) {
|
| 622 |
-
currentId = id;
|
| 623 |
-
await stepModel(false);
|
| 624 |
-
}
|
| 625 |
-
|
| 626 |
-
isWarmingUp = false;
|
| 627 |
-
}
|
| 628 |
-
|
| 629 |
-
function sampleFromLogits(logits) {
|
| 630 |
-
const temperature = Math.max(0.05, Number(tempInput.value));
|
| 631 |
-
const topK = Math.max(1, Number(topKInput.value));
|
| 632 |
-
const scored = [];
|
| 633 |
-
for (let i = 0; i < logits.length; i += 1) {
|
| 634 |
-
const token = itos[i];
|
| 635 |
-
if (token === undefined) continue;
|
| 636 |
-
let score = logits[i] / temperature;
|
| 637 |
-
if (token === "<EOP>" || token === "EOS") score -= 1.0;
|
| 638 |
-
scored.push([i, score]);
|
| 639 |
-
}
|
| 640 |
-
scored.sort((a, b) => b[1] - a[1]);
|
| 641 |
-
const picked = scored.slice(0, Math.min(topK, scored.length));
|
| 642 |
-
const maxScore = picked[0]?.[1] ?? 0;
|
| 643 |
-
let sum = 0;
|
| 644 |
-
for (const item of picked) {
|
| 645 |
-
item[2] = Math.exp(item[1] - maxScore);
|
| 646 |
-
sum += item[2];
|
| 647 |
-
}
|
| 648 |
-
let r = Math.random() * sum;
|
| 649 |
-
for (const item of picked) {
|
| 650 |
-
r -= item[2];
|
| 651 |
-
if (r <= 0) return item[0];
|
| 652 |
-
}
|
| 653 |
-
return picked[picked.length - 1][0];
|
| 654 |
-
}
|
| 655 |
-
|
| 656 |
-
async function stepModel(record = true) {
|
| 657 |
-
const output = await session.run({ input: makeTensorId(currentId), h, c });
|
| 658 |
-
h = output.h_out;
|
| 659 |
-
c = output.c_out;
|
| 660 |
-
currentId = sampleFromLogits(output.logits.data);
|
| 661 |
-
const token = itos[currentId] ?? "?";
|
| 662 |
-
if (record) {
|
| 663 |
-
updateReadouts(token);
|
| 664 |
-
parser.feed(token);
|
| 665 |
-
}
|
| 666 |
-
return token;
|
| 667 |
-
}
|
| 668 |
-
|
| 669 |
-
function durationToSecondsFromAbc(duration) {
|
| 670 |
-
let eighths = 1;
|
| 671 |
-
if (duration) {
|
| 672 |
-
if (duration === "/") eighths = 0.5;
|
| 673 |
-
else if (duration.startsWith("/")) eighths = 1 / Number(duration.slice(1));
|
| 674 |
-
else if (duration.includes("/")) {
|
| 675 |
-
const [a, b] = duration.split("/").map(Number);
|
| 676 |
-
eighths = a / b;
|
| 677 |
-
} else {
|
| 678 |
-
eighths = Number(duration);
|
| 679 |
-
}
|
| 680 |
-
}
|
| 681 |
-
const quarterSeconds = 60 / Number(tempoEl.textContent || 120);
|
| 682 |
-
return Math.max(0.08, eighths * quarterSeconds * 0.5);
|
| 683 |
-
}
|
| 684 |
-
|
| 685 |
-
function durationToSecondsFromEventSteps(steps, grid) {
|
| 686 |
-
const quarterSeconds = 60 / Number(tempoEl.textContent || 120);
|
| 687 |
-
return Math.max(0.035, (Math.max(1, steps) * 4 * quarterSeconds) / Math.max(1, grid));
|
| 688 |
-
}
|
| 689 |
-
|
| 690 |
-
function midiToTonePitch(midi) {
|
| 691 |
-
const names = ["C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B"];
|
| 692 |
-
return `${names[((midi % 12) + 12) % 12]}${Math.floor(midi / 12) - 1}`;
|
| 693 |
-
}
|
| 694 |
-
|
| 695 |
-
function abcToTonePitch(token) {
|
| 696 |
-
const match = token.match(/^(\^\^|__|\^|_|=)?([A-Ga-g])([,']*)$/);
|
| 697 |
-
if (!match) return null;
|
| 698 |
-
const accidental = match[1] || "";
|
| 699 |
-
const step = match[2];
|
| 700 |
-
const marks = match[3] || "";
|
| 701 |
-
let octave = step === step.toLowerCase() ? 5 : 4;
|
| 702 |
-
for (const mark of marks) octave += mark === "'" ? 1 : -1;
|
| 703 |
-
const accidentalText = accidental === "^^" ? "##" : accidental === "__" ? "bb" : accidental === "^" ? "#" : accidental === "_" ? "b" : "";
|
| 704 |
-
return `${step.toUpperCase()}${accidentalText}${octave}`;
|
| 705 |
-
}
|
| 706 |
-
|
| 707 |
function midiFromPitchName(note) {
|
| 708 |
const match = String(note).match(/^([A-G])([#b]{0,2})(-?\d+)$/);
|
| 709 |
if (!match) return null;
|
|
@@ -732,200 +635,20 @@ function recordVisualEvent(event, startTime) {
|
|
| 732 |
}
|
| 733 |
|
| 734 |
function pushEvent(event) {
|
| 735 |
-
if (isWarmingUp) return; // Discard prompt history events
|
| 736 |
if (event.duration < 0 || queue.length > TARGET_QUEUE_EVENTS * 3) return;
|
| 737 |
queue.push(event);
|
| 738 |
generatedEvents += event.type === "note" ? 1 : 0;
|
| 739 |
updateReadouts();
|
| 740 |
}
|
| 741 |
|
| 742 |
-
function
|
| 743 |
-
return {
|
| 744 |
-
pending: null,
|
| 745 |
-
chord: null,
|
| 746 |
-
reset() {
|
| 747 |
-
this.pending = null;
|
| 748 |
-
this.chord = null;
|
| 749 |
-
},
|
| 750 |
-
feed(token) {
|
| 751 |
-
if (token === "Q:") {
|
| 752 |
-
this.pending = { type: "tempoHeader" };
|
| 753 |
-
return;
|
| 754 |
-
}
|
| 755 |
-
if (this.pending?.type === "tempoHeader") {
|
| 756 |
-
const bpm = Number(String(token).split("=").pop());
|
| 757 |
-
if (Number.isFinite(bpm) && bpm >= 40 && bpm <= 220) {
|
| 758 |
-
if (!midiSeedActive || isWarmingUp) {
|
| 759 |
-
if (tempoEl) tempoEl.textContent = String(Math.round(bpm));
|
| 760 |
-
}
|
| 761 |
-
}
|
| 762 |
-
this.pending = null;
|
| 763 |
-
return;
|
| 764 |
-
}
|
| 765 |
-
if (token === "[") {
|
| 766 |
-
this.chord = [];
|
| 767 |
-
this.pending = null;
|
| 768 |
-
return;
|
| 769 |
-
}
|
| 770 |
-
if (this.chord) {
|
| 771 |
-
if (token === "]") {
|
| 772 |
-
this.pending = { type: "chord", notes: this.chord.map(abcToTonePitch).filter(Boolean) };
|
| 773 |
-
this.chord = null;
|
| 774 |
-
return;
|
| 775 |
-
}
|
| 776 |
-
if (abcNotePattern.test(token)) this.chord.push(token);
|
| 777 |
-
return;
|
| 778 |
-
}
|
| 779 |
-
if (token === "z") {
|
| 780 |
-
this.pending = { type: "rest" };
|
| 781 |
-
return;
|
| 782 |
-
}
|
| 783 |
-
if (abcNotePattern.test(token)) {
|
| 784 |
-
this.pending = { type: "note", note: abcToTonePitch(token) };
|
| 785 |
-
return;
|
| 786 |
-
}
|
| 787 |
-
if (token.startsWith("DUR:") && this.pending) {
|
| 788 |
-
this.emitPending(token.slice(4));
|
| 789 |
-
return;
|
| 790 |
-
}
|
| 791 |
-
