Upload 6 files

build/m.d.ts

@@ -0,0 +1,65 @@
+/* tslint:disable */
+/* eslint-disable */
+/**
+*/
+export class Model {
+  free(): void;
+/**
+* @param {Uint8Array} weights
+* @param {Uint8Array} tokenizer
+* @param {boolean} quantized
+*/
+  constructor(weights: Uint8Array, tokenizer: Uint8Array, quantized: boolean);
+/**
+* @param {string} prompt
+* @param {number} temp
+* @param {number} top_p
+* @param {number} repeat_penalty
+* @param {number} repeat_last_n
+* @param {bigint} seed
+* @returns {string}
+*/
+  init_with_prompt(prompt: string, temp: number, top_p: number, repeat_penalty: number, repeat_last_n: number, seed: bigint): string;
+/**
+* @returns {string}
+*/
+  next_token(): string;
+}
+
+export type InitInput = RequestInfo | URL | Response | BufferSource | WebAssembly.Module;
+
+export interface InitOutput {
+  readonly memory: WebAssembly.Memory;
+  readonly __wbg_model_free: (a: number) => void;
+  readonly model_load: (a: number, b: number, c: number, d: number, e: number, f: number) => void;
+  readonly model_init_with_prompt: (a: number, b: number, c: number, d: number, e: number, f: number, g: number, h: number, i: number) => void;
+  readonly model_next_token: (a: number, b: number) => void;
+  readonly main: (a: number, b: number) => number;
+  readonly __wbindgen_add_to_stack_pointer: (a: number) => number;
+  readonly __wbindgen_malloc: (a: number, b: number) => number;
+  readonly __wbindgen_realloc: (a: number, b: number, c: number, d: number) => number;
+  readonly __wbindgen_free: (a: number, b: number, c: number) => void;
+  readonly __wbindgen_exn_store: (a: number) => void;
+  readonly __wbindgen_start: () => void;
+}
+
+export type SyncInitInput = BufferSource | WebAssembly.Module;
+/**
+* Instantiates the given `module`, which can either be bytes or
+* a precompiled `WebAssembly.Module`.
+*
+* @param {SyncInitInput} module
+*
+* @returns {InitOutput}
+*/
+export function initSync(module: SyncInitInput): InitOutput;
+
+/**
+* If `module_or_path` is {RequestInfo} or {URL}, makes a request and
+* for everything else, calls `WebAssembly.instantiate` directly.
+*
+* @param {InitInput | Promise<InitInput>} module_or_path
+*
+* @returns {Promise<InitOutput>}
+*/
+export default function __wbg_init (module_or_path?: InitInput | Promise<InitInput>): Promise<InitOutput>;

build/m.js

@@ -0,0 +1,476 @@
+let wasm;
+
+const cachedTextDecoder = (typeof TextDecoder !== 'undefined' ? new TextDecoder('utf-8', { ignoreBOM: true, fatal: true }) : { decode: () => { throw Error('TextDecoder not available') } } );
+
+if (typeof TextDecoder !== 'undefined') { cachedTextDecoder.decode(); };
+
+let cachedUint8Memory0 = null;
+
+function getUint8Memory0() {
+    if (cachedUint8Memory0 === null || cachedUint8Memory0.byteLength === 0) {
+        cachedUint8Memory0 = new Uint8Array(wasm.memory.buffer);
+    }
+    return cachedUint8Memory0;
+}
+
+function getStringFromWasm0(ptr, len) {
+    ptr = ptr >>> 0;
+    return cachedTextDecoder.decode(getUint8Memory0().subarray(ptr, ptr + len));
+}
+
+const heap = new Array(128).fill(undefined);
+
+heap.push(undefined, null, true, false);
+
+let heap_next = heap.length;
+
+function addHeapObject(obj) {
+    if (heap_next === heap.length) heap.push(heap.length + 1);
+    const idx = heap_next;
+    heap_next = heap[idx];
+
+    heap[idx] = obj;
+    return idx;
+}
+
+function getObject(idx) { return heap[idx]; }
+
+function dropObject(idx) {
+    if (idx < 132) return;
+    heap[idx] = heap_next;
+    heap_next = idx;
+}
+
+function takeObject(idx) {
+    const ret = getObject(idx);
+    dropObject(idx);
+    return ret;
+}
+
+let WASM_VECTOR_LEN = 0;
+
+function passArray8ToWasm0(arg, malloc) {
+    const ptr = malloc(arg.length * 1, 1) >>> 0;
+    getUint8Memory0().set(arg, ptr / 1);
+    WASM_VECTOR_LEN = arg.length;
+    return ptr;
+}
+
+let cachedInt32Memory0 = null;
+
+function getInt32Memory0() {
+    if (cachedInt32Memory0 === null || cachedInt32Memory0.byteLength === 0) {
