wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-16 11:21:10 +08:00
Code Issues Packages Projects Releases Wiki Activity

Enforce quoting style in Travis. (#4589)

Browse Source
This commit is contained in:
justinwyang
2019-04-11 14:24:26 -07:00
committed by Robert Nishihara
parent 6697407ec4
commit e88e706fcc
79 changed files with 777 additions and 778 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ray/tune/automl/search_policy.py
+3 -3
View File
@@ -84,11 +84,11 @@ class AutoMLSearcher(SearchAlgorithm):
for exp in self.experiment_list:
for param_config, extra_arg in zip(raw_param_list, extra_arg_list):
tag = ''
tag = ""
new_spec = copy.deepcopy(exp.spec)
for path, value in param_config.items():
tag += '%s=%s-' % (path.split('.')[-1], value)
deep_insert(path.split('.'), value, new_spec['config'])
tag += "%s=%s-" % (path.split(".")[-1], value)
deep_insert(path.split("."), value, new_spec["config"])
trial = create_trial_from_spec(
new_spec, exp.name, self._parser, experiment_tag=tag)
python/ray/tune/automl/search_space.py
+1 -1
View File
@@ -67,7 +67,7 @@ class ContinuousSpace(ParameterSpace):
certain distribution such as linear.
"""
LINEAR = 'linear'
LINEAR = "linear"
# TODO: logspace
python/ray/tune/automlboard/backend/collector.py
+5 -5
View File
@@ -63,9 +63,9 @@ class CollectorService(object):
"""Initialize logger settings."""
logger = logging.getLogger("AutoMLBoard")
handler = logging.StreamHandler()
formatter = logging.Formatter('[%(levelname)s %(asctime)s] '
'%(filename)s: %(lineno)d '
'%(message)s')
formatter = logging.Formatter("[%(levelname)s %(asctime)s] "
"%(filename)s: %(lineno)d "
"%(message)s")
handler.setFormatter(formatter)
logger.setLevel(log_level)
logger.addHandler(handler)
@@ -294,7 +294,7 @@ class Collector(Thread):
meta = parse_json(meta_file)
if not meta:
job_name = job_dir.split('/')[-1]
job_name = job_dir.split("/")[-1]
user = os.environ.get("USER", None)
meta = {
"job_id": job_name,
@@ -325,7 +325,7 @@ class Collector(Thread):
meta = parse_json(meta_file)
if not meta:
job_id = expr_dir.split('/')[-2]
job_id = expr_dir.split("/")[-2]
trial_id = expr_dir[-8:]
params = parse_json(os.path.join(expr_dir, EXPR_PARARM_FILE))
meta = {
python/ray/tune/automlboard/common/utils.py
+7 -7
View File
@@ -19,9 +19,9 @@ def dump_json(json_info, json_file, overwrite=True):
overwrite(boolean)
"""
if overwrite:
mode = 'w'
mode = "w"
else:
mode = 'w+'
mode = "w+"
try:
with open(json_file, mode) as f:
@@ -45,7 +45,7 @@ def parse_json(json_file):
return None
try:
with open(json_file, 'r') as f:
with open(json_file, "r") as f:
info_str = f.readlines()
info_str = "".join(info_str)
json_info = json.loads(info_str)
@@ -76,11 +76,11 @@ def parse_multiple_json(json_file, offset=None):
return json_info_list
try:
with open(json_file, 'r') as f:
with open(json_file, "r") as f:
if offset:
f.seek(offset)
for line in f:
if line[-1] != '\n':
if line[-1] != "\n":
# Incomplete line
break
json_info = json.loads(line)
@@ -94,7 +94,7 @@ def parse_multiple_json(json_file, offset=None):
def timestamp2date(timestamp):
"""Convert a timestamp to date."""
return time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(timestamp))
return time.strftime("%Y-%m-%d %H:%M:%S", time.localtime(timestamp))
def unicode2str(content):
@@ -109,4 +109,4 @@ def unicode2str(content):
elif isinstance(content, int) or isinstance(content, float):
return content
else:
return content.encode('utf-8')
return content.encode("utf-8")
python/ray/tune/automlboard/frontend/query.py
+5 -5
View File
@@ -36,7 +36,7 @@ def query_job(request):
"success_trials": 4
}
"""
job_id = request.GET.get('job_id')
job_id = request.GET.get("job_id")
jobs = JobRecord.objects.filter(job_id=job_id)
trials = TrialRecord.objects.filter(job_id=job_id)
@@ -68,7 +68,7 @@ def query_job(request):
"progress": progress
}
resp = json.dumps(result)
return HttpResponse(resp, content_type='application/json;charset=utf-8')
return HttpResponse(resp, content_type="application/json;charset=utf-8")
def query_trial(request):
@@ -90,10 +90,10 @@ def query_trial(request):
"trial_id": "2067R2ZD",
}
"""
trial_id = request.GET.get('trial_id')
trial_id = request.GET.get("trial_id")
trials = TrialRecord.objects \
.filter(trial_id=trial_id) \
.order_by('-start_time')
.order_by("-start_time")
if len(trials) == 0:
resp = "Unkonwn trial id %s.\n" % trials
else:
@@ -107,4 +107,4 @@ def query_trial(request):
"params": trial.params
}
resp = json.dumps(result)
return HttpResponse(resp, content_type='application/json;charset=utf-8')
return HttpResponse(resp, content_type="application/json;charset=utf-8")
python/ray/tune/automlboard/frontend/urls.py
+6 -6
View File
@@ -29,10 +29,10 @@ import ray.tune.automlboard.frontend.view as view
import ray.tune.automlboard.frontend.query as query
urlpatterns = [
url(r'^admin/', admin.site.urls),
url(r'^$', view.index),
url(r'^job$', view.job),
url(r'^trial$', view.trial),
url(r'^query_job', query.query_job),
url(r'^query_trial', query.query_trial)
url(r"^admin/", admin.site.urls),
url(r"^$", view.index),
url(r"^job$", view.job),
url(r"^trial$", view.trial),
url(r"^query_job", query.query_job),
url(r"^query_trial", query.query_trial)
]
python/ray/tune/automlboard/frontend/view.py
+36 -36
View File
@@ -16,8 +16,8 @@ import datetime
def index(request):
"""View for the home page."""
