utils.ts 13.2 KB
Newer Older
liuzhe-lz's avatar
liuzhe-lz committed
1
2
// Copyright (c) Microsoft Corporation.
// Licensed under the MIT license.
Deshui Yu's avatar
Deshui Yu committed
3
4
5

'use strict';

6
import * as assert from 'assert';
Deshui Yu's avatar
Deshui Yu committed
7
import { randomBytes } from 'crypto';
8
import * as cpp from 'child-process-promise';
9
10
import * as cp from 'child_process';
import { ChildProcess, spawn, StdioOptions } from 'child_process';
11
import * as dgram from 'dgram';
Deshui Yu's avatar
Deshui Yu committed
12
import * as fs from 'fs';
J-shang's avatar
J-shang committed
13
import * as net from 'net';
Deshui Yu's avatar
Deshui Yu committed
14
15
import * as os from 'os';
import * as path from 'path';
16
import * as lockfile from 'lockfile';
Deshui Yu's avatar
Deshui Yu committed
17
18
19
import { Deferred } from 'ts-deferred';
import { Container } from 'typescript-ioc';
import * as util from 'util';
20
import * as glob from 'glob';
Deshui Yu's avatar
Deshui Yu committed
21
22

import { Database, DataStore } from './datastore';
23
import { getExperimentStartupInfo, setExperimentStartupInfo } from './experimentStartupInfo';
24
import { ExperimentConfig, Manager } from './manager';
25
import { ExperimentManager } from './experimentManager';
QuanluZhang's avatar
QuanluZhang committed
26
import { HyperParameters, TrainingService, TrialJobStatus } from './trainingService';
Deshui Yu's avatar
Deshui Yu committed
27

28
function getExperimentRootDir(): string {
29
    return getExperimentStartupInfo().logDir;
Deshui Yu's avatar
Deshui Yu committed
30
31
}

32
function getLogDir(): string {
Deshui Yu's avatar
Deshui Yu committed
33
34
35
    return path.join(getExperimentRootDir(), 'log');
}

36
function getLogLevel(): string {
37
    return getExperimentStartupInfo().logLevel;
38
39
}

Deshui Yu's avatar
Deshui Yu committed
40
41
42
43
function getDefaultDatabaseDir(): string {
    return path.join(getExperimentRootDir(), 'db');
}

QuanluZhang's avatar
QuanluZhang committed
44
45
46
47
function getCheckpointDir(): string {
    return path.join(getExperimentRootDir(), 'checkpoint');
}

48
49
50
51
function getExperimentsInfoPath(): string {
    return path.join(os.homedir(), 'nni-experiments', '.experiment');
}

Deshui Yu's avatar
Deshui Yu committed
52
53
54
55
56
57
58
59
function mkDirP(dirPath: string): Promise<void> {
    const deferred: Deferred<void> = new Deferred<void>();
    fs.exists(dirPath, (exists: boolean) => {
        if (exists) {
            deferred.resolve();
        } else {
            const parent: string = path.dirname(dirPath);
            mkDirP(parent).then(() => {
60
                fs.mkdir(dirPath, (err: Error | null) => {
Deshui Yu's avatar
Deshui Yu committed
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
                    if (err) {
                        deferred.reject(err);
                    } else {
                        deferred.resolve();
                    }
                });
            }).catch((err: Error) => {
                deferred.reject(err);
            });
        }
    });

    return deferred.promise;
}

function mkDirPSync(dirPath: string): void {
    if (fs.existsSync(dirPath)) {
        return;
    }
    mkDirPSync(path.dirname(dirPath));
    fs.mkdirSync(dirPath);
}

const delay: (ms: number) => Promise<void> = util.promisify(setTimeout);

/**
 * Convert index to character
 * @param index index
 * @returns a mapping character
 */
function charMap(index: number): number {
    if (index < 26) {
        return index + 97;
    } else if (index < 52) {
        return index - 26 + 65;
    } else {
        return index - 52 + 48;
    }
}

/**
 * Generate a unique string by length
 * @param len length of string
 * @returns a unique string
 */
function uniqueString(len: number): string {
    if (len === 0) {
        return '';
    }
    const byteLength: number = Math.ceil((Math.log2(52) + Math.log2(62) * (len - 1)) / 8);
    let num: number = randomBytes(byteLength).reduce((a: number, b: number) => a * 256 + b, 0);
    const codes: number[] = [];
    codes.push(charMap(num % 52));
    num = Math.floor(num / 52);
    for (let i: number = 1; i < len; i++) {
        codes.push(charMap(num % 62));
        num = Math.floor(num / 62);
    }

    return String.fromCharCode(...codes);
}

123
124
125
126
function randomInt(max: number): number {
    return Math.floor(Math.random() * max);
}