+        cachedInt32Memory0 = new Int32Array(wasm.memory.buffer);
+    }
+    return cachedInt32Memory0;
+}
+
+const cachedTextEncoder = (typeof TextEncoder !== 'undefined' ? new TextEncoder('utf-8') : { encode: () => { throw Error('TextEncoder not available') } } );
+
+const encodeString = (typeof cachedTextEncoder.encodeInto === 'function'
+    ? function (arg, view) {
+    return cachedTextEncoder.encodeInto(arg, view);
+}
+    : function (arg, view) {
+    const buf = cachedTextEncoder.encode(arg);
+    view.set(buf);
+    return {
+        read: arg.length,
+        written: buf.length
+    };
+});
+
+function passStringToWasm0(arg, malloc, realloc) {
+
+    if (realloc === undefined) {
+        const buf = cachedTextEncoder.encode(arg);
+        const ptr = malloc(buf.length, 1) >>> 0;
+        getUint8Memory0().subarray(ptr, ptr + buf.length).set(buf);
+        WASM_VECTOR_LEN = buf.length;
+        return ptr;
+    }
+
+    let len = arg.length;
+    let ptr = malloc(len, 1) >>> 0;
+
+    const mem = getUint8Memory0();
+
+    let offset = 0;
+
+    for (; offset < len; offset++) {
+        const code = arg.charCodeAt(offset);
+        if (code > 0x7F) break;
+        mem[ptr + offset] = code;
+    }
+
+    if (offset !== len) {
+        if (offset !== 0) {
+            arg = arg.slice(offset);
+        }
+        ptr = realloc(ptr, len, len = offset + arg.length * 3, 1) >>> 0;
+        const view = getUint8Memory0().subarray(ptr + offset, ptr + len);
+        const ret = encodeString(arg, view);
+
+        offset += ret.written;
+    }
+
+    WASM_VECTOR_LEN = offset;
+    return ptr;
+}
+
+function handleError(f, args) {
+    try {
+        return f.apply(this, args);
+    } catch (e) {
+        wasm.__wbindgen_exn_store(addHeapObject(e));
+    }
+}
+/**
+*/
+export class Model {
+
+    static __wrap(ptr) {
+        ptr = ptr >>> 0;
+        const obj = Object.create(Model.prototype);
+        obj.__wbg_ptr = ptr;
+
+        return obj;
+    }
+
+    __destroy_into_raw() {
+        const ptr = this.__wbg_ptr;
+        this.__wbg_ptr = 0;
+
+        return ptr;
+    }
+
+    free() {
+        const ptr = this.__destroy_into_raw();
+        wasm.__wbg_model_free(ptr);
+    }
+    /**
+    * @param {Uint8Array} weights
+    * @param {Uint8Array} tokenizer
+    * @param {boolean} quantized
+    */
+    constructor(weights, tokenizer, quantized) {
+        try {
+            const retptr = wasm.__wbindgen_add_to_stack_pointer(-16);
+            const ptr0 = passArray8ToWasm0(weights, wasm.__wbindgen_malloc);
+            const len0 = WASM_VECTOR_LEN;
+            const ptr1 = passArray8ToWasm0(tokenizer, wasm.__wbindgen_malloc);
+            const len1 = WASM_VECTOR_LEN;
+            wasm.model_load(retptr, ptr0, len0, ptr1, len1, quantized);
+            var r0 = getInt32Memory0()[retptr / 4 + 0];
+            var r1 = getInt32Memory0()[retptr / 4 + 1];
+            var r2 = getInt32Memory0()[retptr / 4 + 2];
+            if (r2) {
+                throw takeObject(r1);
+            }
+            return Model.__wrap(r0);
+        } finally {
+            wasm.__wbindgen_add_to_stack_pointer(16);
+        }
+    }
+    /**
+    * @param {string} prompt
+    * @param {number} temp
+    * @param {number} top_p
+    * @param {number} repeat_penalty
+    * @param {number} repeat_last_n
+    * @param {bigint} seed
+    * @returns {string}
+    */
+    init_with_prompt(prompt, temp, top_p, repeat_penalty, repeat_last_n, seed) {
+        let deferred3_0;
+        let deferred3_1;
+        try {
+            const retptr = wasm.__wbindgen_add_to_stack_pointer(-16);
+            const ptr0 = passStringToWasm0(prompt, wasm.__wbindgen_malloc, wasm.__wbindgen_realloc);
+            const len0 = WASM_VECTOR_LEN;
+            wasm.model_init_with_prompt(retptr, this.__wbg_ptr, ptr0, len0, temp, top_p, repeat_penalty, repeat_last_n, seed);
+            var r0 = getInt32Memory0()[retptr / 4 + 0];
+            var r1 = getInt32Memory0()[retptr / 4 + 1];