recent_jobs = JobRecord.objects.order_by('-start_time')[0:100]
recent_trials = TrialRecord.objects.order_by('-start_time')[0:500]
recent_jobs = JobRecord.objects.order_by("-start_time")[0:100]
recent_trials = TrialRecord.objects.order_by("-start_time")[0:500]
total_num = len(recent_trials)
running_num = sum(t.trial_status == Trial.RUNNING for t in recent_trials)
@@ -29,31 +29,31 @@ def index(request):
for recent_job in recent_jobs:
job_records.append(get_job_info(recent_job))
context = {
'log_dir': AUTOMLBOARD_LOG_DIR,
'reload_interval': AUTOMLBOARD_RELOAD_INTERVAL,
'recent_jobs': job_records,
'job_num': len(job_records),
'trial_num': total_num,
'running_num': running_num,
'success_num': success_num,
'failed_num': failed_num
"log_dir": AUTOMLBOARD_LOG_DIR,
"reload_interval": AUTOMLBOARD_RELOAD_INTERVAL,
"recent_jobs": job_records,
"job_num": len(job_records),
"trial_num": total_num,
"running_num": running_num,
"success_num": success_num,
"failed_num": failed_num
}
return render(request, 'index.html', context)
return render(request, "index.html", context)
def job(request):
"""View for a single job."""
job_id = request.GET.get('job_id')
recent_jobs = JobRecord.objects.order_by('-start_time')[0:100]
job_id = request.GET.get("job_id")
recent_jobs = JobRecord.objects.order_by("-start_time")[0:100]
recent_trials = TrialRecord.objects \
.filter(job_id=job_id) \
.order_by('-start_time')
.order_by("-start_time")
trial_records = []
for recent_trial in recent_trials:
trial_records.append(get_trial_info(recent_trial))
current_job = JobRecord.objects \
.filter(job_id=job_id) \
.order_by('-start_time')[0]
.order_by("-start_time")[0]
if len(trial_records) > 0:
param_keys = trial_records[0]["params"].keys()
@@ -63,38 +63,38 @@ def job(request):
# TODO: support custom metrics here
metric_keys = ["episode_reward", "accuracy", "loss"]
context = {
'current_job': get_job_info(current_job),
'recent_jobs': recent_jobs,
'recent_trials': trial_records,
'param_keys': param_keys,
'param_num': len(param_keys),
'metric_keys': metric_keys,
'metric_num': len(metric_keys)
"current_job": get_job_info(current_job),
"recent_jobs": recent_jobs,
"recent_trials": trial_records,
"param_keys": param_keys,
"param_num": len(param_keys),
"metric_keys": metric_keys,
"metric_num": len(metric_keys)
}
return render(request, 'job.html', context)
return render(request, "job.html", context)
def trial(request):
"""View for a single trial."""
job_id = request.GET.get('job_id')
trial_id = request.GET.get('trial_id')
job_id = request.GET.get("job_id")
trial_id = request.GET.get("trial_id")
recent_trials = TrialRecord.objects \
.filter(job_id=job_id) \
.order_by('-start_time')
.order_by("-start_time")
recent_results = ResultRecord.objects \
.filter(trial_id=trial_id) \
.order_by('-date')[0:2000]
.order_by("-date")[0:2000]
current_trial = TrialRecord.objects \
.filter(trial_id=trial_id) \
.order_by('-start_time')[0]
.order_by("-start_time")[0]
context = {
'job_id': job_id,
'trial_id': trial_id,
'current_trial': current_trial,
'recent_results': recent_results,
'recent_trials': recent_trials
"job_id": job_id,
"trial_id": trial_id,
"current_trial": current_trial,
"recent_results": recent_results,
"recent_trials": recent_trials
}
return render(request, 'trial.html', context)
return render(request, "trial.html", context)
def get_job_info(current_job):
@@ -133,7 +133,7 @@ def get_job_info(current_job):
def get_trial_info(current_trial):
"""Get job information for current trial."""
if current_trial.end_time and ('_' in current_trial.end_time):
if current_trial.end_time and ("_" in current_trial.end_time):
# end time is parsed from result.json and the format
# is like: yyyy-mm-dd_hh-MM-ss, which will be converted
# to yyyy-mm-dd hh:MM:ss here
@@ -170,7 +170,7 @@ def get_winner(trials):
first_metrics = get_trial_info(trials[0])["metrics"]
if first_metrics and not first_metrics.get("accuracy", None):
sort_key = "episode_reward"
max_metric = float('-Inf')
max_metric = float("-Inf")
for t in trials:
metrics = get_trial_info(t).get("metrics", None)
if metrics and metrics.get(sort_key, None):
python/ray/tune/automlboard/models/apps.py
+1 -1
View File
@@ -9,4 +9,4 @@ from django.apps import AppConfig
class ModelConfig(AppConfig):
"""Model Congig for models."""
name = 'ray.tune.automlboard.models'
name = "ray.tune.automlboard.models"
python/ray/tune/automlboard/run.py
+3 -3
View File
@@ -37,8 +37,8 @@ def run_board(args):
# frontend service
logger.info("Try to start automlboard on port %s\n" % args.port)
command = [
os.path.join(root_path, 'manage.py'), 'runserver',
'0.0.0.0:%s' % args.port, '--noreload'
os.path.join(root_path, "manage.py"), "runserver",
"0.0.0.0:%s" % args.port, "--noreload"
]
execute_from_command_line(command)
@@ -76,7 +76,7 @@ def init_config(args):
os.environ.setdefault("DJANGO_SETTINGS_MODULE",
"ray.tune.automlboard.settings")
django.setup()
command = [os.path.join(root_path, 'manage.py'), 'migrate', '--run-syncdb']
command = [os.path.join(root_path, "manage.py"), "migrate", "--run-syncdb"]
execute_from_command_line(command)
python/ray/tune/automlboard/settings.py
+42 -42
View File
@@ -21,54 +21,54 @@ import os
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
# You can specify your own secret key, here we just pick one randomly.
SECRET_KEY = 'tktks103=$7a#5axn)52&b87!#w_qm(%*72^@hsq!nur%dtk4b'
SECRET_KEY = "tktks103=$7a#5axn)52&b87!#w_qm(%*72^@hsq!nur%dtk4b"