if (this.pending && (token === "|" || token === "<NL>" || token === "<EOP>")) this.emitPending(null);
|
| 792 |
-
},
|
| 793 |
-
emitPending(durationToken) {
|
| 794 |
-
const duration = durationToSecondsFromAbc(durationToken);
|
| 795 |
-
if (this.pending.type === "rest") pushEvent({ type: "rest", duration });
|
| 796 |
-
if (this.pending.type === "note" && this.pending.note) pushEvent({ type: "note", notes: [this.pending.note], duration, advance: duration });
|
| 797 |
-
if (this.pending.type === "chord" && this.pending.notes.length) pushEvent({ type: "note", notes: this.pending.notes.slice(0, 8), duration, advance: duration });
|
| 798 |
-
this.pending = null;
|
| 799 |
-
},
|
| 800 |
-
};
|
| 801 |
-
}
|
| 802 |
-
|
| 803 |
-
function makeEventParser() {
|
| 804 |
-
return {
|
| 805 |
-
grid: 64,
|
| 806 |
-
bar: -1,
|
| 807 |
-
pos: 0,
|
| 808 |
-
pendingPosition: null,
|
| 809 |
-
pendingNotes: [],
|
| 810 |
-
pendingNote: null,
|
| 811 |
-
lastQ: 0,
|
| 812 |
-
reset() {
|
| 813 |
-
this.grid = tokenId("GRID_64") !== null ? 64 : 64;
|
| 814 |
-
this.bar = -1;
|
| 815 |
-
this.pos = 0;
|
| 816 |
-
this.pendingPosition = null;
|
| 817 |
-
this.pendingNotes = [];
|
| 818 |
-
this.pendingNote = null;
|
| 819 |
-
this.lastQ = 0;
|
| 820 |
-
},
|
| 821 |
-
feed(token) {
|
| 822 |
-
if (token.startsWith("BPM_")) {
|
| 823 |
-
const bpm = Number(token.slice(4));
|
| 824 |
-
if (Number.isFinite(bpm) && bpm >= 40 && bpm <= 220) {
|
| 825 |
-
if (!midiSeedActive || isWarmingUp) {
|
| 826 |
-
if (tempoEl) tempoEl.textContent = String(Math.round(bpm));
|
| 827 |
-
}
|
| 828 |
-
}
|
| 829 |
-
return;
|
| 830 |
-
}
|
| 831 |
-
if (token.startsWith("GRID_")) {
|
| 832 |
-
const grid = Number(token.slice(5));
|
| 833 |
-
if (Number.isFinite(grid) && grid > 0) this.grid = grid;
|
| 834 |
-
return;
|
| 835 |
-
}
|
| 836 |
-
if (token === "BAR") {
|
| 837 |
-
this.flushTo(this.absoluteQFor(this.bar + 1, 0));
|
| 838 |
-
this.bar += 1;
|
| 839 |
-
this.pos = 0;
|
| 840 |
-
return;
|
| 841 |
-
}
|
| 842 |
-
if (token.startsWith("POS_")) {
|
| 843 |
-
const pos = Number(token.slice(4));
|
| 844 |
-
if (!Number.isFinite(pos)) return;
|
| 845 |
-
this.flushTo(this.absoluteQFor(this.bar, pos));
|
| 846 |
-
this.pos = pos;
|
| 847 |
-
return;
|
| 848 |
-
}
|
| 849 |
-
if (token.startsWith("NOTE_")) {
|
| 850 |
-
const midi = Number(token.slice(5));
|
| 851 |
-
if (Number.isFinite(midi) && midi >= 0 && midi <= 127) {
|
| 852 |
-
this.pendingNote = { midi, durationSteps: 1, velocity: 0.72 };
|
| 853 |
-
}
|
| 854 |
-
return;
|
| 855 |
-
}
|
| 856 |
-
if (token.startsWith("DUR_") && this.pendingNote) {
|
| 857 |
-
const steps = Number(token.slice(4));
|
| 858 |
-
if (Number.isFinite(steps)) this.pendingNote.durationSteps = Math.max(1, steps);
|
| 859 |
-
return;
|
| 860 |
-
}
|
| 861 |
-
if (token.startsWith("VEL_") && this.pendingNote) {
|
| 862 |
-
const bucket = Number(token.slice(4));
|
| 863 |
-
if (Number.isFinite(bucket)) this.pendingNote.velocity = Math.max(0.2, Math.min(0.95, bucket / 8));
|
| 864 |
-
const q = this.absoluteQFor(this.bar, this.pos);
|
| 865 |
-
this.pendingPosition ??= q;
|
| 866 |
-
this.pendingNotes.push(this.pendingNote);
|
| 867 |
-
this.pendingNote = null;
|
| 868 |
-
}
|
| 869 |
-
},
|
| 870 |
-
absoluteQFor(bar, pos) {
|
| 871 |
-
return Math.max(0, bar) * 4 + (Math.max(0, pos) * 4) / Math.max(1, this.grid);
|
| 872 |
-
},
|
| 873 |
-
flushTo(nextQ) {
|
| 874 |
-
if (this.pendingNote) {
|
| 875 |
-
const q = this.absoluteQFor(this.bar, this.pos);
|
| 876 |
-
this.pendingPosition ??= q;
|
| 877 |
-
this.pendingNotes.push(this.pendingNote);
|
| 878 |
-
this.pendingNote = null;
|
| 879 |
-
}
|
| 880 |
-
if (this.pendingNotes.length && this.pendingPosition !== null) {
|
| 881 |
-
const gap = Math.max(0, this.pendingPosition - this.lastQ);
|
| 882 |
-
if (gap > 0) pushEvent({ type: "rest", duration: this.quartersToSeconds(gap) });
|
| 883 |
-
pushEvent({
|
| 884 |
-
type: "note",
|
| 885 |
-
notes: this.pendingNotes.map((note) => midiToTonePitch(note.midi)),
|
| 886 |
-
perNoteDurations: this.pendingNotes.map((note) => durationToSecondsFromEventSteps(note.durationSteps, this.grid)),
|
| 887 |
-
velocities: this.pendingNotes.map((note) => note.velocity),
|
| 888 |
-
duration: 0,
|
| 889 |
-
advance: 0,
|
| 890 |
-
});
|
| 891 |
-
this.lastQ = this.pendingPosition;
|
| 892 |
-
}
|
| 893 |
-
const finalGap = Math.max(0, nextQ - this.lastQ);
|
| 894 |
-
if (finalGap > 0) pushEvent({ type: "rest", duration: this.quartersToSeconds(finalGap) });
|
| 895 |
-
this.lastQ = Math.max(this.lastQ, nextQ);
|
| 896 |
-
this.pendingPosition = null;
|
| 897 |
-
this.pendingNotes = [];
|
| 898 |
-
},
|
| 899 |
-
quartersToSeconds(quarters) {
|
| 900 |
-
return (quarters * 60) / Number(tempoEl.textContent || 120);
|
| 901 |
-
},
|
| 902 |
-
};
|
| 903 |
-
}
|
| 904 |
-
|
| 905 |
-
function makeParser() {
|
| 906 |
-
return activeTokenizer === "giantmidi_event" ? makeEventParser() : makeAbcParser();
|
| 907 |
-
}
|
| 908 |
-
|
| 909 |
-
async function pumpTokens() {
|
| 910 |
if (!running || pumping || queue.length >= TARGET_QUEUE_EVENTS) return;
|
| 911 |
pumping = true;
|
| 912 |
-
|
| 913 |
-
|
| 914 |
-
|
| 915 |
-
|
| 916 |
-
|
| 917 |
-
steps += 1;
|
| 918 |
-
}
|
| 919 |
-
if (running) {
|
| 920 |
-
setStatus(`Playing ${activeModelName}`, "playing");
|
| 921 |
-
}
|
| 922 |
-
} catch (error) {
|
| 923 |
-
console.error(error);
|
| 924 |
-
setStatus(`Generation error: ${error.message}`, "idle");
|
| 925 |
-
stop();
|
| 926 |
-
} finally {
|
| 927 |
-
pumping = false;
|
| 928 |
-
}
|
| 929 |
}
|
| 930 |
|
| 931 |
function scheduleAudio() {
|
|
@@ -1030,7 +753,7 @@ async function ensurePiano() {
|
|
| 1030 |
}
|
| 1031 |
|
| 1032 |
async function start() {
|
| 1033 |
-
if (running || !