+            var r2 = getInt32Memory0()[retptr / 4 + 2];
+            var r3 = getInt32Memory0()[retptr / 4 + 3];
+            var ptr2 = r0;
+            var len2 = r1;
+            if (r3) {
+                ptr2 = 0; len2 = 0;
+                throw takeObject(r2);
+            }
+            deferred3_0 = ptr2;
+            deferred3_1 = len2;
+            return getStringFromWasm0(ptr2, len2);
+        } finally {
+            wasm.__wbindgen_add_to_stack_pointer(16);
+            wasm.__wbindgen_free(deferred3_0, deferred3_1, 1);
+        }
+    }
+    /**
+    * @returns {string}
+    */
+    next_token() {
+        let deferred2_0;
+        let deferred2_1;
+        try {
+            const retptr = wasm.__wbindgen_add_to_stack_pointer(-16);
+            wasm.model_next_token(retptr, this.__wbg_ptr);
+            var r0 = getInt32Memory0()[retptr / 4 + 0];
+            var r1 = getInt32Memory0()[retptr / 4 + 1];
+            var r2 = getInt32Memory0()[retptr / 4 + 2];
+            var r3 = getInt32Memory0()[retptr / 4 + 3];
+            var ptr1 = r0;
+            var len1 = r1;
+            if (r3) {
+                ptr1 = 0; len1 = 0;
+                throw takeObject(r2);
+            }
+            deferred2_0 = ptr1;
+            deferred2_1 = len1;
+            return getStringFromWasm0(ptr1, len1);
+        } finally {
+            wasm.__wbindgen_add_to_stack_pointer(16);
+            wasm.__wbindgen_free(deferred2_0, deferred2_1, 1);
+        }
+    }
+}
+
+async function __wbg_load(module, imports) {
+    if (typeof Response === 'function' && module instanceof Response) {
+        if (typeof WebAssembly.instantiateStreaming === 'function') {
+            try {
+                return await WebAssembly.instantiateStreaming(module, imports);
+
+            } catch (e) {
+                if (module.headers.get('Content-Type') != 'application/wasm') {
+                    console.warn("`WebAssembly.instantiateStreaming` failed because your server does not serve wasm with `application/wasm` MIME type. Falling back to `WebAssembly.instantiate` which is slower. Original error:\n", e);
+
+                } else {
+                    throw e;
+                }
+            }
+        }
+
+        const bytes = await module.arrayBuffer();
+        return await WebAssembly.instantiate(bytes, imports);
+
+    } else {
+        const instance = await WebAssembly.instantiate(module, imports);
+
+        if (instance instanceof WebAssembly.Instance) {
+            return { instance, module };
+
+        } else {
+            return instance;
+        }
+    }
+}
+
+function __wbg_get_imports() {
+    const imports = {};
+    imports.wbg = {};
+    imports.wbg.__wbindgen_error_new = function(arg0, arg1) {
+        const ret = new Error(getStringFromWasm0(arg0, arg1));
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_new_abda76e883ba8a5f = function() {
+        const ret = new Error();
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_stack_658279fe44541cf6 = function(arg0, arg1) {
+        const ret = getObject(arg1).stack;
+        const ptr1 = passStringToWasm0(ret, wasm.__wbindgen_malloc, wasm.__wbindgen_realloc);
+        const len1 = WASM_VECTOR_LEN;
+        getInt32Memory0()[arg0 / 4 + 1] = len1;
+        getInt32Memory0()[arg0 / 4 + 0] = ptr1;
+    };
+    imports.wbg.__wbg_error_f851667af71bcfc6 = function(arg0, arg1) {
+        let deferred0_0;
+        let deferred0_1;
+        try {
+            deferred0_0 = arg0;
+            deferred0_1 = arg1;
+            console.error(getStringFromWasm0(arg0, arg1));
+        } finally {
+            wasm.__wbindgen_free(deferred0_0, deferred0_1, 1);
+        }
+    };
+    imports.wbg.__wbindgen_object_drop_ref = function(arg0) {
+        takeObject(arg0);
+    };
+    imports.wbg.__wbg_log_ff7e0b5e6573cdff = function(arg0, arg1) {
+        console.log(getStringFromWasm0(arg0, arg1));
+    };
+    imports.wbg.__wbg_crypto_c48a774b022d20ac = function(arg0) {
+        const ret = getObject(arg0).crypto;