# SECURITY WARNING: don't run with debug turned on in production!
DEBUG = True
ALLOWED_HOSTS = ['*']
ALLOWED_HOSTS = ["*"]
# Application definition
INSTALLED_APPS = [
'django.contrib.admin',
'django.contrib.auth',
'django.contrib.contenttypes',
'django.contrib.sessions',
'django.contrib.messages',
'django.contrib.staticfiles',
'ray.tune.automlboard.models',
"django.contrib.admin",
"django.contrib.auth",
"django.contrib.contenttypes",
"django.contrib.sessions",
"django.contrib.messages",
"django.contrib.staticfiles",
"ray.tune.automlboard.models",
]
MIDDLEWARE = [
'django.middleware.security.SecurityMiddleware',
'django.contrib.sessions.middleware.SessionMiddleware',
'django.middleware.common.CommonMiddleware',
'django.middleware.csrf.CsrfViewMiddleware',
'django.contrib.auth.middleware.AuthenticationMiddleware',
'django.contrib.messages.middleware.MessageMiddleware',
'django.middleware.clickjacking.XFrameOptionsMiddleware',
"django.middleware.security.SecurityMiddleware",
"django.contrib.sessions.middleware.SessionMiddleware",
"django.middleware.common.CommonMiddleware",
"django.middleware.csrf.CsrfViewMiddleware",
"django.contrib.auth.middleware.AuthenticationMiddleware",
"django.contrib.messages.middleware.MessageMiddleware",
"django.middleware.clickjacking.XFrameOptionsMiddleware",
]
ROOT_URLCONF = 'ray.tune.automlboard.frontend.urls'
ROOT_URLCONF = "ray.tune.automlboard.frontend.urls"
TEMPLATES = [
{
'BACKEND': 'django.template.backends.django.DjangoTemplates',
'DIRS': [BASE_DIR + "/templates"],
'APP_DIRS': True,
'OPTIONS': {
'context_processors': [
'django.template.context_processors.debug',
'django.template.context_processors.request',
'django.contrib.auth.context_processors.auth',
'django.contrib.messages.context_processors.messages',
"BACKEND": "django.template.backends.django.DjangoTemplates",
"DIRS": [BASE_DIR + "/templates"],
"APP_DIRS": True,
"OPTIONS": {
"context_processors": [
"django.template.context_processors.debug",
"django.template.context_processors.request",
"django.contrib.auth.context_processors.auth",
"django.contrib.messages.context_processors.messages",
],
},
},
]
WSGI_APPLICATION = 'ray.tune.automlboard.frontend.wsgi.application'
WSGI_APPLICATION = "ray.tune.automlboard.frontend.wsgi.application"
DB_ENGINE_NAME_MAP = {
"mysql": "django.db.backends.mysql",
@@ -85,17 +85,17 @@ def lookup_db_engine(name):
# https://docs.djangoproject.com/en/1.11/ref/settings/#databases
if not os.environ.get("AUTOMLBOARD_DB_ENGINE", None):
DATABASES = {
'default': {
'ENGINE': 'django.db.backends.sqlite3',
'NAME': 'automlboard.db',
"default": {
"ENGINE": "django.db.backends.sqlite3",
"NAME": "automlboard.db",
}
}
else:
DATABASES = {
'default': {
'ENGINE': lookup_db_engine(os.environ["AUTOMLBOARD_DB_ENGINE"]),
'NAME': os.environ["AUTOMLBOARD_DB_NAME"],
'USER': os.environ["AUTOMLBOARD_DB_USER"],
"default": {
"ENGINE": lookup_db_engine(os.environ["AUTOMLBOARD_DB_ENGINE"]),
"NAME": os.environ["AUTOMLBOARD_DB_NAME"],
"USER": os.environ["AUTOMLBOARD_DB_USER"],
"PASSWORD": os.environ["AUTOMLBOARD_DB_PASSWORD"],
"HOST": os.environ["AUTOMLBOARD_DB_HOST"],
"PORT": os.environ["AUTOMLBOARD_DB_PORT"]
@@ -109,25 +109,25 @@ VALIDATION_PREFIX = "django.contrib.auth.password_validation."
AUTH_PASSWORD_VALIDATORS = [
{
'NAME': VALIDATION_PREFIX + "UserAttributeSimilarityValidator",
"NAME": VALIDATION_PREFIX + "UserAttributeSimilarityValidator",
},
{
'NAME': VALIDATION_PREFIX + "MinimumLengthValidator",
"NAME": VALIDATION_PREFIX + "MinimumLengthValidator",
},
{
'NAME': VALIDATION_PREFIX + "CommonPasswordValidator",
"NAME": VALIDATION_PREFIX + "CommonPasswordValidator",
},
{
'NAME': VALIDATION_PREFIX + "NumericPasswordValidator",
"NAME": VALIDATION_PREFIX + "NumericPasswordValidator",
},
]
# Internationalization
# https://docs.djangoproject.com/en/1.11/topics/i18n/
LANGUAGE_CODE = 'en-us'
LANGUAGE_CODE = "en-us"
TIME_ZONE = 'Asia/Shanghai'
TIME_ZONE = "Asia/Shanghai"
USE_I18N = True
@@ -138,8 +138,8 @@ USE_TZ = False
# Static files (CSS, JavaScript, Images)
# https://docs.djangoproject.com/en/1.11/howto/static-files/
STATIC_URL = '/static/'
STATICFILES_DIRS = (os.path.join(BASE_DIR, 'static').replace('\\', '/'), )
STATIC_URL = "/static/"
STATICFILES_DIRS = (os.path.join(BASE_DIR, "static").replace("\\", "/"), )
# automlboard settings
AUTOMLBOARD_LOG_DIR = os.environ.get("AUTOMLBOARD_LOGDIR", None)
python/ray/tune/commands.py
+14 -14
View File
@@ -53,12 +53,12 @@ except subprocess.CalledProcessError:
TERM_HEIGHT, TERM_WIDTH = 100, 100
OPERATORS = {
'<': operator.lt,
'<=': operator.le,
'==': operator.eq,
'!=': operator.ne,
'>=': operator.ge,
'>': operator.gt,
"<": operator.lt,
"<=": operator.le,
"==": operator.eq,
"!=": operator.ne,
">=": operator.ge,
">": operator.gt,
}
@@ -89,7 +89,7 @@ def print_format_output(dataframe):
print_df[col] = dataframe[col]
test_table = tabulate(print_df, headers="keys", tablefmt="psql")
if str(test_table).index('\n') > TERM_WIDTH:
if str(test_table).index("\n") > TERM_WIDTH:
# Drop all columns beyond terminal width
print_df.drop(col, axis=1, inplace=True)
dropped_cols += list(dataframe.columns)[i:]
@@ -172,10 +172,10 @@ def list_trials(experiment_path,
if "logdir" in checkpoints_df:
# logdir often too verbose to view in table, so drop experiment_path
checkpoints_df["logdir"] = checkpoints_df["logdir"].str.replace(
experiment_path, '')
experiment_path, "")
if filter_op:
col, op, val = filter_op.split(' ')
col, op, val = filter_op.split(" ")
col_type = checkpoints_df[col].dtype
if is_numeric_dtype(col_type):
val = float(val)
@@ -183,7 +183,7 @@ def list_trials(experiment_path,
val = str(val)
