main.ts 5.41 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
6
7

'use strict';

import { Container, Scope } from 'typescript-ioc';

8
import * as fs from 'fs';
9
import * as path from 'path';
10
import * as component from './common/component';
Deshui Yu's avatar
Deshui Yu committed
11
12
import { Database, DataStore } from './common/datastore';
import { setExperimentStartupInfo } from './common/experimentStartupInfo';
chicm-ms's avatar
chicm-ms committed
13
import { getLogger, Logger, logLevelNameMap } from './common/log';
SparkSnail's avatar
SparkSnail committed
14
import { Manager, ExperimentStartUpMode } from './common/manager';
15
import { ExperimentManager } from './common/experimentManager';
J-shang's avatar
J-shang committed
16
import { TensorboardManager } from './common/tensorboardManager';
17
import { getLogDir, mkDirP, parseArg } from './common/utils';
Deshui Yu's avatar
Deshui Yu committed
18
19
20
import { NNIDataStore } from './core/nniDataStore';
import { NNIManager } from './core/nnimanager';
import { SqlDB } from './core/sqlDatabase';
21
import { NNIExperimentsManager } from './core/nniExperimentsManager';
J-shang's avatar
J-shang committed
22
import { NNITensorboardManager } from './core/nniTensorboardManager';
23
import { NNIRestServer } from './rest_server/nniRestServer';
Deshui Yu's avatar
Deshui Yu committed
24

25

26
function initStartupInfo(
27
    startExpMode: string, experimentId: string, basePort: number, platform: string,
28
    logDirectory: string, experimentLogLevel: string, readonly: boolean, dispatcherPipe: string): void {
SparkSnail's avatar
SparkSnail committed
29
    const createNew: boolean = (startExpMode === ExperimentStartUpMode.NEW);
30
    setExperimentStartupInfo(createNew, experimentId, basePort, platform, logDirectory, experimentLogLevel, readonly, dispatcherPipe);
Deshui Yu's avatar
Deshui Yu committed
31
32
}

33
async function initContainer(foreground: boolean, platformMode: string, logFileName?: string): Promise<void> {
34
35
36
37
38
39
40
41
42
    Container.bind(Manager)
        .to(NNIManager)
        .scope(Scope.Singleton);
    Container.bind(Database)
        .to(SqlDB)
        .scope(Scope.Singleton);
    Container.bind(DataStore)
        .to(NNIDataStore)
        .scope(Scope.Singleton);
43
44
45
    Container.bind(ExperimentManager)
        .to(NNIExperimentsManager)
        .scope(Scope.Singleton);
J-shang's avatar
J-shang committed
46
47
48
    Container.bind(TensorboardManager)
        .to(NNITensorboardManager)
        .scope(Scope.Singleton);
49
50
51
52
53
54
    const DEFAULT_LOGFILE: string = path.join(getLogDir(), 'nnimanager.log');
    if (foreground) {
        logFileName = undefined;
    } else if (logFileName === undefined) {
        logFileName = DEFAULT_LOGFILE;
    }
SparkSnail's avatar
SparkSnail committed
55
56
57
    Container.bind(Logger).provider({
        get: (): Logger => new Logger(logFileName)
    });
Deshui Yu's avatar
Deshui Yu committed
58
59
60
61
62
63
    const ds: DataStore = component.get(DataStore);

    await ds.init();
}

function usage(): void {
64
    console.info('usage: node main.js --port <port> --mode \
SparkSnail's avatar
SparkSnail committed
65
    <local/remote/pai/kubeflow/frameworkcontroller/aml/adl/hybrid> --start_mode <new/resume> --experiment_id <id> --foreground <true/false>');
Deshui Yu's avatar
Deshui Yu committed
66
67
68
}

const strPort: string = parseArg(['--port', '-p']);
goooxu's avatar
goooxu committed
69
70
71
if (!strPort || strPort.length === 0) {
    usage();
    process.exit(1);
Deshui Yu's avatar
Deshui Yu committed
72
73
}

74
75
76
77
78
79
80
81
const foregroundArg: string = parseArg(['--foreground', '-f']);
if (!('true' || 'false').includes(foregroundArg.toLowerCase())) {
    console.log(`FATAL: foreground property should only be true or false`);
    usage();
    process.exit(1);
}
const foreground: boolean = foregroundArg.toLowerCase() === 'true' ? true : false;

goooxu's avatar
goooxu committed
82
83
const port: number = parseInt(strPort, 10);

Deshui Yu's avatar
Deshui Yu committed
84
const mode: string = parseArg(['--mode', '-m']);
SparkSnail's avatar
SparkSnail committed
85
if (!['local', 'remote', 'pai', 'kubeflow', 'frameworkcontroller', 'dlts', 'aml', 'adl', 'hybrid'].includes(mode)) {
86
    console.log(`FATAL: unknown mode: ${mode}`);
Deshui Yu's avatar
Deshui Yu committed
87
88
89
90
91
    usage();
    process.exit(1);
}

const startMode: string = parseArg(['--start_mode', '-s']);
SparkSnail's avatar
SparkSnail committed
92
if (![ExperimentStartUpMode.NEW, ExperimentStartUpMode.RESUME].includes(startMode)) {
93
    console.log(`FATAL: unknown start_mode: ${startMode}`);
Deshui Yu's avatar
Deshui Yu committed
94
95
96
97
98
    usage();
    process.exit(1);
}

const experimentId: string = parseArg(['--experiment_id', '-id']);
99
if (experimentId.trim().length < 1) {
SparkSnail's avatar
SparkSnail committed
100
    console.log(`FATAL: cannot resume the experiment, invalid experiment_id: ${experimentId}`);
Deshui Yu's avatar
Deshui Yu committed
101
102
103
104
    usage();
    process.exit(1);
}

105
106
107
108
109
110
111
112
const logDir: string = parseArg(['--log_dir', '-ld']);
if (logDir.length > 0) {
    if (!fs.existsSync(logDir)) {
        console.log(`FATAL: log_dir ${logDir} does not exist`);
    }
}

const logLevel: string = parseArg(['--log_level', '-ll']);
chicm-ms's avatar
chicm-ms committed
113
if (logLevel.length > 0 && !logLevelNameMap.has(logLevel)) {
114
115
116
    console.log(`FATAL: invalid log_level: ${logLevel}`);
}

SparkSnail's avatar
SparkSnail committed
117
118
119
120
121
122
123
124
const readonlyArg: string = parseArg(['--readonly', '-r']);
if (!('true' || 'false').includes(readonlyArg.toLowerCase())) {
    console.log(`FATAL: readonly property should only be true or false`);
    usage();
    process.exit(1);
}
const readonly = readonlyArg.toLowerCase() == 'true' ? true : false;

125
126
127
const dispatcherPipe: string = parseArg(['--dispatcher_pipe']);

initStartupInfo(startMode, experimentId, port, mode, logDir, logLevel, readonly, dispatcherPipe);
Deshui Yu's avatar
Deshui Yu committed
128

129
130
mkDirP(getLogDir())
    .then(async () => {
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
        try {
            await initContainer(foreground, mode);
            const restServer: NNIRestServer = component.get(NNIRestServer);
            await restServer.start();
            const log: Logger = getLogger();
            log.info(`Rest server listening on: ${restServer.endPoint}`);
        } catch (err) {
            const log: Logger = getLogger();
            log.error(`${err.stack}`);
            throw err;
        }
    })
    .catch((err: Error) => {
        console.error(`Failed to create log dir: ${err.stack}`);
    });
146

147
148
function cleanUp(): void {
    (component.get(Manager) as Manager).stopExperiment();
149
150
151
152
153
}

process.on('SIGTERM', cleanUp);
process.on('SIGBREAK', cleanUp);
process.on('SIGINT', cleanUp);