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbindgen_is_object = function(arg0) {
+        const val = getObject(arg0);
+        const ret = typeof(val) === 'object' && val !== null;
+        return ret;
+    };
+    imports.wbg.__wbg_process_298734cf255a885d = function(arg0) {
+        const ret = getObject(arg0).process;
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_versions_e2e78e134e3e5d01 = function(arg0) {
+        const ret = getObject(arg0).versions;
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_node_1cd7a5d853dbea79 = function(arg0) {
+        const ret = getObject(arg0).node;
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbindgen_is_string = function(arg0) {
+        const ret = typeof(getObject(arg0)) === 'string';
+        return ret;
+    };
+    imports.wbg.__wbg_msCrypto_bcb970640f50a1e8 = function(arg0) {
+        const ret = getObject(arg0).msCrypto;
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_require_8f08ceecec0f4fee = function() { return handleError(function () {
+        const ret = module.require;
+        return addHeapObject(ret);
+    }, arguments) };
+    imports.wbg.__wbindgen_is_function = function(arg0) {
+        const ret = typeof(getObject(arg0)) === 'function';
+        return ret;
+    };
+    imports.wbg.__wbindgen_string_new = function(arg0, arg1) {
+        const ret = getStringFromWasm0(arg0, arg1);
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_getRandomValues_37fa2ca9e4e07fab = function() { return handleError(function (arg0, arg1) {
+        getObject(arg0).getRandomValues(getObject(arg1));
+    }, arguments) };
+    imports.wbg.__wbg_randomFillSync_dc1e9a60c158336d = function() { return handleError(function (arg0, arg1) {
+        getObject(arg0).randomFillSync(takeObject(arg1));
+    }, arguments) };
+    imports.wbg.__wbg_newnoargs_581967eacc0e2604 = function(arg0, arg1) {
+        const ret = new Function(getStringFromWasm0(arg0, arg1));
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_call_cb65541d95d71282 = function() { return handleError(function (arg0, arg1) {
+        const ret = getObject(arg0).call(getObject(arg1));
+        return addHeapObject(ret);
+    }, arguments) };
+    imports.wbg.__wbindgen_object_clone_ref = function(arg0) {
+        const ret = getObject(arg0);
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_self_1ff1d729e9aae938 = function() { return handleError(function () {
+        const ret = self.self;
+        return addHeapObject(ret);
+    }, arguments) };
+    imports.wbg.__wbg_window_5f4faef6c12b79ec = function() { return handleError(function () {
+        const ret = window.window;
+        return addHeapObject(ret);
+    }, arguments) };
+    imports.wbg.__wbg_globalThis_1d39714405582d3c = function() { return handleError(function () {
+        const ret = globalThis.globalThis;
+        return addHeapObject(ret);
+    }, arguments) };
+    imports.wbg.__wbg_global_651f05c6a0944d1c = function() { return handleError(function () {
+        const ret = global.global;
+        return addHeapObject(ret);
+    }, arguments) };
+    imports.wbg.__wbindgen_is_undefined = function(arg0) {
+        const ret = getObject(arg0) === undefined;
+        return ret;
+    };
+    imports.wbg.__wbg_call_01734de55d61e11d = function() { return handleError(function (arg0, arg1, arg2) {
+        const ret = getObject(arg0).call(getObject(arg1), getObject(arg2));
+        return addHeapObject(ret);
+    }, arguments) };
+    imports.wbg.__wbg_now_9c5990bda04c7e53 = function() {
+        const ret = Date.now();
+        return ret;
+    };
+    imports.wbg.__wbg_buffer_085ec1f694018c4f = function(arg0) {
+        const ret = getObject(arg0).buffer;
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_newwithbyteoffsetandlength_6da8e527659b86aa = function(arg0, arg1, arg2) {
+        const ret = new Uint8Array(getObject(arg0), arg1 >>> 0, arg2 >>> 0);
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_new_8125e318e6245eed = function(arg0) {