# TODO(Andrew): add support for datetime and boolean
else:
raise ValueError("Unsupported dtype for '{}': {}".format(
raise ValueError("Unsupported dtype for \"{}\": {}".format(
val, col_type))
op = OPERATORS[op]
filtered_index = op(checkpoints_df[col], val)
@@ -191,7 +191,7 @@ def list_trials(experiment_path,
if sort:
if sort not in checkpoints_df:
raise KeyError("Sort Index '{}' not in: {}".format(
raise KeyError("Sort Index \"{}\" not in: {}".format(
sort, list(checkpoints_df)))
checkpoints_df = checkpoints_df.sort_values(by=sort)
@@ -276,7 +276,7 @@ def list_experiments(project_path,
info_df = info_df[col_keys]
if filter_op:
col, op, val = filter_op.split(' ')
col, op, val = filter_op.split(" ")
col_type = info_df[col].dtype
if is_numeric_dtype(col_type):
val = float(val)
@@ -284,7 +284,7 @@ def list_experiments(project_path,
val = str(val)
# TODO(Andrew): add support for datetime and boolean
else:
raise ValueError("Unsupported dtype for '{}': {}".format(
raise ValueError("Unsupported dtype for \"{}\": {}".format(
val, col_type))
op = OPERATORS[op]
filtered_index = op(info_df[col], val)
@@ -292,7 +292,7 @@ def list_experiments(project_path,
if sort:
if sort not in info_df:
raise KeyError("Sort Index '{}' not in: {}".format(
raise KeyError("Sort Index \"{}\" not in: {}".format(
sort, list(info_df)))
info_df = info_df.sort_values(by=sort)
python/ray/tune/examples/bayesopt_example.py
+1 -1
View File
@@ -32,7 +32,7 @@ if __name__ == "__main__":
args, _ = parser.parse_known_args()
ray.init()
space = {'width': (0, 20), 'height': (-100, 100)}
space = {"width": (0, 20), "height": (-100, 100)}
config = {
"num_samples": 10 if args.smoke_test else 1000,
python/ray/tune/examples/genetic_example.py
+6 -6
View File
@@ -17,7 +17,7 @@ def michalewicz_function(config, reporter):
"""f(x) = -sum{sin(xi) * [sin(i*xi^2 / pi)]^(2m)}"""
import numpy as np
x = np.array(
[config['x1'], config['x2'], config['x3'], config['x4'], config['x5']])
[config["x1"], config["x2"], config["x3"], config["x4"], config["x5"]])
sin_x = np.sin(x)
z = (np.arange(1, 6) / np.pi * (x * x))
sin_z = np.power(np.sin(z), 20) # let m = 20
@@ -37,11 +37,11 @@ if __name__ == "__main__":
ray.init()
space = SearchSpace({
ContinuousSpace('x1', 0, 4, 100),
ContinuousSpace('x2', -2, 2, 100),
ContinuousSpace('x3', 1, 5, 100),
ContinuousSpace('x4', -3, 3, 100),
DiscreteSpace('x5', [-1, 0, 1, 2, 3]),
ContinuousSpace("x1", 0, 4, 100),
ContinuousSpace("x2", -2, 2, 100),
ContinuousSpace("x3", 1, 5, 100),
ContinuousSpace("x4", -3, 3, 100),
DiscreteSpace("x5", [-1, 0, 1, 2, 3]),
})
config = {"stop": {"training_iteration": 100}}
python/ray/tune/examples/hyperopt_example.py
+3 -3
View File
@@ -36,9 +36,9 @@ if __name__ == "__main__":
ray.init()
space = {
'width': hp.uniform('width', 0, 20),
'height': hp.uniform('height', -100, 100),
'activation': hp.choice("activation", ["relu", "tanh"])
"width": hp.uniform("width", 0, 20),
"height": hp.uniform("height", -100, 100),
"activation": hp.choice("activation", ["relu", "tanh"])
}
current_best_params = [
python/ray/tune/examples/mnist_pytorch.py
+28 -28
View File
@@ -10,50 +10,50 @@ import torch.optim as optim
from torchvision import datasets, transforms
# Training settings
parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
parser = argparse.ArgumentParser(description="PyTorch MNIST Example")
parser.add_argument(
'--batch-size',
"--batch-size",
type=int,
default=64,
metavar='N',
help='input batch size for training (default: 64)')
metavar="N",
help="input batch size for training (default: 64)")
parser.add_argument(
'--test-batch-size',
"--test-batch-size",
type=int,
default=1000,
metavar='N',
help='input batch size for testing (default: 1000)')
metavar="N",
help="input batch size for testing (default: 1000)")
parser.add_argument(
'--epochs',
"--epochs",
type=int,
default=1,
metavar='N',
help='number of epochs to train (default: 1)')
metavar="N",
help="number of epochs to train (default: 1)")
parser.add_argument(
'--lr',
"--lr",
type=float,
default=0.01,
metavar='LR',
help='learning rate (default: 0.01)')
metavar="LR",
help="learning rate (default: 0.01)")
parser.add_argument(
'--momentum',
"--momentum",
type=float,
default=0.5,
metavar='M',
help='SGD momentum (default: 0.5)')
metavar="M",
help="SGD momentum (default: 0.5)")
parser.add_argument(
'--no-cuda',
action='store_true',
"--no-cuda",
action="store_true",
default=False,
help='disables CUDA training')
help="disables CUDA training")
parser.add_argument(
'--seed',
"--seed",
type=int,
default=1,
metavar='S',
help='random seed (default: 1)')
metavar="S",
help="random seed (default: 1)")
parser.add_argument(
'--smoke-test', action="store_true", help="Finish quickly for testing")
"--smoke-test", action="store_true", help="Finish quickly for testing")
def train_mnist(args, config, reporter):
@@ -64,10 +64,10 @@ def train_mnist(args, config, reporter):
if args.cuda:
torch.cuda.manual_seed(args.seed)
kwargs = {'num_workers': 1, 'pin_memory': True} if args.cuda else {}
kwargs = {"num_workers": 1, "pin_memory": True} if args.cuda else {}
train_loader = torch.utils.data.DataLoader(
datasets.MNIST(
'~/data',
"~/data",
train=True,
download=False,
transform=transforms.Compose([
@@ -79,7 +79,7 @@ def train_mnist(args, config, reporter):
**kwargs)
test_loader = torch.utils.data.DataLoader(
datasets.MNIST(
'~/data',
"~/data",
train=False,
transform=transforms.Compose([
transforms.ToTensor(),
@@ -135,7 +135,7 @@ def train_mnist(args, config, reporter):
data, target = data.cuda(), target.cuda()
output = model(data)
# sum up batch loss
test_loss += F.nll_loss(output, target, reduction='sum').item()
test_loss += F.nll_loss(output, target, reduction="sum").item()