127
128
129
130
131
function randomSelect<T>(a: T[]): T {
    assert(a !== undefined);

    return a[Math.floor(Math.random() * a.length)];
}
132

Deshui Yu's avatar
Deshui Yu committed
133
134
135
136
137
138
139
140
141
142
143
144
function parseArg(names: string[]): string {
    if (process.argv.length >= 4) {
        for (let i: number = 2; i < process.argv.length - 1; i++) {
            if (names.includes(process.argv[i])) {
                return process.argv[i + 1];
            }
        }
    }

    return '';
}

145
function getCmdPy(): string {
146
    let cmd = 'python3';
147
    if (process.platform === 'win32') {
148
149
150
151
152
        cmd = 'python';
    }
    return cmd;
}

153
/**
154
 * Generate command line to start automl algorithm(s),
QuanluZhang's avatar
QuanluZhang committed
155
 * either start advisor or start a process which runs tuner and assessor
156
 *
chicm-ms's avatar
chicm-ms committed
157
 * @param expParams: experiment startup parameters
158
159
 *
 */
160
function getMsgDispatcherCommand(expParams: ExperimentConfig): string {
chicm-ms's avatar
chicm-ms committed
161
162
163
    const clonedParams = Object.assign({}, expParams);
    delete clonedParams.searchSpace;
    return `${getCmdPy()} -m nni --exp_params ${Buffer.from(JSON.stringify(clonedParams)).toString('base64')}`;
164
165
}

166
167
168
169
/**
 * Generate parameter file name based on HyperParameters object
 * @param hyperParameters HyperParameters instance
 */
chicm-ms's avatar
chicm-ms committed
170
function generateParamFileName(hyperParameters: HyperParameters): string {
171
172
173
    assert(hyperParameters !== undefined);
    assert(hyperParameters.index >= 0);

chicm-ms's avatar
chicm-ms committed
174
    let paramFileName: string;
175
    if (hyperParameters.index == 0) {
176
177
178
179
180
181
182
        paramFileName = 'parameter.cfg';
    } else {
        paramFileName = `parameter_${hyperParameters.index}.cfg`
    }
    return paramFileName;
}

Deshui Yu's avatar
Deshui Yu committed
183
184
185
186
187
188
189
190
191
/**
 * Initialize a pseudo experiment environment for unit test.
 * Must be paired with `cleanupUnitTest()`.
 */
function prepareUnitTest(): void {
    Container.snapshot(Database);
    Container.snapshot(DataStore);
    Container.snapshot(TrainingService);
    Container.snapshot(Manager);
192
    Container.snapshot(ExperimentManager);
Deshui Yu's avatar
Deshui Yu committed
193

194
195
196
    const logLevel: string = parseArg(['--log_level', '-ll']);

    setExperimentStartupInfo(true, 'unittest', 8080, 'unittest', undefined, logLevel);
Deshui Yu's avatar
Deshui Yu committed
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
    mkDirPSync(getLogDir());

    const sqliteFile: string = path.join(getDefaultDatabaseDir(), 'nni.sqlite');
    try {
        fs.unlinkSync(sqliteFile);
    } catch (err) {
        // file not exists, good
    }
}

/**
 * Clean up unit test pseudo experiment.
 * Must be paired with `prepareUnitTest()`.
 */
function cleanupUnitTest(): void {
    Container.restore(Manager);
    Container.restore(TrainingService);
    Container.restore(DataStore);
    Container.restore(Database);
216
    Container.restore(ExperimentManager);
217
218
    const logLevel: string = parseArg(['--log_level', '-ll']);
    setExperimentStartupInfo(true, 'unittest', 8080, 'unittest', undefined, logLevel);
Deshui Yu's avatar
Deshui Yu committed
219
220
}

221
222
let cachedIpv4Address: string | null = null;

223
/**
224
 * Get IPv4 address of current machine.
225
226
 */
function getIPV4Address(): string {
227
228
    if (cachedIpv4Address !== null) {
        return cachedIpv4Address;
229
    }
230

