mirror of
https://github.com/LostRuins/koboldcpp.git
synced 2026-08-15 11:25:12 +00:00
updated lite (+4 squashed commit)
Squashed commit: [ecaa653df] updated lite [1154c44a7] updated lite [db2e5e43d] allow whisper interrogate mode for audio files [490b13af8] whitespace
This commit is contained in:
Concedo
parent
b0b7a07b34
commit
e1849f9a0b
2 changed files with 84 additions and 25 deletions
|
|
@ -199,11 +199,7 @@
|
|||
"assistant_start": "ASSISTANT: ",
|
||||
"assistant_end": "</s>\n"
|
||||
}
|
||||
},
|
||||
|
||||
|
||||
|
||||
{
|
||||
}, {
|
||||
"search": ["[/INST]"],
|
||||
"name": "Mistral (Generic)",
|
||||
"adapter": {
|
||||
|
|
|
|||
103
klite.embd
103
klite.embd
|
|
@ -3459,6 +3459,8 @@ Current version indicated by LITEVER below.
|
|||
};
|
||||
|
||||
const defaultsettings = JSON.parse(JSON.stringify(localsettings));
|
||||
//visionmode 0=disabled, 1=hordeinterrogate, 2=localinterrogate, 3=multimodal
|
||||
//type 0=img, 1=audio
|
||||
const default_imgs_meta = {prompt:"", desc:"", visionmode:0, aspect:0, ref:"", len:0, type:0, data: ""};
|
||||
|
||||
//a list of presets users can choose from
|
||||
|
|
@ -14725,7 +14727,7 @@ Current version indicated by LITEVER below.
|
|||
let userinput = getInputBoxValue();
|
||||
if(userinput.trim()!="")
|
||||
{
|
||||
var sentence = userinput.trim().substring(0, 380);
|
||||
var sentence = userinput.trim().substring(0, 540);
|
||||
do_manual_gen_image(sentence);
|
||||
}
|
||||
},false);
|
||||
|
|
@ -16554,7 +16556,7 @@ Current version indicated by LITEVER below.
|
|||
{
|
||||
let txt = "I'll try and create that image.";
|
||||
gametext_arr.push(txt);
|
||||
var sentence = img_gen_trigger_prompt.trim().substring(0, 380);
|
||||
var sentence = img_gen_trigger_prompt.trim().substring(0, 540);
|
||||
do_manual_gen_image(sentence);
|
||||
}
|
||||
|
||||
|
|
@ -17916,11 +17918,18 @@ Current version indicated by LITEVER below.
|
|||
}
|
||||
|
||||
}
|
||||
function zoomed_transcribe_btn(audiohash,onDone)
|
||||
function zoomed_transcribe_btn(audiohash)
|
||||
{
|
||||
transcribe_audio_file(audiohash,(txt)=>{
|
||||
msgbox(txt,"Transcribed Audio");
|
||||
});
|
||||
}
|
||||
function transcribe_audio_file(audiohash,onDone)
|
||||
{
|
||||
let fetchedblob = data_hash_to_blob_lookup[audiohash];
|
||||
if(!fetchedblob)
|
||||
{
|
||||
onDone(null);
|
||||
return;
|
||||
}
|
||||
fetch(fetchedblob.blob)
|
||||
|
|
@ -17950,10 +17959,14 @@ Current version indicated by LITEVER below.
|
|||
console.log(resp);
|
||||
if(resp && resp.text && resp.text!="")
|
||||
{
|
||||
msgbox(resp.text,"Transcribed Audio");
|
||||
onDone(resp.text);
|
||||
}else
|
||||
{
|
||||
onDone(null);
|
||||
}
|
||||
}).catch((error) => {
|
||||
console.log("Transcribe Error: " + error);
|
||||
onDone(null);
|
||||
});
|
||||
}
|
||||
reader.readAsDataURL(wavblob);
|
||||
|
|
@ -17968,14 +17981,34 @@ Current version indicated by LITEVER below.