# get the index of the max log-probability
pred = output.argmax(dim=1, keepdim=True)
correct += pred.eq(
@@ -151,7 +151,7 @@ def train_mnist(args, config, reporter):
if __name__ == "__main__":
datasets.MNIST('~/data', train=True, download=True)
datasets.MNIST("~/data", train=True, download=True)
args = parser.parse_args()
import numpy as np
python/ray/tune/examples/mnist_pytorch_trainable.py
+28 -28
View File
@@ -13,50 +13,50 @@ from torchvision import datasets, transforms
from ray.tune import Trainable
# Training settings
parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
parser = argparse.ArgumentParser(description="PyTorch MNIST Example")
parser.add_argument(
'--batch-size',
"--batch-size",
type=int,
default=64,
metavar='N',
help='input batch size for training (default: 64)')
metavar="N",
help="input batch size for training (default: 64)")
parser.add_argument(
'--test-batch-size',
"--test-batch-size",
type=int,
default=1000,
metavar='N',
help='input batch size for testing (default: 1000)')
metavar="N",
help="input batch size for testing (default: 1000)")
parser.add_argument(
'--epochs',
"--epochs",
type=int,
default=1,
metavar='N',
help='number of epochs to train (default: 1)')
metavar="N",
help="number of epochs to train (default: 1)")
parser.add_argument(
'--lr',
"--lr",
type=float,
default=0.01,
metavar='LR',
help='learning rate (default: 0.01)')
metavar="LR",
help="learning rate (default: 0.01)")
parser.add_argument(
'--momentum',
"--momentum",
type=float,
default=0.5,
metavar='M',
help='SGD momentum (default: 0.5)')
metavar="M",
help="SGD momentum (default: 0.5)")
parser.add_argument(
'--no-cuda',
action='store_true',
"--no-cuda",
action="store_true",
default=False,
help='disables CUDA training')
help="disables CUDA training")
parser.add_argument(
'--seed',
"--seed",
type=int,
default=1,
metavar='S',
help='random seed (default: 1)')
metavar="S",
help="random seed (default: 1)")
parser.add_argument(
'--smoke-test', action="store_true", help="Finish quickly for testing")
"--smoke-test", action="store_true", help="Finish quickly for testing")
class Net(nn.Module):
@@ -88,10 +88,10 @@ class TrainMNIST(Trainable):
if args.cuda:
torch.cuda.manual_seed(args.seed)
kwargs = {'num_workers': 1, 'pin_memory': True} if args.cuda else {}
kwargs = {"num_workers": 1, "pin_memory": True} if args.cuda else {}
self.train_loader = torch.utils.data.DataLoader(
datasets.MNIST(
'~/data',
"~/data",
train=True,
download=False,
transform=transforms.Compose([
@@ -103,7 +103,7 @@ class TrainMNIST(Trainable):
**kwargs)
self.test_loader = torch.utils.data.DataLoader(
datasets.MNIST(
'~/data',
"~/data",
train=False,
transform=transforms.Compose([
transforms.ToTensor(),
@@ -142,7 +142,7 @@ class TrainMNIST(Trainable):
data, target = data.cuda(), target.cuda()
output = self.model(data)
# sum up batch loss
test_loss += F.nll_loss(output, target, reduction='sum').item()
test_loss += F.nll_loss(output, target, reduction="sum").item()
# get the index of the max log-probability
pred = output.argmax(dim=1, keepdim=True)
correct += pred.eq(
@@ -166,7 +166,7 @@ class TrainMNIST(Trainable):
if __name__ == "__main__":
datasets.MNIST('~/data', train=True, download=True)
datasets.MNIST("~/data", train=True, download=True)
args = parser.parse_args()
import numpy as np
python/ray/tune/examples/sigopt_example.py
+10 -10
View File
@@ -38,19 +38,19 @@ if __name__ == "__main__":
space = [
{
'name': 'width',
'type': 'int',
'bounds': {
'min': 0,
'max': 20
"name": "width",
"type": "int",
"bounds": {
"min": 0,
"max": 20
},
},
{
'name': 'height',
'type': 'int',
'bounds': {
'min': -100,
'max': 100
"name": "height",
"type": "int",
"bounds": {
"min": -100,
"max": 100
},
},
]
python/ray/tune/examples/tune_mnist_async_hyperband.py
+31 -31
View File
@@ -60,32 +60,32 @@ def deepnn(x):
# Reshape to use within a convolutional neural net.
# Last dimension is for "features" - there is only one here, since images
# are grayscale -- it would be 3 for an RGB image, 4 for RGBA, etc.
with tf.name_scope('reshape'):
with tf.name_scope("reshape"):
x_image = tf.reshape(x, [-1, 28, 28, 1])
# First convolutional layer - maps one grayscale image to 32 feature maps.
with tf.name_scope('conv1'):
with tf.name_scope("conv1"):
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
h_conv1 = activation_fn(conv2d(x_image, W_conv1) + b_conv1)
# Pooling layer - downsamples by 2X.
with tf.name_scope('pool1'):
with tf.name_scope("pool1"):
h_pool1 = max_pool_2x2(h_conv1)
# Second convolutional layer -- maps 32 feature maps to 64.
with tf.name_scope('conv2'):
with tf.name_scope("conv2"):
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = activation_fn(conv2d(h_pool1, W_conv2) + b_conv2)
# Second pooling layer.
with tf.name_scope('pool2'):
with tf.name_scope("pool2"):
h_pool2 = max_pool_2x2(h_conv2)
# Fully connected layer 1 -- after 2 round of downsampling, our 28x28 image
# is down to 7x7x64 feature maps -- maps this to 1024 features.
with tf.name_scope('fc1'):
with tf.name_scope("fc1"):
W_fc1 = weight_variable([7 * 7 * 64, 1024])
b_fc1 = bias_variable([1024])
@@ -94,12 +94,12 @@ def deepnn(x):
# Dropout - controls the complexity of the model, prevents co-adaptation of
# features.
with tf.name_scope('dropout'):
with tf.name_scope("dropout"):
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
# Map the 1024 features to 10 classes, one for each digit
with tf.name_scope('fc2'):
with tf.name_scope("fc2"):
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
@@ -109,13 +109,13 @@ def deepnn(x):
def conv2d(x, W):
"""conv2d returns a 2d convolution layer with full stride."""