|
| 1034 |
startBtn.disabled = true;
|
| 1035 |
setStatus("Starting audio...", "loading");
|
| 1036 |
|
|
@@ -1042,11 +765,21 @@ async function start() {
|
|
| 1042 |
visualNotes = [];
|
| 1043 |
if (tempoEl) tempoEl.textContent = midiSeedActive ? String(Math.round(midiBpm)) : "120";
|
| 1044 |
|
| 1045 |
-
parser = makeParser();
|
| 1046 |
-
parser.reset();
|
| 1047 |
-
|
| 1048 |
setStatus("Warming AI model prompt...", "buffering");
|
| 1049 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1050 |
|
| 1051 |
running = true;
|
| 1052 |
stopBtn.disabled = false;
|
|
@@ -1059,7 +792,6 @@ async function start() {
|
|
| 1059 |
midiStartBarInput.disabled = true;
|
| 1060 |
|
| 1061 |
setStatus(`Playing ${activeModelName}`, "playing");
|
| 1062 |
-
await pumpTokens();
|
| 1063 |
|
| 1064 |
nextNoteTime = Tone.now() + 0.12;
|
| 1065 |
startVisualizer();
|
|
@@ -1073,7 +805,10 @@ function stop() {
|
|
| 1073 |
running = false;
|
| 1074 |
window.clearTimeout(schedulerId);
|
| 1075 |
schedulerId = 0;
|
| 1076 |
-
|
|
|
|
|
|
|
|
|
|
| 1077 |
stopBtn.disabled = true;
|
| 1078 |
if (newSongBtn) newSongBtn.disabled = true;
|
| 1079 |
if (recordBtn) recordBtn.disabled = true;
|
|
@@ -1096,7 +831,7 @@ function stop() {
|
|
| 1096 |
}
|
| 1097 |
if (visualFrame) cancelAnimationFrame(visualFrame);
|
| 1098 |
visualFrame = 0;
|
| 1099 |
-
setStatus(
|
| 1100 |
}
|
| 1101 |
|
| 1102 |
// MIDI Recording functions
|
|
@@ -1245,7 +980,7 @@ function downloadMidi(fileBytes, filename = "nanomaestro_performance.mid") {
|
|
| 1245 |
|
| 1246 |
async function loadSelectedModel() {
|
| 1247 |
stop();
|
| 1248 |
-
|
| 1249 |
startBtn.disabled = true;
|
| 1250 |
if (!modelFileInput || !vocabFileInput) {
|
| 1251 |
setStatus("Select model_int8.onnx and matching vocab.json in Advanced settings", "idle");
|
|
@@ -1261,18 +996,21 @@ async function loadSelectedModel() {
|
|
| 1261 |
activeModelName = modelFile.name;
|
| 1262 |
setStatus(`Reading ${vocabFile.name}...`, "loading");
|
| 1263 |
const vocab = JSON.parse(await vocabFile.text());
|
| 1264 |
-
stoi = vocab.stoi;
|
| 1265 |
-
itos = Object.fromEntries(Object.entries(vocab.itos).map(([key, value]) => [Number(key), value]));
|
| 1266 |
-
vocabSize = vocab.vocab_size;
|
| 1267 |
-
activeTokenizer = vocab.tokenizer || "abc";
|
| 1268 |
|
| 1269 |
setStatus(`Loading ${modelFile.name}...`, "loading");
|
| 1270 |
const modelBuffer = await modelFile.arrayBuffer();
|
| 1271 |
-
|
| 1272 |
-
|
| 1273 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1274 |
});
|
| 1275 |
-
|
| 1276 |
setStatus(`Ready: ${activeModelName}`, "idle");
|
| 1277 |
startBtn.disabled = false;
|
| 1278 |
}
|
|
@@ -1595,10 +1333,6 @@ async function init() {
|
|
| 1595 |
// Step 1. Load Vocabulary
|
| 1596 |
updateProgress(5, "Downloading vocabulary metadata...");
|
| 1597 |
const vocab = await fetchWithCache("https://huggingface.co/utkucoban/NanoMaestro-Realtime/resolve/main/NanoMaestro-Realtime/NM2.4%20ONNX%20int8/vocab.json", null, true);
|
| 1598 |
-
stoi = vocab.stoi;
|
| 1599 |
-
itos = Object.fromEntries(Object.entries(vocab.itos).map(([key, value]) => [Number(key), value]));
|
| 1600 |
-
vocabSize = vocab.vocab_size;
|
| 1601 |
-
activeTokenizer = vocab.tokenizer || "abc";
|
| 1602 |
|
| 1603 |
// Step 2. Download ONNX Model
|
| 1604 |
updateProgress(15, "Downloading AI model...");
|
|
@@ -1610,18 +1344,20 @@ async function init() {
|
|
| 1610 |
updateProgress(pct, `Downloading AI model (${loadedMB}MB / ${totalMB}MB)...`);
|
| 1611 |
}, false);
|
| 1612 |
|
| 1613 |
-
// Step 3. Initialize inference session
|
| 1614 |
updateProgress(82, "Initializing neural network engine...");
|
| 1615 |
-
ort.env.wasm.wasmPaths = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.20.1/dist/";
|
| 1616 |
-
ort.env.wasm.numThreads = Math.min(4, navigator.hardwareConcurrency || 1);
|
| 1617 |
|
| 1618 |
const modelBuffer = await modelBlob.arrayBuffer();
|
| 1619 |
-
|
| 1620 |
-
|
| 1621 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1622 |
});
|
| 1623 |
-
numLayers = 2;
|
| 1624 |
-
hiddenSize = 1024;
|
| 1625 |
activeModelName = "model_int8.onnx (v2.4)";
|
| 1626 |
|
| 1627 |
// Step 4. Loading Piano Samples
|
|
|
|
| 44 |
|
| 45 |
// Inference & Audio Variables
|
| 46 |
let activeModelName = "";
|
| 47 |
+
let isModelReady = false;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
let running = false;
|
| 49 |
let pumping = false;
|
| 50 |
let schedulerId = 0;
|
| 51 |
+
|
| 52 |
+
// Web Worker instance & routing
|
| 53 |
+
const worker = new Worker("./worker.js");
|
| 54 |
+
let modelInitResolve = null;
|
| 55 |
+
let modelInitReject = null;
|
| 56 |
+
let startResolve = null;
|
| 57 |
+
let startReject = null;
|
| 58 |
+
|
| 59 |
+
worker.onmessage = function (e) {
|
| 60 |
+
const data = e.data;
|
| 61 |
+
switch (data.action) {
|
| 62 |
+
case "initialized":
|
| 63 |
+
isModelReady = true;
|
| 64 |
+
