cturan
/
llama.cpp
mirror of https://github.com/cturan/llama.cpp


			
							123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449
							<html>

<head>
  <meta charset="UTF-8">
  <meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=1" />
  <title>llama.cpp - chat</title>

  <style>
    body {
      background-color: #fff;
      color: #000;
      font-family: system-ui;
      font-size: 90%;
    }

    #container {
      margin: 0em auto;
      display: flex;
      flex-direction: column;
      justify-content: space-between;
      height: 100%;
    }

    main {
      margin: 3px;
      display: flex;
      flex-direction: column;
      justify-content: space-between;
      gap: 1em;

      flex-grow: 1;
      overflow-y: auto;

      border: 1px solid #ccc;
      border-radius: 5px;
      padding: 0.5em;
    }

    body {
      max-width: 600px;
      min-width: 300px;
      line-height: 1.2;
      margin: 0 auto;
      padding: 0 0.5em;
    }

    p {
      overflow-wrap: break-word;
      word-wrap: break-word;
      hyphens: auto;
      margin-top: 0.5em;
      margin-bottom: 0.5em;
    }

    #write form {
      margin: 1em 0 0 0;
      display: flex;
      flex-direction: column;
      gap: 0.5em;
      align-items: stretch;
    }

    .right {
      display: flex;
      flex-direction: row;
      gap: 0.5em;
      justify-content: flex-end;
    }

    fieldset {
      border: none;
      padding: 0;
      margin: 0;
    }

    fieldset.two {
      display: grid;
      grid-template: "a a";
      gap: 1em;
    }

    fieldset.three {
      display: grid;
      grid-template: "a a a";
      gap: 1em;
    }

    details {
      border: 1px solid #aaa;
      border-radius: 4px;
      padding: 0.5em 0.5em 0;
      margin-top: 0.5em;
    }

    summary {
      font-weight: bold;
      margin: -0.5em -0.5em 0;
      padding: 0.5em;
      cursor: pointer;
    }

    details[open] {
      padding: 0.5em;
    }


    textarea {
      padding: 5px;
      flex-grow: 1;
      width: 100%;
    }

    pre code {
      display: block;
      background-color: #222;
      color: #ddd;
    }
    code {
      font-family: monospace;
      padding: 0.1em 0.3em;
      border-radius: 3px;
    }

    fieldset label {
      margin: 0.5em 0;
      display: block;
    }

    header, footer {
      text-align: center;
    }

    footer {
      font-size: 80%;
      color: #888;
    }
  </style>

  <script type="module">
    import {
      html, h, signal, effect, computed, render, useSignal, useEffect, useRef
    } from '/index.js';

    import { llama } from '/completion.js';

    const session = signal({
      prompt: "This is a conversation between user and llama, a friendly chatbot. respond in simple markdown.",
      template: "{{prompt}}\n\n{{history}}\n{{char}}:",
      historyTemplate: "{{name}}: {{message}}",
      transcript: [],
      type: "chat",
      char: "llama",
      user: "User",
    })

    const params = signal({
      n_predict: 400,
      temperature: 0.7,
      repeat_last_n: 256, // 0 = disable penalty, -1 = context size
      repeat_penalty: 1.18, // 1.0 = disabled
      top_k: 40, // <= 0 to use vocab size
      top_p: 0.5, // 1.0 = disabled
      tfs_z: 1.0, // 1.0 = disabled
      typical_p: 1.0, // 1.0 = disabled
      presence_penalty: 0.0, // 0.0 = disabled
      frequency_penalty: 0.0, // 0.0 = disabled
      mirostat: 0, // 0/1/2
      mirostat_tau: 5, // target entropy
      mirostat_eta: 0.1, // learning rate
    })

    const llamaStats = signal(null)
    const controller = signal(null)

    const generating = computed(() => controller.value == null )
    const chatStarted = computed(() => session.value.transcript.length > 0)

    const transcriptUpdate = (transcript) => {
      session.value = {
        ...session.value,
        transcript
      }
    }

    // simple template replace
    const template = (str, extraSettings) => {
      let settings = session.value;
      if (extraSettings) {
        settings = { ...settings, ...extraSettings };
      }
      return String(str).replaceAll(/\{\{(.*?)\}\}/g, (_, key) => template(settings[key]));
    }