+        const ret = new Uint8Array(getObject(arg0));
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_set_5cf90238115182c3 = function(arg0, arg1, arg2) {
+        getObject(arg0).set(getObject(arg1), arg2 >>> 0);
+    };
+    imports.wbg.__wbg_newwithlength_e5d69174d6984cd7 = function(arg0) {
+        const ret = new Uint8Array(arg0 >>> 0);
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbg_subarray_13db269f57aa838d = function(arg0, arg1, arg2) {
+        const ret = getObject(arg0).subarray(arg1 >>> 0, arg2 >>> 0);
+        return addHeapObject(ret);
+    };
+    imports.wbg.__wbindgen_throw = function(arg0, arg1) {
+        throw new Error(getStringFromWasm0(arg0, arg1));
+    };
+    imports.wbg.__wbindgen_memory = function() {
+        const ret = wasm.memory;
+        return addHeapObject(ret);
+    };
+
+    return imports;
+}
+
+function __wbg_init_memory(imports, maybe_memory) {
+
+}
+
+function __wbg_finalize_init(instance, module) {
+    wasm = instance.exports;
+    __wbg_init.__wbindgen_wasm_module = module;
+    cachedInt32Memory0 = null;
+    cachedUint8Memory0 = null;
+
+    wasm.__wbindgen_start();
+    return wasm;
+}
+
+function initSync(module) {
+    if (wasm !== undefined) return wasm;
+
+    const imports = __wbg_get_imports();
+
+    __wbg_init_memory(imports);
+
+    if (!(module instanceof WebAssembly.Module)) {
+        module = new WebAssembly.Module(module);
+    }
+
+    const instance = new WebAssembly.Instance(module, imports);
+
+    return __wbg_finalize_init(instance, module);
+}
+
+async function __wbg_init(input) {
+    if (wasm !== undefined) return wasm;
+
+    if (typeof input === 'undefined') {
+        input = new URL('m_bg.wasm', import.meta.url);
+    }
+    const imports = __wbg_get_imports();
+
+    if (typeof input === 'string' || (typeof Request === 'function' && input instanceof Request) || (typeof URL === 'function' && input instanceof URL)) {
+        input = fetch(input);
+    }
+
+    __wbg_init_memory(imports);
+
+    const { instance, module } = await __wbg_load(await input, imports);
+
+    return __wbg_finalize_init(instance, module);
+}
+
+export { initSync }
+export default __wbg_init;

build/m_bg.wasm

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a62ccdcce6e333016d5f26d45ceaa3066123a2396f3ea46d419a046f02363ae6
+size 4349308

build/m_bg.wasm.d.ts

@@ -0,0 +1,14 @@
+/* tslint:disable */
+/* eslint-disable */
+export const memory: WebAssembly.Memory;
+export function __wbg_model_free(a: number): void;
+export function model_load(a: number, b: number, c: number, d: number, e: number, f: number): void;
+export function model_init_with_prompt(a: number, b: number, c: number, d: number, e: number, f: number, g: number, h: number, i: number): void;
+export function model_next_token(a: number, b: number): void;
+export function main(a: number, b: number): number;
+export function __wbindgen_add_to_stack_pointer(a: number): number;
+export function __wbindgen_malloc(a: number, b: number): number;
+export function __wbindgen_realloc(a: number, b: number, c: number, d: number): number;
+export function __wbindgen_free(a: number, b: number, c: number): void;
+export function __wbindgen_exn_store(a: number): void;
+export function __wbindgen_start(): void;

index.html

@@ -1,19 +1,366 @@
+<html>
+  <head>
+    <meta content="text/html;charset=utf-8" http-equiv="Content-Type" />
+    <title>Candle Phi 1.5 Rust/WASM</title>
+  </head>
+  <body></body>
+</html>
+
 <!DOCTYPE html>
 <html>
-	<head>
-		<meta charset="utf-8" />
-		<meta name="viewport" content="width=device-width" />
-		<title>My static Space</title>
-		<link rel="stylesheet" href="style.css" />
-	</head>
-	<body>
-		<div class="card">
-			<h1>Welcome to your static Space!</h1>
-			<p>You can modify this app directly by editing <i>index.html</i> in the Files and versions tab.</p>
-			<p>
-				Also don't forget to check the
-				<a href="https://huggingface.co/docs/hub/spaces" target="_blank">Spaces documentation</a>.