if (modelInitResolve) modelInitResolve();
|
| 65 |
+
break;
|
| 66 |
+
case "started":
|
| 67 |
+
if (data.events) {
|
| 68 |
+
data.events.forEach(event => pushEvent(event));
|
| 69 |
+
}
|
| 70 |
+
if (data.lastToken) {
|
| 71 |
+
updateReadouts(data.lastToken);
|
| 72 |
+
}
|
| 73 |
+
if (startResolve) startResolve();
|
| 74 |
+
break;
|
| 75 |
+
case "events":
|
| 76 |
+
if (data.events) {
|
| 77 |
+
data.events.forEach(event => pushEvent(event));
|
| 78 |
+
}
|
| 79 |
+
if (data.lastToken) {
|
| 80 |
+
updateReadouts(data.lastToken);
|
| 81 |
+
}
|
| 82 |
+
pumping = false;
|
| 83 |
+
break;
|
| 84 |
+
case "tempo":
|
| 85 |
+
if (tempoEl) tempoEl.textContent = String(Math.round(data.bpm));
|
| 86 |
+
break;
|
| 87 |
+
case "error":
|
| 88 |
+
console.error("Worker error:", data.message);
|
| 89 |
+
setStatus(data.message, "idle");
|
| 90 |
+
stop();
|
| 91 |
+
if (modelInitReject) modelInitReject(new Error(data.message));
|
| 92 |
+
if (startReject) startReject(new Error(data.message));
|
| 93 |
+
break;
|
| 94 |
+
}
|
| 95 |
+
};
|
| 96 |
let nextNoteTime = 0;
|
| 97 |
let generatedEvents = 0;
|
| 98 |
let queue = [];
|
|
|
|
| 607 |
if (topKValue) topKValue.textContent = topKInput.value;
|
| 608 |
}
|
| 609 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 610 |
function midiFromPitchName(note) {
|
| 611 |
const match = String(note).match(/^([A-G])([#b]{0,2})(-?\d+)$/);
|
| 612 |
if (!match) return null;
|
|
|
|
| 635 |
}
|
| 636 |
|
| 637 |
function pushEvent(event) {
|
|
|
|
| 638 |
if (event.duration < 0 || queue.length > TARGET_QUEUE_EVENTS * 3) return;
|
| 639 |
queue.push(event);
|
| 640 |
generatedEvents += event.type === "note" ? 1 : 0;
|
| 641 |
updateReadouts();
|
| 642 |
}
|
| 643 |
|
| 644 |
+
function pumpTokens() {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 645 |
if (!running || pumping || queue.length >= TARGET_QUEUE_EVENTS) return;
|
| 646 |
pumping = true;
|
| 647 |
+
worker.postMessage({
|
| 648 |
+
action: "pump",
|
| 649 |
+
temperature: Math.max(0.05, Number(tempInput.value)),
|
| 650 |
+
topK: Math.max(1, Number(topKInput.value))
|
| 651 |
+
});
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 652 |
}
|
| 653 |
|
| 654 |
function scheduleAudio() {
|
|
|
|
| 753 |
}
|
| 754 |
|
| 755 |
async function start() {
|
| 756 |
+
if (running || !isModelReady) return;
|
| 757 |
startBtn.disabled = true;
|
| 758 |
setStatus("Starting audio...", "loading");
|
| 759 |
|
|
|
|
| 765 |
visualNotes = [];
|
| 766 |
if (tempoEl) tempoEl.textContent = midiSeedActive ? String(Math.round(midiBpm)) : "120";
|
| 767 |
|
|
|
|
|
|
|
|
|
|
| 768 |
setStatus("Warming AI model prompt...", "buffering");
|
| 769 |
+
|
| 770 |
+
await new Promise((resolve, reject) => {
|
| 771 |
+
startResolve = resolve;
|
| 772 |
+
startReject = reject;
|
| 773 |
+
worker.postMessage({
|
| 774 |
+
action: "start",
|
| 775 |
+
temperature: Math.max(0.05, Number(tempInput.value)),
|
| 776 |
+
topK: Math.max(1, Number(topKInput.value)),
|
| 777 |
+
bpm: Number(tempoEl.textContent || 120),
|
| 778 |
+
midiSeedActive: midiSeedActive,
|
| 779 |
+
midiTokens: midiSeedActive ? midiTokens : [],
|
| 780 |
+
midiStartBar: midiStartBar
|
| 781 |
+
});
|
| 782 |
+
});
|
| 783 |
|
| 784 |
running = true;
|
| 785 |
stopBtn.disabled = false;
|
|
|
|
| 792 |
midiStartBarInput.disabled = true;
|
| 793 |
|
| 794 |
setStatus(`Playing ${activeModelName}`, "playing");
|
|
|
|
| 795 |
|
| 796 |
nextNoteTime = Tone.now() + 0.12;
|
| 797 |
startVisualizer();
|
|
|
|
| 805 |
running = false;
|
| 806 |
window.clearTimeout(schedulerId);
|
| 807 |
schedulerId = 0;
|
| 808 |
+
|
| 809 |
+
worker.postMessage({ action: "stop" });
|
| 810 |
+
|
| 811 |
+
startBtn.disabled = !isModelReady;
|
| 812 |
stopBtn.disabled = true;
|
| 813 |
if (newSongBtn) newSongBtn.disabled = true;
|
| 814 |
if (recordBtn) recordBtn.disabled = true;
|
|
|
|
| 831 |
}
|
| 832 |
if (visualFrame) cancelAnimationFrame(visualFrame);
|
| 833 |
visualFrame = 0;
|
| 834 |
+
setStatus(isModelReady ? `Ready: ${activeModelName}` : "Stopped", "idle");
|
| 835 |
}
|
| 836 |
|
| 837 |
// MIDI Recording functions
|
|
|
|
| 980 |
|
| 981 |
async function loadSelectedModel() {
|
| 982 |
stop();
|
| 983 |
+
isModelReady = false;
|
| 984 |
startBtn.disabled = true;
|
| 985 |
if (!modelFileInput || !vocabFileInput) {
|
| 986 |
setStatus("Select model_int8.onnx and matching vocab.json in Advanced settings", "idle");
|
|
|
|
| 996 |
activeModelName = modelFile.name;
|
| 997 |
setStatus(`Reading ${vocabFile.name}...`, "loading");
|
| 998 |
const vocab = JSON.parse(await vocabFile.text());
|
|
|
|
|
|
|
|
|
|
|
|
|
| 999 |
|
| 1000 |
setStatus(`Loading ${modelFile.name}...`, "loading");
|
| 1001 |
const modelBuffer = await modelFile.arrayBuffer();
|
| 1002 |
+
|
| 1003 |
+
await new Promise((resolve, reject) => {
|
| 1004 |
+
modelInitResolve = resolve;
|
| 1005 |
+
modelInitReject = reject;