|
|||
if(savedmeta)
|
||||
{
|
||||
savedmeta.visionmode = document.getElementById("aivisionmode").value;
|
||||
if(!savedmeta.desc && (savedmeta.visionmode==1 || savedmeta.visionmode==2))
|
||||
if(savedmeta.type==1) //audio
|
||||
{
|
||||
//request a new interrogation
|
||||
var alreadysent = Object.values(interrogation_db).some(item => item.imghash === imghash);
|
||||
if(!alreadysent && document.getElementById("zoomedimg"))
|
||||
if(!savedmeta.desc && savedmeta.visionmode==2)
|
||||
{
|
||||
let b64 = document.getElementById("zoomedimg").src;
|
||||
interrogate_new_image(b64,imghash,(savedmeta.visionmode==1));
|
||||
var alreadysent = Object.values(interrogation_db).some(item => item.imghash === imghash);
|
||||
if(!alreadysent && document.getElementById("zoomedaudio"))
|
||||
{
|
||||
transcribe_audio_file(imghash,(txt)=>{
|
||||
if(txt)
|
||||
{
|
||||
savedmeta.desc = txt;
|
||||
update_clicked_image(imghash);
|
||||
}
|
||||
});
|
||||
}
|
||||
}
|
||||
}
|
||||
else //images
|
||||
{
|
||||
if(!savedmeta.desc && (savedmeta.visionmode==1 || savedmeta.visionmode==2))
|
||||
{
|
||||
//request a new interrogation
|
||||
var alreadysent = Object.values(interrogation_db).some(item => item.imghash === imghash);
|
||||
if(!alreadysent && document.getElementById("zoomedimg"))
|
||||
{
|
||||
let b64 = document.getElementById("zoomedimg").src;
|
||||
interrogate_new_image(b64,imghash,(savedmeta.visionmode==1));
|
||||
}
|
||||
}
|
||||
}
|
||||
update_clicked_image(imghash);
|
||||
|
|
@ -18066,6 +18099,7 @@ Current version indicated by LITEVER below.
|
|||
}
|
||||
let togglebtn = `<select class="form-control" id="aivisionmode" style="display:inline;height:24px;width: 140px; padding: 2px; margin: 3px; font-size:12px;" onchange="toggle_ai_vision(\'`+imghash+`\')">
|
||||
<option value="0">Disabled</option>
|
||||
<option value="2">Transcribe (Local)</option>
|
||||
<option value="3">Multimodal Audio</option>
|
||||
</select>`;
|
||||
document.getElementById("zoomedimgdesc").innerHTML = `
|
||||
|
|
@ -18093,7 +18127,7 @@ Current version indicated by LITEVER below.
|
|||
let userinput = getInputBoxValue();
|
||||
if(userinput.trim()!="" && document.getElementById("zoomedimg"))
|
||||
{
|
||||
var sentence = userinput.trim().substring(0, 380);
|
||||
var sentence = userinput.trim().substring(0, 540);
|
||||
let b64 = document.getElementById("zoomedimg").src;
|
||||
do_manual_gen_image(sentence, b64);
|
||||
document.getElementById("zoomedimgcontainer").classList.add("hidden");
|
||||
|
|
@ -19059,9 +19093,23 @@ Current version indicated by LITEVER below.
|
|||
reader.onloadend = function () {
|
||||
const arrayBuffer = reader.result;
|
||||
const audioContext = new (window.AudioContext || window.webkitAudioContext)();
|
||||
audioContext.decodeAudioData(arrayBuffer, function (buffer) {
|
||||
audioContext.decodeAudioData(arrayBuffer, function (obuf) {
|
||||
resampleAudioBuffer(obuf,22050,(buffer)=>{
|
||||
const samplefreq = buffer.sampleRate;
|
||||
const samples = buffer.getChannelData(0); // mono
|
||||
const numberOfChannels = buffer.numberOfChannels;
|
||||
const length = buffer.length;
|
||||
//first, mix all down into mono
|
||||
let samples = new Float32Array(length);
|
||||
for (let channel = 0; channel < numberOfChannels; channel++) {
|
||||
const channelData = buffer.getChannelData(channel);
|
||||
for (let i = 0; i < length; i++) {
|
||||
samples[i] += channelData[i];
|
||||
}
|
||||
}
|
||||
for (let i = 0; i < length; i++) {
|
||||
samples[i] /= numberOfChannels;
|
||||
}
|
||||
|
||||
const durationInSeconds = buffer.duration;
|
||||
const mp3encoder = new lamejs.Mp3Encoder(1, samplefreq, 40); // mono, 16kHz, 40kbps
|
||||
const sampleBlockSize = 1152; //can be anything but make it a multiple of 576 to make encoders life easier
|
||||
|
|
@ -19069,13 +19117,14 @@ Current version indicated by LITEVER below.