-			</p>
-		</div>
-	</body>
+  <head>
+    <meta charset="UTF-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1.0" />
+    <link
+      rel="stylesheet"
+      href="https://cdn.jsdelivr.net/gh/highlightjs/cdn-release@11.8.0/build/styles/default.min.css" />
+    <style>
+      @import url("https://fonts.googleapis.com/css2?family=Source+Code+Pro:wght@200;300;400&family=Source+Sans+3:wght@100;200;300;400;500;600;700;800;900&display=swap");
+      html,
+      body {
+        font-family: "Source Sans 3", sans-serif;
+      }
+      code,
+      output,
+      select,
+      pre {
+        font-family: "Source Code Pro", monospace;
+      }
+    </style>
+    <style type="text/tailwindcss">
+      .link {
+        @apply underline hover:text-blue-500 hover:no-underline;
+      }
+    </style>
+    <script src="https://cdn.tailwindcss.com"></script>
+    <script type="module">
+      import snarkdown from "https://cdn.skypack.dev/snarkdown";
+      import hljs from "https://cdn.skypack.dev/highlight.js";
+
+      const TOKENIZER_URL =
+        "https://huggingface.co/microsoft/phi-1_5/raw/main/tokenizer.json";
+      // models base url
+      const MODELS = {
+        phi_1_5_quantized: {
+          base_url:
+            "https://huggingface.co/lmz/candle-quantized-phi/resolve/main/",
+          model: "model-q4k.gguf",
+          quantized: true,
+          seq_len: 2048,
+        },
+        phi_1_5_quantized_2: {
+          base_url:
+            "https://huggingface.co/lmz/candle-quantized-phi/resolve/main/",
+          model: "model-q80.gguf",
+          quantized: true,
+          seq_len: 2048,
+        },
+      };
+
+      const phiWorker = new Worker("./phiWorker.js", {
+        type: "module",
+      });
+      async function generateSequence(controller) {
+        const getValue = (id) => document.querySelector(`#${id}`).value;
+        const modelID = getValue("model");
+        const model = MODELS[modelID];
+        const weightsURL = model.base_url + model.model;
+
+        const prompt = getValue("prompt").trim();
+        const temperature = getValue("temperature");
+        const topP = getValue("top-p");
+        const repeatPenalty = getValue("repeat_penalty");
+        const seed = getValue("seed");
+        const maxSeqLen = getValue("max-seq");
+
+        function updateStatus(data) {
+          const outStatus = document.querySelector("#output-status");
+          const outGen = document.querySelector("#output-generation");
+          const outCounter = document.querySelector("#output-counter");
+
+          switch (data.status) {
+            case "loading":
+              outStatus.hidden = false;
+              outStatus.textContent = data.message;
+              outGen.hidden = true;
+              outCounter.hidden = true;
+              break;
+            case "generating":
+              const { message, prompt, sentence, tokensSec, totalTime } = data;
+              outStatus.hidden = true;
+              outCounter.hidden = false;
+              outGen.hidden = false;
+
+              const p = prompt.replace(/\n|\\n/g, "  \n");
+              const generated = sentence.replace(/\n|\\n/g, "  \n");
+              outGen.innerHTML = snarkdown(p + generated);
+              outCounter.innerHTML = `${(totalTime / 1000).toFixed(
+                2
+              )}s (${tokensSec.toFixed(2)} tok/s)`;
+              hljs.highlightAll();
+              break;
+            case "complete":
+              outStatus.hidden = true;
+              outGen.hidden = false;
+              break;
+          }
+        }
+
+        return new Promise((resolve, reject) => {
+          phiWorker.postMessage({
+            weightsURL,
+            modelID,
+            tokenizerURL: TOKENIZER_URL,
+            quantized: model.quantized,
+            prompt,
+            temp: temperature,
+            top_p: topP,
+            repeatPenalty,
+            seed: seed,
+            maxSeqLen,
+            command: "start",
+          });
+
+          const handleAbort = () => {
+            phiWorker.postMessage({ command: "abort" });
+          };
+          const handleMessage = (event) => {
+            const { status, error, message, prompt, sentence } = event.data;
+            if (status) updateStatus(event.data);
+            if (error) {
+              phiWorker.removeEventListener("message", handleMessage);
+              reject(new Error(error));
+            }
+            if (status === "abort") {
+              phiWorker.removeEventListener("message", handleMessage);
+              resolve(event.data);
+            }
+            if (status === "complete") {
+              phiWorker.removeEventListener("message", handleMessage);
+              resolve(event.data);
+            }
+          };
+
+          controller.signal.addEventListener("abort", handleAbort);
+          phiWorker.addEventListener("message", handleMessage);
+        });
+      }
+
+      const form = document.querySelector("#form");
+      const prompt = document.querySelector("#prompt");
+      const clearBtn = document.querySelector("#clear-btn");
+      const runBtn = document.querySelector("#run");
+      const modelSelect = document.querySelector("#model");
+      let runController = new AbortController();
+      let isRunning = false;
+
+      modelSelect.addEventListener("change", (e) => {
+        const model = MODELS[e.target.value];
+        document.querySelector("#max-seq").max = model.seq_len;
+        document.querySelector("#max-seq").nextElementSibling.value = 200;
+      });
+
+      form.addEventListener("submit", async (e) => {
+        e.preventDefault();
+        if (isRunning) {
+          stopRunning();
+        } else {
+          startRunning();
+          await generateSequence(runController);
+          stopRunning();
+        }
+      });
+
+      function startRunning() {
+        isRunning = true;
+        runBtn.textContent = "Stop";
+      }
+
+      function stopRunning() {
+        runController.abort();
+        runController = new AbortController();
+        runBtn.textContent = "Run";
+        isRunning = false;
+      }
+      clearBtn.addEventListener("click", (e) => {
+        e.preventDefault();
+        prompt.value = "";
+        clearBtn.classList.add("invisible");
+        runBtn.disabled = true;
+        stopRunning();
+      });
+      prompt.addEventListener("input", (e) => {
+        runBtn.disabled = false;
+        if (e.target.value.length > 0) {
+          clearBtn.classList.remove("invisible");
+        } else {
+          clearBtn.classList.add("invisible");
+        }
+      });
+    </script>
+  </head>
+  <body class="container max-w-4xl mx-auto p-4 text-gray-800">
+    <main class="grid grid-cols-1 gap-8 relative">
+      <span class="absolute text-5xl -ml-[1em]"> 🕯️ </span>
+      <div>
+        <h1 class="text-5xl font-bold">Candle Phi 1.5</h1>
+        <h2 class="text-2xl font-bold">Rust/WASM Demo</h2>
+        <p class="max-w-lg">
+          The
+          <a
+            href="https://huggingface.co/microsoft/phi-1_5"
+            class="link"
+            target="_blank">Phi-1.5</a
+          >
+          model achieves state-of-the-art performance with only 1.3 billion
+          parameters, compared to models with up to 10 billion. You can try the
+          quantized version of the model here. Additional prompt examples are
+          available in the
+          <a
+            href="https://arxiv.org/pdf/2309.05463.pdf#page=8"
+            class="link"
+            target="_blank">
+            technical report </a
+          >.