|
| 1006 |
+
worker.postMessage({
|
| 1007 |
+
action: "init",
|
| 1008 |
+
activeModelName: modelFile.name,
|
| 1009 |
+
vocab: vocab,
|
| 1010 |
+
modelBuffer: modelBuffer
|
| 1011 |
+
}, [modelBuffer]);
|
| 1012 |
});
|
| 1013 |
+
|
| 1014 |
setStatus(`Ready: ${activeModelName}`, "idle");
|
| 1015 |
startBtn.disabled = false;
|
| 1016 |
}
|
|
|
|
| 1333 |
// Step 1. Load Vocabulary
|
| 1334 |
updateProgress(5, "Downloading vocabulary metadata...");
|
| 1335 |
const vocab = await fetchWithCache("https://huggingface.co/utkucoban/NanoMaestro-Realtime/resolve/main/NanoMaestro-Realtime/NM2.4%20ONNX%20int8/vocab.json", null, true);
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1336 |
|
| 1337 |
// Step 2. Download ONNX Model
|
| 1338 |
updateProgress(15, "Downloading AI model...");
|
|
|
|
| 1344 |
updateProgress(pct, `Downloading AI model (${loadedMB}MB / ${totalMB}MB)...`);
|
| 1345 |
}, false);
|
| 1346 |
|
| 1347 |
+
// Step 3. Initialize inference session inside worker
|
| 1348 |
updateProgress(82, "Initializing neural network engine...");
|
|
|
|
|
|
|
| 1349 |
|
| 1350 |
const modelBuffer = await modelBlob.arrayBuffer();
|
| 1351 |
+
await new Promise((resolve, reject) => {
|
| 1352 |
+
modelInitResolve = resolve;
|
| 1353 |
+
modelInitReject = reject;
|
| 1354 |
+
worker.postMessage({
|
| 1355 |
+
action: "init",
|
| 1356 |
+
activeModelName: "model_int8.onnx (v2.4)",
|
| 1357 |
+
vocab: vocab,
|
| 1358 |
+
modelBuffer: modelBuffer
|
| 1359 |
+
}, [modelBuffer]);
|
| 1360 |
});
|
|
|
|
|
|
|
| 1361 |
activeModelName = "model_int8.onnx (v2.4)";
|
| 1362 |
|
| 1363 |
// Step 4. Loading Piano Samples
|
worker.js
ADDED
|
@@ -0,0 +1,469 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
// NanoMaestro Web Worker for isolated model inference and parsing
|
| 2 |
+
importScripts("https://cdn.jsdelivr.net/npm/onnxruntime-web@1.20.1/dist/ort.min.js");
|
| 3 |
+
|
| 4 |
+
// ONNX runtime configuration
|
| 5 |
+
ort.env.wasm.wasmPaths = "https://cdn.jsdelivr.net/npm/onnxruntime-web@1.20.1/dist/";
|
| 6 |
+
ort.env.wasm.numThreads = Math.min(4, navigator.hardwareConcurrency || 1);
|
| 7 |
+
|
| 8 |
+
// State variables
|
| 9 |
+
let session = null;
|
| 10 |
+
let stoi = null;
|
| 11 |
+
let itos = null;
|
| 12 |
+
let vocabSize = 0;
|
| 13 |
+
let activeTokenizer = "abc";
|
| 14 |
+
|
| 15 |
+
let h = null;
|
| 16 |
+
let c = null;
|
| 17 |
+
let currentId = 0;
|
| 18 |
+
let activeModelName = "";
|
| 19 |
+
|
| 20 |
+
// Model parameters
|
| 21 |
+
let hiddenSize = 1024;
|
| 22 |
+
let numLayers = 2;
|
| 23 |
+
|
| 24 |
+
// Playback settings
|
| 25 |
+
let temperature = 0.85;
|
| 26 |
+
let topK = 40;
|
| 27 |
+
let currentBpm = 120;
|
| 28 |
+
|
| 29 |
+
// Seeding settings
|
| 30 |
+
let midiSeedActive = false;
|
| 31 |
+
let midiTokens = [];
|
| 32 |
+
let midiStartBar = 0;
|
| 33 |
+
let isWarmingUp = false;
|
| 34 |
+
|
| 35 |
+
// Temporal queues for gathering parsed events
|
| 36 |
+
let tempQueue = [];
|
| 37 |
+
let parser = null;
|
| 38 |
+
|
| 39 |
+
// MIDI to Pitch Name helpers
|
| 40 |
+
const abcNotePattern = /^(?:\^\^|__|\^|_|=)?[A-Ga-g][,']*$/;
|
| 41 |
+
|
| 42 |
+
function tokenId(token) {
|
| 43 |
+
return stoi?.[token] ?? null;
|
| 44 |
+
}
|
| 45 |
+
|
| 46 |
+
function makeTensorId(id) {
|
| 47 |
+
return new ort.Tensor("int64", BigInt64Array.from([BigInt(id)]), [1, 1]);
|
| 48 |
+
}
|
| 49 |
+
|
| 50 |
+
function zeroState() {
|
| 51 |
+
return new ort.Tensor("float32", new Float32Array(numLayers * hiddenSize), [numLayers, 1, hiddenSize]);
|
| 52 |
+
}
|
| 53 |
+
|
| 54 |
+
function updateModelDimensions(session) {
|
| 55 |
+
try {
|
| 56 |
+
const inputNames = session.inputNames;
|
| 57 |
+
if (inputNames.includes("h") && session.handler && session.handler._model) {
|
| 58 |
+
const inputs = session.handler._model.graph.inputs;
|
| 59 |
+
const hInput = inputs.find(i => i.name === "h");
|
| 60 |
+
if (hInput && hInput.type && hInput.type.tensorType && hInput.type.tensorType.shape) {
|
| 61 |
+
const shape = hInput.type.tensorType.shape.dim;
|
| 62 |
+
const layers = Number(shape[0].dimValue);
|
| 63 |
+
const hidden = Number(shape[2].dimValue);
|
| 64 |
+
if (layers > 0 && hidden > 0) {
|
| 65 |
+
numLayers = layers;
|
| 66 |
+
hiddenSize = hidden;
|
| 67 |
+
console.log(`Worker set model dimensions: layers=${numLayers}, hiddenSize=${hiddenSize}`);
|
| 68 |
+
}
|
| 69 |
+
}
|
| 70 |
+
}
|
| 71 |
+
} catch (e) {
|
| 72 |
+
console.error("Worker failed to detect model dimensions:", e);
|
| 73 |
+
// Fall back to default
|
| 74 |
+
numLayers = 2;
|
| 75 |
+
hiddenSize = 1024;
|
| 76 |
+
}
|
| 77 |
+
}
|
| 78 |
+
|
| 79 |
+
// Model Step Execution
|
| 80 |
+
async function stepModel(record = true) {
|
| 81 |
+
if (!session) return "?";
|
| 82 |
+
const output = await session.run({ input: makeTensorId(currentId), h, c });
|
| 83 |
+
h = output.h_out;
|
| 84 |
+
c = output.c_out;
|
| 85 |