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding="SAME")
def max_pool_2x2(x):
"""max_pool_2x2 downsamples a feature map by 2X."""
return tf.nn.max_pool(
x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
def weight_variable(shape):
@@ -148,21 +148,21 @@ def main(_):
# Build the graph for the deep net
y_conv, keep_prob = deepnn(x)
with tf.name_scope('loss'):
with tf.name_scope("loss"):
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(
labels=y_, logits=y_conv)
cross_entropy = tf.reduce_mean(cross_entropy)
with tf.name_scope('adam_optimizer'):
with tf.name_scope("adam_optimizer"):
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
with tf.name_scope('accuracy'):
with tf.name_scope("accuracy"):
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
correct_prediction = tf.cast(correct_prediction, tf.float32)
accuracy = tf.reduce_mean(correct_prediction)
graph_location = tempfile.mkdtemp()
print('Saving graph to: %s' % graph_location)
print("Saving graph to: %s" % graph_location)
train_writer = tf.summary.FileWriter(graph_location)
train_writer.add_graph(tf.get_default_graph())
@@ -182,14 +182,14 @@ def main(_):
status_reporter(
timesteps_total=i, mean_accuracy=train_accuracy)
print('step %d, training accuracy %g' % (i, train_accuracy))
print("step %d, training accuracy %g" % (i, train_accuracy))
train_step.run(feed_dict={
x: batch[0],
y_: batch[1],
keep_prob: 0.5
})
print('test accuracy %g' % accuracy.eval(feed_dict={
print("test accuracy %g" % accuracy.eval(feed_dict={
x: mnist.test.images,
y_: mnist.test.labels,
keep_prob: 1.0
@@ -197,16 +197,16 @@ def main(_):
# !!! Entrypoint for ray.tune !!!
def train(config={'activation': 'relu'}, reporter=None):
def train(config={"activation": "relu"}, reporter=None):
global FLAGS, status_reporter, activation_fn
status_reporter = reporter
activation_fn = getattr(tf.nn, config['activation'])
activation_fn = getattr(tf.nn, config["activation"])
parser = argparse.ArgumentParser()
parser.add_argument(
'--data_dir',
"--data_dir",
type=str,
default='/tmp/tensorflow/mnist/input_data',
help='Directory for storing input data')
default="/tmp/tensorflow/mnist/input_data",
help="Directory for storing input data")
FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
@@ -215,29 +215,29 @@ def train(config={'activation': 'relu'}, reporter=None):
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
'--smoke-test', action='store_true', help='Finish quickly for testing')
"--smoke-test", action="store_true", help="Finish quickly for testing")
args, _ = parser.parse_known_args()
mnist_spec = {
'num_samples': 10,
'stop': {
'mean_accuracy': 0.99,
'timesteps_total': 600,
"num_samples": 10,
"stop": {
"mean_accuracy": 0.99,
"timesteps_total": 600,
},
'config': {
'activation': grid_search(['relu', 'elu', 'tanh']),
"config": {
"activation": grid_search(["relu", "elu", "tanh"]),
},
}
if args.smoke_test:
mnist_spec['stop']['training_iteration'] = 2
mnist_spec['num_samples'] = 1
mnist_spec["stop"]["training_iteration"] = 2
mnist_spec["num_samples"] = 1
ray.init()
from ray.tune.schedulers import AsyncHyperBandScheduler
run(train,
name='tune_mnist_test',
name="tune_mnist_test",
scheduler=AsyncHyperBandScheduler(
time_attr="timesteps_total",
reward_attr="mean_accuracy",
python/ray/tune/examples/tune_mnist_keras.py
+36 -36
View File
@@ -50,7 +50,7 @@ def train_mnist(args, cfg, reporter):
# the data, split between train and test sets
(x_train, y_train), (x_test, y_test) = mnist.load_data()
if K.image_data_format() == 'channels_first':
if K.image_data_format() == "channels_first":
x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)
x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
input_shape = (1, img_rows, img_cols)
@@ -59,13 +59,13 @@ def train_mnist(args, cfg, reporter):
x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
input_shape = (img_rows, img_cols, 1)
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train = x_train.astype("float32")
x_test = x_test.astype("float32")
x_train /= 255
x_test /= 255
print('x_train shape:', x_train.shape)
print(x_train.shape[0], 'train samples')
print(x_test.shape[0], 'test samples')
print("x_train shape:", x_train.shape)
print(x_train.shape[0], "train samples")
print(x_test.shape[0], "test samples")
# convert class vectors to binary class matrices
y_train = keras.utils.to_categorical(y_train, num_classes)
@@ -76,20 +76,20 @@ def train_mnist(args, cfg, reporter):
Conv2D(
32,
kernel_size=(args.kernel1, args.kernel1),
activation='relu',
activation="relu",
input_shape=input_shape))
model.add(Conv2D(64, (args.kernel2, args.kernel2), activation='relu'))
model.add(Conv2D(64, (args.kernel2, args.kernel2), activation="relu"))
model.add(MaxPooling2D(pool_size=(args.poolsize, args.poolsize)))
model.add(Dropout(args.dropout1))
model.add(Flatten())
model.add(Dense(args.hidden, activation='relu'))
model.add(Dense(args.hidden, activation="relu"))
model.add(Dropout(args.dropout2))
model.add(Dense(num_classes, activation='softmax'))
model.add(Dense(num_classes, activation="softmax"))
model.compile(
loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.SGD(lr=args.lr, momentum=args.momentum),
metrics=['accuracy'])
metrics=["accuracy"])
model.fit(
x_train,
@@ -102,66 +102,66 @@ def train_mnist(args, cfg, reporter):
def create_parser():
parser = argparse.ArgumentParser(description='Keras MNIST Example')
parser = argparse.ArgumentParser(description="Keras MNIST Example")
parser.add_argument(
"--smoke-test", action="store_true", help="Finish quickly for testing")
parser.add_argument(
"--use-gpu", action="store_true", help="Use GPU in training.")
parser.add_argument(
'--jobs',
"--jobs",
type=int,
default=1,
help='number of jobs to run concurrently (default: 1)')
help="number of jobs to run concurrently (default: 1)")
parser.add_argument(
'--threads',
"--threads",
type=int,
default=2,
help='threads used in operations (default: 2)')
help="threads used in operations (default: 2)")
parser.add_argument(
'--steps',
"--steps",
type=float,
default=0.01,
metavar='LR',
help='learning rate (default: 0.01)')
metavar="LR",
help="learning rate (default: 0.01)")
parser.add_argument(
'--lr',
"--lr",
type=float,
default=0.01,
metavar='LR',
help='learning rate (default: 0.01)')
metavar="LR",
help="learning rate (default: 0.01)")
parser.add_argument(
'--momentum',
"--momentum",
type=float,
default=0.5,
metavar='M',
help='SGD momentum (default: 0.5)')
metavar="M",
help="SGD momentum (default: 0.5)")
parser.add_argument(
'--kernel1',
"--kernel1",
type=int,
default=3,
help='Size of first kernel (default: 3)')
help="Size of first kernel (default: 3)")
parser.add_argument(
'--kernel2',
"--kernel2",
type=int,
default=3,
help='Size of second kernel (default: 3)')
help="Size of second kernel (default: 3)")
parser.add_argument(
'--poolsize', type=int, default=2, help='Size of Pooling (default: 2)')
"--poolsize", type=int, default=2, help="Size of Pooling (default: 2)")
parser.add_argument(
'--dropout1',
"--dropout1",
type=float,
default=0.25,
help='Size of first kernel (default: 0.25)')
help="Size of first kernel (default: 0.25)")
parser.add_argument(
'--hidden',
"--hidden",
type=int,
default=128,
help='Size of Hidden Layer (default: 128)')
help="Size of Hidden Layer (default: 128)")
parser.add_argument(
'--dropout2',
"--dropout2",
type=float,
default=0.5,
help='Size of first kernel (default: 0.5)')
help="Size of first kernel (default: 0.5)")
return parser
python/ray/tune/examples/tune_mnist_ray.py
+31 -31
View File
@@ -62,32 +62,32 @@ def deepnn(x):