    // send message to server
    const chat = async (msg) => {
      if (controller.value) {
        console.log('already running...');
        return;
      }
      controller.value = new AbortController();

      transcriptUpdate([...session.value.transcript, ["{{user}}", msg]])

      const prompt = template(session.value.template, {
        message: msg,
        history: session.value.transcript.flatMap(([name, message]) => template(session.value.historyTemplate, {name, message})).join("\n"),
      });

      let currentMessage = '';
      const history = session.value.transcript

      const llamaParams = {
        ...params.value,
        stop: ["</s>", template("{{char}}:"), template("{{user}}:")],
      }

      for await (const chunk of llama(prompt, llamaParams, { controller: controller.value })) {
        const data = chunk.data;
        currentMessage += data.content;

        // remove leading whitespace
        currentMessage = currentMessage.replace(/^\s+/, "")

        transcriptUpdate([...history, ["{{char}}", currentMessage]])

        if (data.stop) {
          console.log("Completion finished: '", currentMessage, "', summary: ", data);
        }

        if (data.timings) {
          llamaStats.value = data.timings;
        }
      }

      controller.value = null;
    }

    function MessageInput() {
      const message = useSignal("")

      const stop = (e) => {
        e.preventDefault();
        if (controller.value) {
          controller.value.abort();
          controller.value = null;
        }
      }

      const reset = (e) => {
        stop(e);
        transcriptUpdate([]);
      }

      const submit = (e) => {
        stop(e);
        chat(message.value);
        message.value = "";
      }

      const enterSubmits = (event) => {
        if (event.which === 13 && !event.shiftKey) {
          submit(event);
        }
      }

      return html`
        <form onsubmit=${submit}>
          <div>
            <textarea type="text" rows=2 onkeypress=${enterSubmits} value="${message}" oninput=${(e) => message.value = e.target.value} placeholder="Say something..."/>
          </div>
          <div class="right">
            <button type="submit" disabled=${!generating.value} >Send</button>
            <button onclick=${stop} disabled=${generating}>Stop</button>
            <button onclick=${reset}>Reset</button>
          </div>
        </form>
      `
    }

    const ChatLog = (props) => {
      const messages = session.value.transcript;
      const container = useRef(null)

      useEffect(() => {
        // scroll to bottom (if needed)
        if (container.current && container.current.scrollHeight <= container.current.scrollTop + container.current.offsetHeight + 300) {
          container.current.scrollTo(0, container.current.scrollHeight)
        }
      }, [messages])

      const chatLine = ([user, msg]) => {
        return html`<p key=${msg}><strong>${template(user)}:</strong> <${Markdownish} text=${template(msg)} /></p>`
      };

      return html`
        <section id="chat" ref=${container}>
          ${messages.flatMap(chatLine)}
        </section>`;
    };

    const ConfigForm = (props) => {
      const updateSession = (el) => session.value = { ...session.value, [el.target.name]: el.target.value }
      const updateParams = (el) => params.value = { ...params.value, [el.target.name]: el.target.value }
      const updateParamsFloat = (el) => params.value = { ...params.value, [el.target.name]: parseFloat(el.target.value) }
      const updateParamsInt = (el) => params.value = { ...params.value, [el.target.name]: Math.floor(parseFloat(el.target.value)) }

      const FloatField = ({label, max, min, name, step, value}) => {
        return html`
          <div>
            <label for="${name}">${label}</label>
            <input type="range" id="${name}" min="${min}" max="${max}" step="${step}" name="${name}" value="${value}" oninput=${updateParamsFloat} />
            <span>${value}</span>
          </div>
        `
      };

      const IntField = ({label, max, min, name, value}) => {
        return html`
          <div>
            <label for="${name}">${label}</label>
            <input type="range" id="${name}" min="${min}" max="${max}" name="${name}" value="${value}" oninput=${updateParamsInt} />
            <span>${value}</span>
          </div>
        `
      };

      return html`
        <form>
          <fieldset>
            <div>
              <label for="prompt">Prompt</label>
              <textarea type="text" name="prompt" value="${session.value.prompt}" rows=4 oninput=${updateSession}/>
            </div>
          </fieldset>

          <fieldset class="two">
            <div>
              <label for="user">User name</label>
              <input type="text" name="user" value="${session.value.user}" oninput=${updateSession} />
            </div>

            <div>
              <label for="bot">Bot name</label>
              <input type="text" name="char" value="${session.value.char}" oninput=${updateSession} />
            </div>
          </fieldset>

          <fieldset>
            <div>
              <label for="template">Prompt template</label>
              <textarea id="template" name="template" value="${session.value.template}" rows=4 oninput=${updateSession}/>
            </div>

            <div>
              <label for="template">Chat history template</label>
              <textarea id="template" name="historyTemplate" value="${session.value.historyTemplate}" rows=1 oninput=${updateSession}/>
            </div>
          </fieldset>