+
currentId = sampleFromLogits(output.logits.data);
|
| 86 |
+
const token = itos[currentId] ?? "?";
|
| 87 |
+
if (record && parser) {
|
| 88 |
+
parser.feed(token);
|
| 89 |
+
}
|
| 90 |
+
return token;
|
| 91 |
+
}
|
| 92 |
+
|
| 93 |
+
function sampleFromLogits(logits) {
|
| 94 |
+
const temp = Math.max(0.05, temperature);
|
| 95 |
+
const k = Math.max(1, topK);
|
| 96 |
+
const scored = [];
|
| 97 |
+
for (let i = 0; i < logits.length; i += 1) {
|
| 98 |
+
const token = itos[i];
|
| 99 |
+
if (token === undefined) continue;
|
| 100 |
+
let score = logits[i] / temp;
|
| 101 |
+
if (token === "<EOP>" || token === "EOS") score -= 1.0;
|
| 102 |
+
scored.push([i, score]);
|
| 103 |
+
}
|
| 104 |
+
scored.sort((a, b) => b[1] - a[1]);
|
| 105 |
+
const picked = scored.slice(0, Math.min(k, scored.length));
|
| 106 |
+
const maxScore = picked[0]?.[1] ?? 0;
|
| 107 |
+
let sum = 0;
|
| 108 |
+
for (const item of picked) {
|
| 109 |
+
item[2] = Math.exp(item[1] - maxScore);
|
| 110 |
+
sum += item[2];
|
| 111 |
+
}
|
| 112 |
+
let r = Math.random() * sum;
|
| 113 |
+
for (const item of picked) {
|
| 114 |
+
r -= item[2];
|
| 115 |
+
if (r <= 0) return item[0];
|
| 116 |
+
}
|
| 117 |
+
return picked[picked.length - 1][0];
|
| 118 |
+
}
|
| 119 |
+
|
| 120 |
+
// Duration converters
|
| 121 |
+
function durationToSecondsFromAbc(duration) {
|
| 122 |
+
let eighths = 1;
|
| 123 |
+
if (duration) {
|
| 124 |
+
if (duration === "/") eighths = 0.5;
|
| 125 |
+
else if (duration.startsWith("/")) eighths = 1 / Number(duration.slice(1));
|
| 126 |
+
else if (duration.includes("/")) {
|
| 127 |
+
const [a, b] = duration.split("/").map(Number);
|
| 128 |
+
eighths = a / b;
|
| 129 |
+
} else {
|
| 130 |
+
eighths = Number(duration);
|
| 131 |
+
}
|
| 132 |
+
}
|
| 133 |
+
const quarterSeconds = 60 / currentBpm;
|
| 134 |
+
return Math.max(0.08, eighths * quarterSeconds * 0.5);
|
| 135 |
+
}
|
| 136 |
+
|
| 137 |
+
function durationToSecondsFromEventSteps(steps, grid) {
|
| 138 |
+
const quarterSeconds = 60 / currentBpm;
|
| 139 |
+
return Math.max(0.035, (Math.max(1, steps) * 4 * quarterSeconds) / Math.max(1, grid));
|
| 140 |
+
}
|
| 141 |
+
|
| 142 |
+
function midiToTonePitch(midi) {
|
| 143 |
+
const names = ["C", "C#", "D", "D#", "E", "F", "F#", "G", "G#", "A", "A#", "B"];
|
| 144 |
+
return `${names[((midi % 12) + 12) % 12]}${Math.floor(midi / 12) - 1}`;
|
| 145 |
+
}
|
| 146 |
+
|
| 147 |
+
function abcToTonePitch(token) {
|
| 148 |
+
const match = token.match(/^(\^\^|__|\^|_|=)?([A-Ga-g])([,']*)$/);
|
| 149 |
+
if (!match) return null;
|
| 150 |
+
const accidental = match[1] || "";
|
| 151 |
+
const step = match[2];
|
| 152 |
+
const marks = match[3] || "";
|
| 153 |
+
let octave = step === step.toLowerCase() ? 5 : 4;
|
| 154 |
+
for (const mark of marks) octave += mark === "'" ? 1 : -1;
|
| 155 |
+
const accidentalText = accidental === "^^" ? "##" : accidental === "__" ? "bb" : accidental === "^" ? "#" : accidental === "_" ? "b" : "";
|
| 156 |
+
return `${step.toUpperCase()}${accidentalText}${octave}`;
|
| 157 |
+
}
|
| 158 |
+
|
| 159 |
+
function pushEvent(event) {
|
| 160 |
+
if (isWarmingUp) return; // Discard prompt history events
|
| 161 |
+
tempQueue.push(event);
|
| 162 |
+
}
|
| 163 |
+
|
| 164 |
+
// Token Slicers
|
| 165 |
+
function getTokensUpToBar(tokens, targetBar) {
|
| 166 |
+
let barCount = 0;
|
| 167 |
+
const sliced = [];
|
| 168 |
+
for (const t of tokens) {
|
| 169 |
+
if (t === "BAR") {
|
| 170 |
+
barCount += 1;
|
| 171 |
+
if (barCount > targetBar) {
|
| 172 |
+
break;
|
| 173 |
+
}
|
| 174 |
+
}
|
| 175 |
+
sliced.push(t);
|
| 176 |
+
}
|
| 177 |
+
return sliced;
|
| 178 |
+
}
|
| 179 |
+
|
| 180 |
+
// ABC Parser
|
| 181 |
+
function makeAbcParser() {
|
| 182 |
+
return {
|
| 183 |
+
pending: null,
|
| 184 |
+
chord: null,
|
| 185 |
+
reset() {
|
| 186 |
+
this.pending = null;
|
| 187 |
+
this.chord = null;
|
| 188 |
+
},
|
| 189 |
+
feed(token) {
|
| 190 |
+
if (token === "Q:") {
|
| 191 |
+
this.pending = { type: "tempoHeader" };
|
| 192 |
+
return;
|
| 193 |
+
}
|
| 194 |
+
if (this.pending?.type === "tempoHeader") {
|
| 195 |
+
const bpm = Number(String(token).split("=").pop());
|
| 196 |
+
if (Number.isFinite(bpm) && bpm >= 40 && bpm <= 220) {
|
| 197 |
+
if (!midiSeedActive || isWarmingUp) {
|
| 198 |
+
currentBpm = bpm;
|
| 199 |
+
self.postMessage({ action: "tempo", bpm: bpm });
|
| 200 |
+
}
|
| 201 |
+
}
|
| 202 |
+
this.pending = null;
|
| 203 |
+
return;
|
| 204 |
+
}
|
| 205 |
+
if (token === "[") {
|
| 206 |
+
this.chord = [];
|
| 207 |
+
this.pending = null;
|
| 208 |
+
return;
|
| 209 |
+
}
|
| 210 |
+
if (this.chord) {
|
| 211 |
+
if (token === "]") {
|
| 212 |
+
this.pending = { type: "chord", notes: this.chord.map(abcToTonePitch).filter(Boolean) };
|
| 213 |
+
this.chord = null;
|
| 214 |
+
return;
|
| 215 |
+
}
|
| 216 |
+
if (abcNotePattern.test(token)) this.chord.push(token);