+        </p>
+      </div>
+
+      <div>
+        <label for="model" class="font-medium">Models Options: </label>
+        <select
+          id="model"
+          class="border-2 border-gray-500 rounded-md font-light">
+          <option value="phi_1_5_quantized" selected>
+            phi 1.5 quantized q4k (800 MB)
+          </option>
+          <option value="phi_1_5_quantized_2">
+            phi 1.5 quantized q80 (1.51 GB)
+          </option>
+          <!-- <option value="phi_1_5">phi 1.5 (2.84 GB)</option> -->
+        </select>
+      </div>
+      <form
+        id="form"
+        class="flex text-normal px-1 py-1 border border-gray-700 rounded-md items-center">
+        <input type="submit" hidden />
+        <textarea
+          type="text"
+          id="prompt"
+          class="font-light w-full px-3 py-2 mx-1 resize-none outline-none"
+          oninput="this.style.height = 0;this.style.height = this.scrollHeight + 'px'"
+          placeholder="Add your prompt here...">
+Write a detailed analogy between mathematics and a lighthouse.
+Answer:</textarea
+        >
+        <button id="clear-btn">
+          <svg
+            fill="none"
+            xmlns="http://www.w3.org/2000/svg"
+            width="40"
+            viewBox="0 0 70 40">
+            <path opacity=".5" d="M39 .2v40.2" stroke="#1F2937" />
+            <path
+              d="M1.5 11.5 19 29.1m0-17.6L1.5 29.1"
+              opacity=".5"
+              stroke="#1F2937"
+              stroke-width="2" />
+          </svg>
+        </button>
+        <button
+          id="run"
+          class="bg-gray-700 hover:bg-gray-800 text-white font-normal py-2 w-16 rounded disabled:bg-gray-300 disabled:cursor-not-allowed">
+          Run
+        </button>
+      </form>
+      <div class="grid grid-cols-3 max-w-md items-center gap-3">
+        <label class="text-sm font-medium" for="max-seq">Maximum length </label>
+        <input
+          type="range"
+          id="max-seq"
+          name="max-seq"
+          min="1"
+          max="2048"
+          step="1"
+          value="200"
+          oninput="this.nextElementSibling.value = Number(this.value)" />
+        <output
+          class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md">
+          200</output
+        >
+        <label class="text-sm font-medium" for="temperature">Temperature</label>
+        <input
+          type="range"
+          id="temperature"
+          name="temperature"
+          min="0"
+          max="2"
+          step="0.01"
+          value="0.00"
+          oninput="this.nextElementSibling.value = Number(this.value).toFixed(2)" />
+        <output
+          class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md">
+          0.00</output
+        >
+        <label class="text-sm font-medium" for="top-p">Top-p</label>
+        <input
+          type="range"
+          id="top-p"
+          name="top-p"
+          min="0"
+          max="1"
+          step="0.01"
+          value="1.00"
+          oninput="this.nextElementSibling.value = Number(this.value).toFixed(2)" />
+        <output
+          class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md">
+          1.00</output
+        >
+
+        <label class="text-sm font-medium" for="repeat_penalty"
+          >Repeat Penalty</label
+        >
+
+        <input
+          type="range"
+          id="repeat_penalty"
+          name="repeat_penalty"
+          min="1"
+          max="2"
+          step="0.01"
+          value="1.10"
+          oninput="this.nextElementSibling.value = Number(this.value).toFixed(2)" />
+        <output
+          class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md"
+          >1.10</output
+        >
+        <label class="text-sm font-medium" for="seed">Seed</label>
+        <input
+          type="number"
+          id="seed"
+          name="seed"
+          value="299792458"
+          class="font-light border border-gray-700 text-right rounded-md p-2" />
+        <button
+          id="run"
+          onclick="document.querySelector('#seed').value = Math.floor(Math.random() * Number.MAX_SAFE_INTEGER)"