231
232
233
234
235
236
    // creates "udp connection" to a non-exist target, and get local address of the connection.
    // since udp is connectionless, this does not send actual packets.
    const socket = dgram.createSocket('udp4');
    socket.connect(1, '192.0.2.0');
    cachedIpv4Address = socket.address().address;
    socket.close();
237

238
    return cachedIpv4Address;
239
240
}

QuanluZhang's avatar
QuanluZhang committed
241
242
243
244
245
246
247
/**
 * Get the status of canceled jobs according to the hint isEarlyStopped
 */
function getJobCancelStatus(isEarlyStopped: boolean): TrialJobStatus {
    return isEarlyStopped ? 'EARLY_STOPPED' : 'USER_CANCELED';
}

248
249
250
251
/**
 * Utility method to calculate file numbers under a directory, recursively
 * @param directory directory name
 */
chicm-ms's avatar
chicm-ms committed
252
function countFilesRecursively(directory: string): Promise<number> {
253
    if (!fs.existsSync(directory)) {
254
255
256
257
258
        throw Error(`Direcotory ${directory} doesn't exist`);
    }

    const deferred: Deferred<number> = new Deferred<number>();

chicm-ms's avatar
chicm-ms committed
259
260
    let timeoutId: NodeJS.Timer
    const delayTimeout: Promise<number> = new Promise((resolve: Function, reject: Function): void => {
261
262
263
264
265
266
267
        // Set timeout and reject the promise once reach timeout (5 seconds)
        timeoutId = setTimeout(() => {
            reject(new Error(`Timeout: path ${directory} has too many files`));
        }, 5000);
    });

    let fileCount: number = -1;
268
    let cmd: string;
269
    if (process.platform === "win32") {
270
271
        cmd = `powershell "Get-ChildItem -Path ${directory} -Recurse -File | Measure-Object | %{$_.Count}"`
    } else {
272
        cmd = `find ${directory} -type f | wc -l`;
273
274
    }
    cpp.exec(cmd).then((result) => {
275
        if (result.stdout && parseInt(result.stdout)) {
276
            fileCount = parseInt(result.stdout);
277
278
279
280
281
282
283
284
        }
        deferred.resolve(fileCount);
    });
    return Promise.race([deferred.promise, delayTimeout]).finally(() => {
        clearTimeout(timeoutId);
    });
}

285
286
287
288
/**
 * get the version of current package
 */
async function getVersion(): Promise<string> {
chicm-ms's avatar
chicm-ms committed
289
    const deferred: Deferred<string> = new Deferred<string>();
290
    import(path.join(__dirname, '..', 'package.json')).then((pkg) => {
291
        deferred.resolve(pkg.version);
292
293
    }).catch(() => {
        deferred.resolve('999.0.0-developing');
294
295
    });
    return deferred.promise;
296
}
297

298
299
300
/**
 * run command as ChildProcess
 */
J-shang's avatar
J-shang committed
301
function getTunerProc(command: string, stdio: StdioOptions, newCwd: string, newEnv: any, newShell: boolean = true, isDetached: boolean = false): ChildProcess {
302
303
    let cmd: string = command;
    let arg: string[] = [];
304
    if (process.platform === "win32") {
305
        cmd = command.split(" ", 1)[0];
306
        arg = command.substr(cmd.length + 1).split(" ");
307
        newShell = false;
308
        isDetached = true;
309
310
311
312
313
    }
    const tunerProc: ChildProcess = spawn(cmd, arg, {
        stdio,
        cwd: newCwd,
        env: newEnv,
314
315
        shell: newShell,
        detached: isDetached
316
317
318
319
320
321
322
    });
    return tunerProc;
}

/**
 * judge whether the process is alive
 */
Yuge Zhang's avatar
Yuge Zhang committed
323
async function isAlive(pid: any): Promise<boolean> {
chicm-ms's avatar
chicm-ms committed
324
    const deferred: Deferred<boolean> = new Deferred<boolean>();
325
    let alive: boolean = false;
Yuge Zhang's avatar
Yuge Zhang committed
326
    if (process.platform === 'win32') {
327
328
329
330
331
332
333
        try {
            const str = cp.execSync(`powershell.exe Get-Process -Id ${pid} -ErrorAction SilentlyContinue`).toString();
            if (str) {
                alive = true;
            }
        }
        catch (error) {
chicm-ms's avatar
chicm-ms committed
334
            //ignore
335
336
        }
    }
Yuge Zhang's avatar
Yuge Zhang committed
337
    else {
338
339
340
341
342
343
344
345
346
347
348
349
        try {
            await cpp.exec(`kill -0 ${pid}`);
            alive = true;
        } catch (error) {
            //ignore
        }
    }
    deferred.resolve(alive);
    return deferred.promise;
}