|
| 217 |
+
return;
|
| 218 |
+
}
|
| 219 |
+
if (token === "z") {
|
| 220 |
+
this.pending = { type: "rest" };
|
| 221 |
+
return;
|
| 222 |
+
}
|
| 223 |
+
if (abcNotePattern.test(token)) {
|
| 224 |
+
this.pending = { type: "note", note: abcToTonePitch(token) };
|
| 225 |
+
return;
|
| 226 |
+
}
|
| 227 |
+
if (token.startsWith("DUR:") && this.pending) {
|
| 228 |
+
this.emitPending(token.slice(4));
|
| 229 |
+
return;
|
| 230 |
+
}
|
| 231 |
+
if (this.pending && (token === "|" || token === "<NL>" || token === "<EOP>")) this.emitPending(null);
|
| 232 |
+
},
|
| 233 |
+
emitPending(durationToken) {
|
| 234 |
+
const duration = durationToSecondsFromAbc(durationToken);
|
| 235 |
+
if (this.pending.type === "rest") pushEvent({ type: "rest", duration });
|
| 236 |
+
if (this.pending.type === "note" && this.pending.note) pushEvent({ type: "note", notes: [this.pending.note], duration, advance: duration });
|
| 237 |
+
if (this.pending.type === "chord" && this.pending.notes.length) pushEvent({ type: "note", notes: this.pending.notes.slice(0, 8), duration, advance: duration });
|
| 238 |
+
this.pending = null;
|
| 239 |
+
},
|
| 240 |
+
};
|
| 241 |
+
}
|
| 242 |
+
|
| 243 |
+
// Event Parser
|
| 244 |
+
function makeEventParser() {
|
| 245 |
+
return {
|
| 246 |
+
grid: 64,
|
| 247 |
+
bar: -1,
|
| 248 |
+
pos: 0,
|
| 249 |
+
pendingPosition: null,
|
| 250 |
+
pendingNotes: [],
|
| 251 |
+
pendingNote: null,
|
| 252 |
+
lastQ: 0,
|
| 253 |
+
reset() {
|
| 254 |
+
this.grid = 64;
|
| 255 |
+
this.bar = -1;
|
| 256 |
+
this.pos = 0;
|
| 257 |
+
this.pendingPosition = null;
|
| 258 |
+
this.pendingNotes = [];
|
| 259 |
+
this.pendingNote = null;
|
| 260 |
+
this.lastQ = 0;
|
| 261 |
+
},
|
| 262 |
+
feed(token) {
|
| 263 |
+
if (token.startsWith("BPM_")) {
|
| 264 |
+
const bpm = Number(token.slice(4));
|
| 265 |
+
if (Number.isFinite(bpm) && bpm >= 40 && bpm <= 220) {
|
| 266 |
+
if (!midiSeedActive || isWarmingUp) {
|
| 267 |
+
currentBpm = bpm;
|
| 268 |
+
self.postMessage({ action: "tempo", bpm: bpm });
|
| 269 |
+
}
|
| 270 |
+
}
|
| 271 |
+
return;
|
| 272 |
+
}
|
| 273 |
+
if (token.startsWith("GRID_")) {
|
| 274 |
+
const grid = Number(token.slice(5));
|
| 275 |
+
if (Number.isFinite(grid) && grid > 0) this.grid = grid;
|
| 276 |
+
return;
|
| 277 |
+
}
|
| 278 |
+
if (token === "BAR") {
|
| 279 |
+
this.flushTo(this.absoluteQFor(this.bar + 1, 0));
|
| 280 |
+
this.bar += 1;
|
| 281 |
+
this.pos = 0;
|
| 282 |
+
return;
|
| 283 |
+
}
|
| 284 |
+
if (token.startsWith("POS_")) {
|
| 285 |
+
const pos = Number(token.slice(4));
|
| 286 |
+
if (!Number.isFinite(pos)) return;
|
| 287 |
+
this.flushTo(this.absoluteQFor(this.bar, pos));
|
| 288 |
+
this.pos = pos;
|
| 289 |
+
return;
|
| 290 |
+
}
|
| 291 |
+
if (token.startsWith("NOTE_")) {
|
| 292 |
+
const midi = Number(token.slice(5));
|
| 293 |
+
if (Number.isFinite(midi) && midi >= 0 && midi <= 127) {
|
| 294 |
+
this.pendingNote = { midi, durationSteps: 1, velocity: 0.72 };
|
| 295 |
+
}
|
| 296 |
+
return;
|
| 297 |
+
}
|
| 298 |
+
if (token.startsWith("DUR_") && this.pendingNote) {
|
| 299 |
+
const steps = Number(token.slice(4));
|
| 300 |
+
if (Number.isFinite(steps)) this.pendingNote.durationSteps = Math.max(1, steps);
|
| 301 |
+
return;
|
| 302 |
+
}
|
| 303 |
+
if (token.startsWith("VEL_") && this.pendingNote) {
|
| 304 |
+
const bucket = Number(token.slice(4));
|
| 305 |
+
if (Number.isFinite(bucket)) this.pendingNote.velocity = Math.max(0.2, Math.min(0.95, bucket / 8));
|
| 306 |
+
const q = this.absoluteQFor(this.bar, this.pos);
|
| 307 |
+
this.pendingPosition ??= q;
|
| 308 |
+
this.pendingNotes.push(this.pendingNote);
|
| 309 |
+
this.pendingNote = null;
|
| 310 |
+
}
|
| 311 |
+
},
|
| 312 |
+
absoluteQFor(bar, pos) {
|
| 313 |
+
return Math.max(0, bar) * 4 + (Math.max(0, pos) * 4) / Math.max(1, this.grid);
|
| 314 |
+
},
|
| 315 |
+
flushTo(nextQ) {
|
| 316 |
+
if (this.pendingNote) {
|
| 317 |
+
const q = this.absoluteQFor(this.bar, this.pos);
|
| 318 |
+
this.pendingPosition ??= q;
|
| 319 |
+
this.pendingNotes.push(this.pendingNote);
|
| 320 |
+
this.pendingNote = null;
|
| 321 |
+
}
|
| 322 |
+
if (this.pendingNotes.length && this.pendingPosition !== null) {
|
| 323 |
+
const gap = Math.max(0, this.pendingPosition - this.lastQ);
|
| 324 |
+
if (gap > 0) pushEvent({ type: "rest", duration: this.quartersToSeconds(gap) });
|
| 325 |
+
pushEvent({
|
| 326 |
+
type: "note",
|
| 327 |
+
notes: this.pendingNotes.map((note) => midiToTonePitch(note.midi)),
|
| 328 |
+
perNoteDurations: this.pendingNotes.map((note) => durationToSecondsFromEventSteps(note.durationSteps, this.grid)),
|
| 329 |
+
velocities: this.pendingNotes.map((note) => note.velocity),
|
| 330 |
+
duration: 0,
|
| 331 |
+
advance: 0,
|
| 332 |
+
});
|
| 333 |
+
this.lastQ = this.pendingPosition;
|
| 334 |
+
}
|
| 335 |
+
const finalGap = Math.max(0, nextQ - this.lastQ);