# Reshape to use within a convolutional neural net.
# Last dimension is for "features" - there is only one here, since images
# are grayscale -- it would be 3 for an RGB image, 4 for RGBA, etc.
with tf.name_scope('reshape'):
with tf.name_scope("reshape"):
x_image = tf.reshape(x, [-1, 28, 28, 1])
# First convolutional layer - maps one grayscale image to 32 feature maps.
with tf.name_scope('conv1'):
with tf.name_scope("conv1"):
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
h_conv1 = activation_fn(conv2d(x_image, W_conv1) + b_conv1)
# Pooling layer - downsamples by 2X.
with tf.name_scope('pool1'):
with tf.name_scope("pool1"):
h_pool1 = max_pool_2x2(h_conv1)
# Second convolutional layer -- maps 32 feature maps to 64.
with tf.name_scope('conv2'):
with tf.name_scope("conv2"):
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = activation_fn(conv2d(h_pool1, W_conv2) + b_conv2)
# Second pooling layer.
with tf.name_scope('pool2'):
with tf.name_scope("pool2"):
h_pool2 = max_pool_2x2(h_conv2)
# Fully connected layer 1 -- after 2 round of downsampling, our 28x28 image
# is down to 7x7x64 feature maps -- maps this to 1024 features.
with tf.name_scope('fc1'):
with tf.name_scope("fc1"):
W_fc1 = weight_variable([7 * 7 * 64, 1024])
b_fc1 = bias_variable([1024])
@@ -96,12 +96,12 @@ def deepnn(x):
# Dropout - controls the complexity of the model, prevents co-adaptation of
# features.
with tf.name_scope('dropout'):
with tf.name_scope("dropout"):
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
# Map the 1024 features to 10 classes, one for each digit
with tf.name_scope('fc2'):
with tf.name_scope("fc2"):
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
@@ -111,13 +111,13 @@ def deepnn(x):
def conv2d(x, W):
"""conv2d returns a 2d convolution layer with full stride."""
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding="SAME")
def max_pool_2x2(x):
"""max_pool_2x2 downsamples a feature map by 2X."""
return tf.nn.max_pool(
x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
def weight_variable(shape):
@@ -150,21 +150,21 @@ def main(_):
# Build the graph for the deep net
y_conv, keep_prob = deepnn(x)
with tf.name_scope('loss'):
with tf.name_scope("loss"):
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(
labels=y_, logits=y_conv)
cross_entropy = tf.reduce_mean(cross_entropy)
with tf.name_scope('adam_optimizer'):
with tf.name_scope("adam_optimizer"):
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
with tf.name_scope('accuracy'):
with tf.name_scope("accuracy"):
correct_prediction = tf.equal(tf.argmax(y_conv, 1), tf.argmax(y_, 1))
correct_prediction = tf.cast(correct_prediction, tf.float32)
accuracy = tf.reduce_mean(correct_prediction)
graph_location = tempfile.mkdtemp()
print('Saving graph to: %s' % graph_location)
print("Saving graph to: %s" % graph_location)
train_writer = tf.summary.FileWriter(graph_location)
train_writer.add_graph(tf.get_default_graph())
@@ -184,14 +184,14 @@ def main(_):
status_reporter(
timesteps_total=i, mean_accuracy=train_accuracy)
print('step %d, training accuracy %g' % (i, train_accuracy))
print("step %d, training accuracy %g" % (i, train_accuracy))
train_step.run(feed_dict={
x: batch[0],
y_: batch[1],
keep_prob: 0.5
})
print('test accuracy %g' % accuracy.eval(feed_dict={
print("test accuracy %g" % accuracy.eval(feed_dict={
x: mnist.test.images,
y_: mnist.test.labels,
keep_prob: 1.0
@@ -199,16 +199,16 @@ def main(_):
# !!! Entrypoint for ray.tune !!!
def train(config={'activation': 'relu'}, reporter=None):
def train(config={"activation": "relu"}, reporter=None):
global FLAGS, status_reporter, activation_fn
status_reporter = reporter
activation_fn = getattr(tf.nn, config['activation'])
activation_fn = getattr(tf.nn, config["activation"])
parser = argparse.ArgumentParser()
parser.add_argument(
'--data_dir',
"--data_dir",
type=str,
default='/tmp/tensorflow/mnist/input_data',
help='Directory for storing input data')
default="/tmp/tensorflow/mnist/input_data",
help="Directory for storing input data")
FLAGS, unparsed = parser.parse_known_args()
tf.app.run(main=main, argv=[sys.argv[0]] + unparsed)
@@ -217,25 +217,25 @@ def train(config={'activation': 'relu'}, reporter=None):
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
'--smoke-test', action='store_true', help='Finish quickly for testing')
"--smoke-test", action="store_true", help="Finish quickly for testing")
args, _ = parser.parse_known_args()
register_trainable('train_mnist', train)
register_trainable("train_mnist", train)
mnist_spec = {
'stop': {
'mean_accuracy': 0.99,
'time_total_s': 600,
"stop": {
"mean_accuracy": 0.99,
"time_total_s": 600,
},
'config': {
'activation': grid_search(['relu', 'elu', 'tanh']),
"config": {
"activation": grid_search(["relu", "elu", "tanh"]),
# You can pass any serializable object as well
'foo': grid_search([np.array([1, 2]),
"foo": grid_search([np.array([1, 2]),
np.array([2, 3])]),
},
}
if args.smoke_test:
mnist_spec['stop']['training_iteration'] = 2
mnist_spec["stop"]["training_iteration"] = 2
ray.init()
tune.run('train_mnist', name='tune_mnist_test', **mnist_spec)
tune.run("train_mnist", name="tune_mnist_test", **mnist_spec)
python/ray/tune/examples/tune_mnist_ray_hyperband.py
+25 -25
View File
@@ -55,32 +55,32 @@ def setupCNN(x):