/**
350
 * kill process
351
 */
Yuge Zhang's avatar
Yuge Zhang committed
352
async function killPid(pid: any): Promise<void> {
chicm-ms's avatar
chicm-ms committed
353
    const deferred: Deferred<void> = new Deferred<void>();
354
355
    try {
        if (process.platform === "win32") {
Yuge Zhang's avatar
Yuge Zhang committed
356
            await cpp.exec(`cmd.exe /c taskkill /PID ${pid} /F`);
357
        }
358
        else {
359
360
361
362
363
364
365
366
367
            await cpp.exec(`kill -9 ${pid}`);
        }
    } catch (error) {
        // pid does not exist, do nothing here
    }
    deferred.resolve();
    return deferred.promise;
}

368
function getNewLine(): string {
369
370
371
    if (process.platform === "win32") {
        return "\r\n";
    }
372
    else {
373
374
375
376
        return "\n";
    }
}

377
378
/**
 * Use '/' to join path instead of '\' for all kinds of platform
379
 * @param path
380
381
382
383
384
385
386
 */
function unixPathJoin(...paths: any[]): string {
    const dir: string = paths.filter((path: any) => path !== '').join('/');
    if (dir === '') return '.';
    return dir;
}

387
388
389
390
391
392
393
394
395
/**
 * lock a file sync
 */
function withLockSync(func: Function, filePath: string, lockOpts: {[key: string]: any}, ...args: any): any {
    const lockName = path.join(path.dirname(filePath), path.basename(filePath) + `.lock.${process.pid}`);
    if (typeof lockOpts.stale === 'number'){
        const lockPath = path.join(path.dirname(filePath), path.basename(filePath) + '.lock.*');
        const lockFileNames: string[] = glob.sync(lockPath);
        const canLock: boolean = lockFileNames.map((fileName) => {
396
397
            return fs.existsSync(fileName) && Date.now() - fs.statSync(fileName).mtimeMs < lockOpts.stale;
        }).filter(unexpired=>unexpired === true).length === 0;
398
399
400
401
402
403
404
405
406
407
        if (!canLock) {
            throw new Error('File has been locked.');
        }
    }
    lockfile.lockSync(lockName, lockOpts);
    const result = func(...args);
    lockfile.unlockSync(lockName);
    return result;
}

J-shang's avatar
J-shang committed
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
async function isPortOpen(host: string, port: number): Promise<boolean> {
    return new Promise<boolean>((resolve, reject) => {
        try{
            const stream = net.createConnection(port, host);
            const id = setTimeout(() => {
                stream.destroy();
                resolve(false);
            }, 1000);

            stream.on('connect', () => {
                clearTimeout(id);
                stream.destroy();
                resolve(true);
            });

            stream.on('error', () => {
                clearTimeout(id);
                stream.destroy();
                resolve(false);
            });
        } catch (error) {
            reject(error);
        }
    });
}

async function getFreePort(host: string, start: number, end: number): Promise<number> {
    if (start > end) {
        throw new Error(`no more free port`);
    }
    if (await isPortOpen(host, start)) {
        return await getFreePort(host, start + 1, end);
    } else {
        return start;
    }
}

445
446
447
448
449
export function importModule(modulePath: string): any {
    module.paths.unshift(path.dirname(modulePath));
    return require(path.basename(modulePath));
}

450
export {
451
    countFilesRecursively, generateParamFileName, getMsgDispatcherCommand, getCheckpointDir, getExperimentsInfoPath,
J-shang's avatar
J-shang committed
452
    getLogDir, getExperimentRootDir, getJobCancelStatus, getDefaultDatabaseDir, getIPV4Address, unixPathJoin, withLockSync, getFreePort, isPortOpen,
453
454
    mkDirP, mkDirPSync, delay, prepareUnitTest, parseArg, cleanupUnitTest, uniqueString, randomInt, randomSelect, getLogLevel, getVersion, getCmdPy, getTunerProc, isAlive, killPid, getNewLine
};