|
| 336 |
+
if (finalGap > 0) pushEvent({ type: "rest", duration: this.quartersToSeconds(finalGap) });
|
| 337 |
+
this.lastQ = Math.max(this.lastQ, nextQ);
|
| 338 |
+
this.pendingPosition = null;
|
| 339 |
+
this.pendingNotes = [];
|
| 340 |
+
},
|
| 341 |
+
quartersToSeconds(quarters) {
|
| 342 |
+
return (quarters * 60) / currentBpm;
|
| 343 |
+
},
|
| 344 |
+
};
|
| 345 |
+
}
|
| 346 |
+
|
| 347 |
+
function makeParser() {
|
| 348 |
+
return activeTokenizer === "giantmidi_event" ? makeEventParser() : makeAbcParser();
|
| 349 |
+
}
|
| 350 |
+
|
| 351 |
+
async function warmPrompt() {
|
| 352 |
+
h = zeroState();
|
| 353 |
+
c = zeroState();
|
| 354 |
+
|
| 355 |
+
const allTokensStr = midiSeedActive && midiTokens.length > 0
|
| 356 |
+
? getTokensUpToBar(midiTokens, midiStartBar)
|
| 357 |
+
: (activeTokenizer === "giantmidi_event"
|
| 358 |
+
? ["BOS", "BPM_120", "GRID_64", "BAR", "POS_0"]
|
| 359 |
+
: ["X:", "1", "<NL>", "T:", "piece", "<NL>", "M:", "4/4", "<NL>", "L:", "1/8", "<NL>", "Q:", "1/4=120", "<NL>", "K:", "C", "<NL>"]);
|
| 360 |
+
|
| 361 |
+
isWarmingUp = true;
|
| 362 |
+
|
| 363 |
+
// 1. Feed all prompt tokens to parser to advance its clock
|
| 364 |
+
if (parser) {
|
| 365 |
+
for (const token of allTokensStr) {
|
| 366 |
+
parser.feed(token);
|
| 367 |
+
}
|
| 368 |
+
}
|
| 369 |
+
|
| 370 |
+
// 2. Slice model warm-up tokens to last 256 (matching model's context window)
|
| 371 |
+
const modelTokensStr = allTokensStr.slice(-256);
|
| 372 |
+
const ids = modelTokensStr.map(tokenId).filter((id) => id !== null);
|
| 373 |
+
|
| 374 |
+
for (const id of ids) {
|
| 375 |
+
currentId = id;
|
| 376 |
+
await stepModel(false);
|
| 377 |
+
}
|
| 378 |
+
|
| 379 |
+
isWarmingUp = false;
|
| 380 |
+
}
|
| 381 |
+
|
| 382 |
+
async function pumpTokens(targetCount = 96) {
|
| 383 |
+
tempQueue = [];
|
| 384 |
+
let steps = 0;
|
| 385 |
+
// Keep stepping the model until we collect enough parsed events or hit loop bounds
|
| 386 |
+
while (tempQueue.length < targetCount && steps < 300) {
|
| 387 |
+
await stepModel(true);
|
| 388 |
+
steps += 1;
|
| 389 |
+
}
|
| 390 |
+
return {
|
| 391 |
+
events: tempQueue,
|
| 392 |
+
lastToken: itos[currentId] ?? "?"
|
| 393 |
+
};
|
| 394 |
+
}
|
| 395 |
+
|
| 396 |
+
// Message Router
|
| 397 |
+
self.onmessage = async function (e) {
|
| 398 |
+
const data = e.data;
|
| 399 |
+
|
| 400 |
+
switch (data.action) {
|
| 401 |
+
case "init":
|
| 402 |
+
try {
|
| 403 |
+
console.log(`Worker loading model: ${data.activeModelName}`);
|
| 404 |
+
activeModelName = data.activeModelName;
|
| 405 |
+
stoi = data.vocab.stoi;
|
| 406 |
+
itos = Object.fromEntries(Object.entries(data.vocab.itos).map(([key, value]) => [Number(key), value]));
|
| 407 |
+
vocabSize = data.vocab.vocab_size;
|
| 408 |
+
activeTokenizer = data.vocab.tokenizer || "abc";
|
| 409 |
+
|
| 410 |
+
session = await ort.InferenceSession.create(data.modelBuffer, {
|
| 411 |
+
executionProviders: ["wasm"],
|
| 412 |
+
graphOptimizationLevel: "all",
|
| 413 |
+
});
|
| 414 |
+
|
| 415 |
+
updateModelDimensions(session);
|
| 416 |
+
self.postMessage({ action: "initialized" });
|
| 417 |
+
} catch (err) {
|
| 418 |
+
console.error("Worker initialization failed:", err);
|
| 419 |
+
self.postMessage({ action: "error", message: `Init failed: ${err.message}` });
|
| 420 |
+
}
|
| 421 |
+
break;
|
| 422 |
+
|
| 423 |
+
case "start":
|
| 424 |
+
try {
|
| 425 |
+
temperature = data.temperature;
|
| 426 |
+
topK = data.topK;
|
| 427 |
+
currentBpm = data.bpm;
|
| 428 |
+
midiSeedActive = data.midiSeedActive;
|
| 429 |
+
midiTokens = data.midiTokens;
|
| 430 |
+
midiStartBar = data.midiStartBar;
|
| 431 |
+
|
| 432 |
+
parser = makeParser();
|
| 433 |
+
parser.reset();
|
| 434 |
+
|
| 435 |
+
await warmPrompt();
|
| 436 |
+
|
| 437 |
+
const initialBatch = await pumpTokens(96);
|
| 438 |
+
self.postMessage({
|
| 439 |
+
action: "started",
|
| 440 |
+
events: initialBatch.events,
|
| 441 |
+
lastToken: initialBatch.lastToken
|
| 442 |
+
});
|
| 443 |
+
} catch (err) {
|
| 444 |
+
console.error("Worker start failed:", err);
|
| 445 |
+
self.postMessage({ action: "error", message: `Start failed: ${err.message}` });
|
| 446 |
+
}
|
| 447 |
+
break;
|
| 448 |
+
|
| 449 |
+
case "pump":
|
| 450 |
+
try {
|
| 451 |
+
temperature = data.temperature;
|
| 452 |
+
topK = data.topK;
|
| 453 |
+
const batch = await pumpTokens(96);
|
| 454 |
+
self.postMessage({
|
| 455 |
+
action: "events",
|
| 456 |
+
events: batch.events,
|
| 457 |
+
lastToken: batch.lastToken
|
| 458 |
+
});
|
| 459 |
+
} catch (err) {
|
| 460 |
+
console.error("Worker pump failed:", err);
|
| 461 |
+
self.postMessage({ action: "error", message: `Pump failed: ${err.message}` });
|
| 462 |
+
}
|
| 463 |
+
break;
|
| 464 |
+
|
| 465 |
+
case "stop":
|
| 466 |
+
// Halts ongoing actions, resets states if needed
|
| 467 |
+
break;
|
| 468 |
+
}
|
| 469 |
+
};
|