+          class="bg-gray-700 hover:bg-gray-800 text-white font-normal py-1 w-[50px] rounded disabled:bg-gray-300 disabled:cursor-not-allowed text-sm">
+          Rand
+        </button>
+      </div>
+      <div>
+        <h3 class="font-medium">Generation:</h3>
+        <div
+          class="min-h-[250px] bg-slate-100 text-gray-500 p-4 rounded-md flex flex-col gap-2">
+          <div
+            id="output-counter"
+            hidden
+            class="ml-auto font-semibold grid-rows-1 text-sm"></div>
+          <p hidden id="output-generation" class="grid-rows-2"></p>
+          <span id="output-status" class="m-auto font-light"
+            >No output yet</span
+          >
+        </div>
+      </div>
+    </main>
+  </body>
 </html>

phiWorker.js

@@ -0,0 +1,131 @@
+import init, { Model } from "./build/m.js";
+
+async function fetchArrayBuffer(url) {
+  const cacheName = "phi-mixformer-candle-cache";
+  const cache = await caches.open(cacheName);
+  const cachedResponse = await cache.match(url);
+  if (cachedResponse) {
+    const data = await cachedResponse.arrayBuffer();
+    return new Uint8Array(data);
+  }
+  const res = await fetch(url, { cache: "force-cache" });
+  cache.put(url, res.clone());
+  return new Uint8Array(await res.arrayBuffer());
+}
+class Phi {
+  static instance = {};
+
+  static async getInstance(weightsURL, modelID, tokenizerURL, quantized) {
+    // load individual modelID only once
+    if (!this.instance[modelID]) {
+      await init();
+
+      self.postMessage({ status: "loading", message: "Loading Model" });
+
+      const [weightsArrayU8, tokenizerArrayU8] = await Promise.all([
+        fetchArrayBuffer(weightsURL),
+        fetchArrayBuffer(tokenizerURL),
+      ]);
+
+      this.instance[modelID] = new Model(
+        weightsArrayU8,
+        tokenizerArrayU8,
+        quantized
+      );
+    }
+    return this.instance[modelID];
+  }
+}
+
+let controller = null;
+self.addEventListener("message", (event) => {
+  if (event.data.command === "start") {
+    controller = new AbortController();
+    generate(event.data);
+  } else if (event.data.command === "abort") {
+    controller.abort();
+  }
+});
+
+async function generate(data) {
+  const {
+    weightsURL,
+    modelID,
+    tokenizerURL,
+    quantized,
+    prompt,
+    temp,
+    top_p,
+    repeatPenalty,
+    seed,
+    maxSeqLen,
+  } = data;
+  try {
+    self.postMessage({ status: "loading", message: "Starting Phi" });
+    const model = await Phi.getInstance(
+      weightsURL,
+      modelID,
+      tokenizerURL,
+      quantized
+    );
+
+    self.postMessage({ status: "loading", message: "Initializing model" });
+    model.init_with_prompt(
+      prompt,
+      temp,
+      top_p,
+      repeatPenalty,
+      64,
+      BigInt(seed)
+    );
+    const seq_len = 2048;
+
+    let sentence = "";
+    let maxTokens = maxSeqLen ? maxSeqLen : seq_len - prompt.length - 1;
+    let startTime = performance.now();
+    let tokensCount = 0;
+    while (tokensCount < maxTokens) {
+      await new Promise(async (resolve) => {
+        if (controller && controller.signal.aborted) {
+          self.postMessage({
+            status: "aborted",
+            message: "Aborted",
+            output: prompt + sentence,
+          });
+          return;
+        }
+        const token = await model.next_token();
+        if (token === "<|endoftext|>") {
+          self.postMessage({
+            status: "complete",
+            message: "complete",
+            output: prompt + sentence,
+          });
+          return;
+        }
+        const tokensSec =
+          ((tokensCount + 1) / (performance.now() - startTime)) * 1000;
+
+        sentence += token;
+        self.postMessage({
+          status: "generating",
+          message: "Generating token",
+          token: token,
+          sentence: sentence,
+          totalTime: performance.now() - startTime,
+          tokensSec,
+          prompt: prompt,
+        });
+        setTimeout(resolve, 0);
+      });
+      tokensCount++;
+    }
+    self.postMessage({
+      status: "complete",
+      message: "complete",
+      output: prompt + sentence,
+    });
+  } catch (e) {
+    self.postMessage({ error: e });
+  }
+}