|
|||
for (let i = 0; i < samples.length; i += sampleBlockSize) {
|
||||
let sampleChunk = samples.subarray(i, i + sampleBlockSize);
|
||||
|
||||
// Convert Float32 [-1, 1] to Int16 [-32768, 32767]
|
||||
let int16Buffer = new Int16Array(sampleChunk.length);
|
||||
let buff = new Float32Array(sampleChunk.length);
|
||||
for (let j = 0; j < sampleChunk.length; j++) {
|
||||
let s = Math.max(-1, Math.min(1, sampleChunk[j]));
|
||||
int16Buffer[j] = s < 0 ? s * 0x8000 : s * 0x7FFF;
|
||||
let s = sampleChunk[j];
|
||||
s = s*32767.0;
|
||||
s = Math.max(-32767.0, Math.min(32767.0, s));
|
||||
buff[j] = s;
|
||||
}
|
||||
const mp3buf = mp3encoder.encodeBuffer(int16Buffer);
|
||||
const mp3buf = mp3encoder.encodeBuffer(buff);
|
||||
if (mp3buf.length > 0) {
|
||||
mp3Data.push(new Uint8Array(mp3buf));
|
||||
}
|
||||
|
|
@ -19091,6 +19140,7 @@ Current version indicated by LITEVER below.
|
|||
onDone(mp3Base64,durationInSeconds);
|
||||
};
|
||||
fileReader.readAsDataURL(mp3Blob);
|
||||
});
|
||||
}, function (err) {
|
||||
console.log("Audio decode failed.");
|
||||
onDone(null,null);
|
||||
|
|
@ -20000,7 +20050,12 @@ Current version indicated by LITEVER below.
|
|||
let data = foundmeta.data;
|
||||
if(foundmeta.desc && (foundmeta.visionmode==1||foundmeta.visionmode==2))
|
||||
{
|
||||
return "\n(Attached Image: " + foundmeta.desc + ")\n";
|
||||
if(foundmeta.type==1)//audio
|
||||
{
|
||||
return "\n(Attached Audio: " + foundmeta.desc + ")\n";
|
||||
}else{
|
||||
return "\n(Attached Image: " + foundmeta.desc + ")\n";
|
||||
}
|
||||
}
|
||||
else if(foundmeta.visionmode==3)
|
||||
{
|
||||
|
|
@ -21315,7 +21370,15 @@ Current version indicated by LITEVER below.
|
|||
for(var i=0;i<chatunits.length;++i)
|
||||
{
|
||||
let curr = chatunits[i];
|
||||
curr = repack_postprocess_turn(curr, countmap);
|
||||
if(corpo_editing_turn != i)
|
||||
{
|
||||
curr = repack_postprocess_turn(curr, countmap); //change whole object
|
||||
}
|
||||
else
|
||||
{
|
||||
let reshaped = repack_postprocess_turn(JSON.parse(JSON.stringify(curr)), countmap);
|
||||
curr.name = reshaped.name; //only change name
|
||||
}
|
||||
|
||||
let resendbtn = ((curr.myturn && i<chatunits.length-1)?`<button type="button" class="btn btn-primary corpoeditbtn" style="float:right;" onclick="corpo_edit_chunk_resend(${i})">Resend</button>`:``);
|
||||
let bodypart = (corpo_editing_turn == i ?
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue