CspFileReader.py 12.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os
import time
import json
import glob
import logging
20
from multiprocessing import Lock
21

22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
""" Some terms to clarify the code
    in most case, one or more paremeters may be set as input args for a class or a function
    in form of single variable or k-v dict

    1.  trainerId
    2.  gpuId
    3.  rankId
    4.  gpuPerTrainer
    5.  groupSize
    6.  groupId
    7.  groupNum
    8.  displaySize
    9.  dataPath
    10. resultPath
    11. fileOrganizeForm -- "byRank" OR "byTrainer" or "other"

"""

PIPELINEINFO_TRACE_NUM = 1

dcgmMetricParameterMap = {
    "02_gpuUtility": [("GPUTL", "GPUTL"), ("GRACT", "GRACT")],
    "03_smUtility": [("SMACT", "SMACT"), ("SMOCC", "SMOCC")],
    "04_memUtility": [("FB_USED_RATIO", "FB_USED_RATIO"), ("DRAMA", "DRAMA")],
46 47 48 49 50 51 52 53 54 55 56
    "05_txUtility": [
        ("NVLTX", "NVLTX"),
        ("NVLRX", "NVLRX"),
        ("PCITX", "PCITX"),
        ("PCIRX", "PCIRX"),
    ],
    "06_calUtility": [
        ("FP32A", "FP32A"),
        ("FP16A", "FP16A"),
        ("TENSO", "TENSO"),
    ],
57 58 59 60 61 62 63 64 65 66 67 68 69
}
DCGMINFO_TRACE_NUM = len(dcgmMetricParameterMap.keys())
NETINFO_TRACE_NUM = 2

DCGM_PATH = "dcgm"
NET_PATH = "net"
TIME_PATH = "time"
PROFILE_PATH = "profile"

FILEORGANIZEFORM_BYRANK = "byRank"
FILEORGANIZEFORM_BYTRAINER = "byTrainer"
FILEORGANIZEFORM_BYOTHER = "other"
FILEORGANIZEFORM = [
70 71 72
    FILEORGANIZEFORM_BYRANK,
    FILEORGANIZEFORM_BYTRAINER,
    FILEORGANIZEFORM_BYOTHER,
73 74 75
]


76
class FileReader:
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
    def __init__(self, logger, args):
        self._logger = logger
        self._args = args

        self._fileList = []
        self._fileNum = 0

        self._dataPath = ""
        self._groupSize = 0
        self._displaySize = 0
        self._organizeForm = FILEORGANIZEFORM_BYOTHER
        self._gpuPerTrainer = 0

        self._checkArgs()
        self._getFileList()

        self._lock = Lock()

    def printArgs(self):
        self._logger.info("dataPath:")
        self._logger.info(self._dataPath)
        self._logger.info("groupSize:")
        self._logger.info(self._groupSize)
        self._logger.info("displaySize:")
        self._logger.info(self._displaySize)
        self._logger.info("organizeForm:")
        self._logger.info(self._organizeForm)
        self._logger.info("gpuPerTrainer:")
        self._logger.info(self._gpuPerTrainer)
        self._logger.info("minTimeStamp:")
        self._logger.info(self._minTimeStamp)

    def _checkArgsKey(self, key, type):
110
        if key not in self._args:
111 112 113 114
            raise KeyError("args should has key [%s]!" % key)

        if not isinstance(self._args[key], type):
            raise TypeError(
115 116 117
                "Invalid type of key [%s] in args dict, it should be a %s!"
                % (key, type)
            )
118 119 120 121 122 123 124 125 126 127 128

        exec("self._%s = self._args[\"%s\"]" % (key, key))

    def _align_ts(self, ts):
        return ts - self._minTimeStamp

    def _checkArgs(self):
        if not isinstance(self._args, dict):
            raise TypeError("Invalid type of args, it should be a dict!")

        self._checkArgsKey("organizeForm", str)
129 130 131 132
        if (
            self._organizeForm not in FILEORGANIZEFORM
            or self._organizeForm == FILEORGANIZEFORM_BYOTHER
        ):
133
            raise NotImplementedError(
134 135 136
                "we have not known how to process this form of file [%s]!"
                % self._organizeForm
            )
137 138 139 140 141

        self._checkArgsKey("gpuPerTrainer", int)

        self._checkArgsKey("dataPath", str)
        if not os.path.exists(self._dataPath):
142 143 144
            raise IOError(
                "input data path [%s] not existed!" % (self._dataPath)
            )
145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170

        self._checkArgsKey("groupSize", int)
        self._checkArgsKey("displaySize", int)
        self._checkArgsKey("minTimeStamp", int)

    def getFileListByGroup(self, groupId):
        lIndext = 0
        rIndext = 0

        if self._organizeForm == FILEORGANIZEFORM_BYTRAINER:
            lIndext = groupId * self._groupSize
            rIndext = (groupId + 1) * self._groupSize
        elif self._organizeForm == FILEORGANIZEFORM_BYRANK:
            lIndext = groupId * self._groupSize * self._gpuPerTrainer
            rIndext = (groupId + 1) * self._groupSize * self._gpuPerTrainer

        try:
            return self._fileList[lIndext:rIndext]
        except IndexError:
            raise IndexError("invalid index of file list")

    def getFileList(self):
        return self._getFileList

    def _cmp(self, x, y):
        return self._getId(x, self._organizeForm) - self._getId(
171 172
            y, self._organizeForm
        )
173 174 175 176 177 178 179 180 181 182 183 184 185 186

    def _getFileList(self):
        self._fileList = glob.glob(os.path.join(self._dataPath, "*.*"))

        # check unique
        idList = []
        newFileList = []
        for file in self._fileList:
            id = self._getId(file, self._organizeForm)
            if id not in idList:
                idList.append(id)
                newFileList.append(file)
            else:
                raise NotImplementedError(
187 188 189
                    "[%s] is repeated by id, we don not how to process it!"
                    % file
                )
190 191

        if not self._fileList:
192 193 194
            if (
                self._getId(self._fileList[-1]) - self._getId(self._fileList[0])
            ) != len(self._fileList) - 1:
195
                raise Exception("The file id should be countious!")
196

197 198 199 200 201 202 203
        # sort
        def _sortBySuffix(elem):
            return int(elem.split(".")[-1])

        self._fileList.sort(key=_sortBySuffix)

        if not self._fileList:
204 205 206
            self._logger.warning(
                "we can not find any file in dir [%s]!" % self._dataPath
            )
207
        else:
208 209 210 211
            self._logger.info(
                "file list in dir [%s] is : %s !"
                % (self._dataPath, ',  '.join(self._fileList))
            )
212 213 214 215 216

        return self._fileList

    def _getId(self, fileName, organizeForm, sed="."):
        if self._organizeForm != organizeForm:
217
            raise TypeError(
218 219 220
                "Can not get rank id when organizer form is not %s!"
                % organizeForm
            )
221 222 223 224 225 226 227 228

        if not os.path.isfile(fileName):
            raise IOError("[%s] is not a valid file!" % (fileName))

        try:
            prefix_str = fileName.split(sed)[-1]
            try:
                return int(prefix_str)
N
Nyakku Shigure 已提交
229 230
            except ValueError as e:
                print(e)
231 232
                raise TypeError("invalid fileName [%s]" % fileName)

N
Nyakku Shigure 已提交
233 234
        except IndexError as e:
            print(e)
235
            raise TypeError(
236 237 238
                "invalid fileName [%s], the prefix should be a number!"
                % fileName
            )
239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274

    def getRankId(self, fileName, sed="."):
        return self._getId(fileName, FILEORGANIZEFORM_BYRANK, sed)

    def getRankNum(self):
        if self._organizeForm == FILEORGANIZEFORM_BYRANK:
            return len(self._fileList)

        elif self._organizeForm == FILEORGANIZEFORM_BYTRAINER:
            return len(self._fileList) * self._gpuPerTrainer

    def getTrainerNum(self):
        if self._organizeForm == FILEORGANIZEFORM_BYRANK:
            return len(self._fileList) / self._gpuPerTrainer

        elif self._organizeForm == FILEORGANIZEFORM_BYTRAINER:
            return len(self._fileList)

    def getTrainerId(self, fileName, sed="."):
        return self._getId(fileName, FILEORGANIZEFORM_BYTRAINER, sed)

    def _splitTaskListForMultiProcess(self, ls, n):
        if not isinstance(ls, list) or not isinstance(n, int):
            return []
        ls_len = len(ls)
        if n <= 0 or 0 == ls_len:
            return []
        if n >= ls_len:
            return [[i] for i in ls]
        else:
            j = int((ls_len + n - 1) / n)
            k = ls_len % n
            ls_return = []
            end = 0
            for i in range(0, (n) * j, j):
                if i < len(ls) and (i + j) < len(ls):
275
                    ls_return.append(ls[i : i + j])
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307
                    end = i + j
            ls_return.append(ls[end:])
            return ls_return

    def getOpInfoFileName(self, groupId, gpuId, tmpPath="./tmp"):
        return self.getFileName("opinfo", groupId, gpuId, tmpPath)

    def getPipeLineInfoFileName(self, groupId, gpuId, tmpPath="./tmp"):
        return self.getFileName("pipilineinfo", groupId, gpuId, tmpPath)

    def getDCGMInfoFileName(self, groupId, gpuId, tmpPath="./tmp"):
        return self.getFileName("dcgm", groupId, gpuId, tmpPath)

    def getFileName(self, name, groupId, gpuId, tmpPath="./tmp"):
        return os.path.join(tmpPath, "%s_%d_%d.json" % (name, groupId, gpuId))

    def getOpInfoDict(self, groupId, gpuId, tmpPath="./tmp"):
        return self.getDict("opinfo", groupId, gpuId, tmpPath)

    def getDcgmInfoDict(self, groupId, gpuId, tmpPath="./tmp"):
        return self.getDict("dcgm", groupId, gpuId, tmpPath)

    def getDict(self, name, groupId, gpuId, tmpPath="./tmp"):
        fileName = self.getFileName(name, groupId, gpuId, tmpPath)
        if not os.path.isfile(fileName):
            raise IOError("[%s] is not existed!" % fileName)

        data = {}
        with open(fileName, "r") as rf:
            try:
                data = json.load(rf)
            except Exception:
308 309 310 311 312 313
                self._logger.error(
                    "read [%s] error. not a json file!" % (fileName)
                )
                raise TypeError(
                    "read [%s] error. not a json file!" % (fileName)
                )
314 315
        return data

316 317 318 319 320 321
    def dumpOpInfoDict(
        self, data, groupId, gpuId, pretty=False, tmpPath="./tmp"
    ):
        return self.dumpDict(
            data, "opinfo", groupId, gpuId, pretty=False, tmpPath="./tmp"
        )
322 323

    def dumpDCGMDict(self, data, groupId, gpuId, pretty=False, tmpPath="./tmp"):
324 325 326 327 328 329 330
        return self.dumpDict(
            data, "dcgm", groupId, gpuId, pretty=False, tmpPath="./tmp"
        )

    def dumpDict(
        self, data, name, groupId, gpuId, pretty=False, tmpPath="./tmp"
    ):
331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375
        self._lock.acquire()
        if not os.path.exists(tmpPath):
            os.makedirs(tmpPath)
        self._lock.release()
        if pretty:
            jsObj = json.dumps(data, indent=4, separators=(',', ': '))
        else:
            jsObj = json.dumps(data, separators=(',', ':'))

        fileName = self.getFileName(name, groupId, gpuId, tmpPath)
        if os.path.isfile(fileName):
            os.remove(fileName)

        fileObject = open(fileName, 'w')
        fileObject.write(jsObj)
        fileObject.close()
        self._logger.info("dump [%s] sucessfully!" % fileName)


def getLogger():
    logger = logging.getLogger()
    logger.setLevel(logging.DEBUG)

    rq = time.strftime('%Y%m%d%H%M.%s', time.localtime(time.time()))
    log_path = os.path.dirname(os.getcwd()) + '/Logs/'
    if not os.path.exists(log_path):
        os.makedirs(log_path)

    log_name = log_path + rq + '.log'
    logfile = log_name
    fh = logging.FileHandler(logfile, mode='w')
    fh.setLevel(logging.DEBUG)

    formatter = logging.Formatter(
        "%(asctime)s - %(filename)s[line:%(lineno)d] - %(process)d - %(levelname)s: %(message)s"
    )
    fh.setFormatter(formatter)

    logger.addHandler(fh)
    return logger


def test_FileReader(args):
    try:
        testReader = FileReader(None, args)
N
Nyakku Shigure 已提交
376 377
    except Exception as e:
        print(e)
378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
    else:
        testReader.printArgs()


if __name__ == "__main__":
    args = 0
    test_FileReader(args)

    args = {
        "dataPath": ".",
        "groupSize": 1,
        "displaySize": 1,
        "gpuPerTrainer": 8,
        "organizeForm": FILEORGANIZEFORM_BYOTHER,
    }
    test_FileReader(args)

    args = {
        "dataPath": ".",
        "groupSize": 1,
        "displaySize": 1,
        "gpuPerTrainer": 8,
        "organizeForm": FILEORGANIZEFORM_BYTRAINER,
    }
    test_FileReader(args)

    args = {
        "dataPath": "./res",
        "groupSize": 1,
        "displaySize": 1,
        "gpuPerTrainer": 8,
        "organizeForm": FILEORGANIZEFORM_BYTRAINER,
    }
    test_FileReader(args)

    args = {
        "dataPath": ".",
        "groupSize": "",
        "displaySize": 1,
        "gpuPerTrainer": 8,
        "organizeForm": FILEORGANIZEFORM_BYTRAINER,
    }
    test_FileReader(args)