          <fieldset class="two">
            ${IntField({label: "Predictions", max: 2048, min: -1, name: "n_predict", value: params.value.n_predict})}
            ${FloatField({label: "Temperature", max: 1.5, min: 0.0, name: "temperature", step: 0.01, value: params.value.temperature})}
            ${FloatField({label: "Penalize repeat sequence", max: 2.0, min: 0.0, name: "repeat_penalty", step: 0.01, value: params.value.repeat_penalty})}
            ${IntField({label: "Consider N tokens for penalize", max: 2048, min: 0, name: "repeat_last_n", value: params.value.repeat_last_n})}
            ${IntField({label: "Top-K sampling", max: 100, min: -1, name: "top_k", value: params.value.top_k})}
            ${FloatField({label: "Top-P sampling", max: 1.0, min: 0.0, name: "top_p", step: 0.01, value: params.value.top_p})}
          </fieldset>
          <details>
            <summary>More options</summary>
            <fieldset class="two">
              ${FloatField({label: "TFS-Z", max: 1.0, min: 0.0, name: "tfs_z", step: 0.01, value: params.value.tfs_z})}
              ${FloatField({label: "Typical P", max: 1.0, min: 0.0, name: "typical_p", step: 0.01, value: params.value.typical_p})}
              ${FloatField({label: "Presence penalty", max: 1.0, min: 0.0, name: "presence_penalty", step: 0.01, value: params.value.presence_penalty})}
              ${FloatField({label: "Frequency penalty", max: 1.0, min: 0.0, name: "frequency_penalty", step: 0.01, value: params.value.frequency_penalty})}
            </fieldset>
            <hr />
            <fieldset class="three">
              <div>
                <label><input type="radio" name="mirostat" value="0" checked=${params.value.mirostat == 0} oninput=${updateParamsInt} /> no Mirostat</label>
                <label><input type="radio" name="mirostat" value="1" checked=${params.value.mirostat == 1} oninput=${updateParamsInt} /> Mirostat v1</label>
                <label><input type="radio" name="mirostat" value="2" checked=${params.value.mirostat == 2} oninput=${updateParamsInt} /> Mirostat v2</label>
              </div>
              ${FloatField({label: "Mirostat tau", max: 10.0, min: 0.0, name: "mirostat_tau", step: 0.01, value: params.value.mirostat_tau})}
              ${FloatField({label: "Mirostat eta", max: 1.0, min: 0.0, name: "mirostat_eta", step: 0.01, value: params.value.mirostat_eta})}
            </fieldset>
          </details>
        </form>
      `
    }
    // poor mans markdown replacement
    const Markdownish = (params) => {
      const md = params.text
        .replace(/&/g, '&amp;')
        .replace(/</g, '&lt;')
        .replace(/>/g, '&gt;')
        .replace(/^#{1,6} (.*)$/gim, '<h3>$1</h3>')
        .replace(/\*\*(.*?)\*\*/g, '<strong>$1</strong>')
        .replace(/__(.*?)__/g, '<strong>$1</strong>')
        .replace(/\*(.*?)\*/g, '<em>$1</em>')
        .replace(/_(.*?)_/g, '<em>$1</em>')
        .replace(/```.*?\n([\s\S]*?)```/g, '<pre><code>$1</code></pre>')
        .replace(/`(.*?)`/g, '<code>$1</code>')
        .replace(/\n/gim, '<br />');
      return html`<span dangerouslySetInnerHTML=${{ __html: md }} />`;
    };

    const ModelGenerationInfo = (params) => {
      if (!llamaStats.value) {
        return html`<span/>`
      }
      return html`
        <span>
          ${llamaStats.value.predicted_per_token_ms.toFixed()}ms per token, ${llamaStats.value.predicted_per_second.toFixed(2)} tokens per second
        </span>
      `
    }

    function App(props) {

      return html`
        <div id="container">
          <header>
            <h1>llama.cpp</h1>
          </header>

          <main id="content">
            <${chatStarted.value ? ChatLog : ConfigForm} />
          </main>

          <section id="write">
            <${MessageInput} />
          </section>

          <footer>
            <p><${ModelGenerationInfo} /></p>
            <p>Powered by <a href="https://github.com/ggerganov/llama.cpp">llama.cpp</a> and <a href="https://ggml.ai">ggml.ai</a>.</p>
          </footer>
        </div>
      `;
    }

    render(h(App), document.body);
  </script>
</head>

<body>
</body>

</html>