# Reshape to use within a convolutional neural net.
# Last dimension is for "features" - there is only one here, since images
# are grayscale -- it would be 3 for an RGB image, 4 for RGBA, etc.
with tf.name_scope('reshape'):
with tf.name_scope("reshape"):
x_image = tf.reshape(x, [-1, 28, 28, 1])
# First convolutional layer - maps one grayscale image to 32 feature maps.
with tf.name_scope('conv1'):
with tf.name_scope("conv1"):
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
h_conv1 = activation_fn(conv2d(x_image, W_conv1) + b_conv1)
# Pooling layer - downsamples by 2X.
with tf.name_scope('pool1'):
with tf.name_scope("pool1"):
h_pool1 = max_pool_2x2(h_conv1)
# Second convolutional layer -- maps 32 feature maps to 64.
with tf.name_scope('conv2'):
with tf.name_scope("conv2"):
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = activation_fn(conv2d(h_pool1, W_conv2) + b_conv2)
# Second pooling layer.
with tf.name_scope('pool2'):
with tf.name_scope("pool2"):
h_pool2 = max_pool_2x2(h_conv2)
# Fully connected layer 1 -- after 2 round of downsampling, our 28x28 image
# is down to 7x7x64 feature maps -- maps this to 1024 features.
with tf.name_scope('fc1'):
with tf.name_scope("fc1"):
W_fc1 = weight_variable([7 * 7 * 64, 1024])
b_fc1 = bias_variable([1024])
@@ -89,12 +89,12 @@ def setupCNN(x):
# Dropout - controls the complexity of the model, prevents co-adaptation of
# features.
with tf.name_scope('dropout'):
with tf.name_scope("dropout"):
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
# Map the 1024 features to 10 classes, one for each digit
with tf.name_scope('fc2'):
with tf.name_scope("fc2"):
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
@@ -104,13 +104,13 @@ def setupCNN(x):
def conv2d(x, W):
"""conv2d returns a 2d convolution layer with full stride."""
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding='SAME')
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding="SAME")
def max_pool_2x2(x):
"""max_pool_2x2 downsamples a feature map by 2X."""
return tf.nn.max_pool(
x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")
def weight_variable(shape):
@@ -148,23 +148,23 @@ class TrainMNIST(Trainable):
self.x = tf.placeholder(tf.float32, [None, 784])
self.y_ = tf.placeholder(tf.float32, [None, 10])
activation_fn = getattr(tf.nn, config['activation'])
activation_fn = getattr(tf.nn, config["activation"])
# Build the graph for the deep net
y_conv, self.keep_prob = setupCNN(self.x)
with tf.name_scope('loss'):
with tf.name_scope("loss"):
cross_entropy = tf.nn.softmax_cross_entropy_with_logits(
labels=self.y_, logits=y_conv)
cross_entropy = tf.reduce_mean(cross_entropy)
with tf.name_scope('adam_optimizer'):
with tf.name_scope("adam_optimizer"):
train_step = tf.train.AdamOptimizer(
config['learning_rate']).minimize(cross_entropy)
config["learning_rate"]).minimize(cross_entropy)
self.train_step = train_step
with tf.name_scope('accuracy'):
with tf.name_scope("accuracy"):
correct_prediction = tf.equal(
tf.argmax(y_conv, 1), tf.argmax(self.y_, 1))
correct_prediction = tf.cast(correct_prediction, tf.float32)
@@ -212,24 +212,24 @@ class TrainMNIST(Trainable):
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
'--smoke-test', action='store_true', help='Finish quickly for testing')
"--smoke-test", action="store_true", help="Finish quickly for testing")
args, _ = parser.parse_known_args()
mnist_spec = {
'stop': {
'mean_accuracy': 0.99,
'time_total_s': 600,
"stop": {
"mean_accuracy": 0.99,
"time_total_s": 600,
},
'config': {
'learning_rate': sample_from(
"config": {
"learning_rate": sample_from(
lambda spec: 10**np.random.uniform(-5, -3)),
'activation': grid_search(['relu', 'elu', 'tanh']),
"activation": grid_search(["relu", "elu", "tanh"]),
},
"num_samples": 10,
}
if args.smoke_test:
mnist_spec['stop']['training_iteration'] = 20
mnist_spec['num_samples'] = 2
mnist_spec["stop"]["training_iteration"] = 20
mnist_spec["num_samples"] = 2
ray.init()
hyperband = HyperBandScheduler(
@@ -237,6 +237,6 @@ if __name__ == "__main__":
tune.run(
TrainMNIST,
name='mnist_hyperband_test',
name="mnist_hyperband_test",
scheduler=hyperband,
**mnist_spec)
python/ray/tune/log_sync.py
+4 -4
View File
@@ -148,8 +148,8 @@ class _LogSyncer(object):
if not distutils.spawn.find_executable("rsync"):
logger.error("Log sync requires rsync to be installed.")
return
source = '{}/'.format(self.local_dir)
target = '{}@{}:{}/'.format(ssh_user, self.worker_ip, self.local_dir)
source = "{}/".format(self.local_dir)
target = "{}@{}:{}/".format(ssh_user, self.worker_ip, self.local_dir)
final_cmd = (("""rsync -savz -e "ssh -i {} -o ConnectTimeout=120s """
"""-o StrictHostKeyChecking=no" {} {}""").format(
quote(ssh_key), quote(source), quote(target)))
@@ -180,9 +180,9 @@ class _LogSyncer(object):
if not distutils.spawn.find_executable("rsync"):
logger.error("Log sync requires rsync to be installed.")
return
source = '{}@{}:{}/'.format(ssh_user, self.worker_ip,
source = "{}@{}:{}/".format(ssh_user, self.worker_ip,
self.local_dir)
target = '{}/'.format(self.local_dir)
target = "{}/".format(self.local_dir)
worker_to_local_sync_cmd = ((
"""rsync -savz -e "ssh -i {} -o ConnectTimeout=120s """
"""-o StrictHostKeyChecking=no" {} {}""").format(
python/ray/tune/ray_trial_executor.py
+1 -1
View File
@@ -166,7 +166,7 @@ class RayTrialExecutor(TrialExecutor):
try:
trial.write_error_log(error_msg)
if hasattr(trial, 'runner') and trial.runner:
if hasattr(trial, "runner") and trial.runner:
if (not error and self._reuse_actors
and self._cached_actor is None):
logger.debug("Reusing actor for {}".format(trial.runner))
python/ray/tune/schedulers/async_hyperband.py
+2 -2
View File
@@ -39,8 +39,8 @@ class AsyncHyperBandScheduler(FIFOScheduler):
"""
def __init__(self,
time_attr='training_iteration',
reward_attr='episode_reward_mean',
time_attr="training_iteration",
reward_attr="episode_reward_mean",
max_t=100,
grace_period=10,
reduction_factor=3,
python/ray/tune/schedulers/hyperband.py
+2 -2
View File
@@ -73,8 +73,8 @@ class HyperBandScheduler(FIFOScheduler):
"""
def __init__(self,
time_attr='training_iteration',
reward_attr='episode_reward_mean',
time_attr="training_iteration",
reward_attr="episode_reward_mean",
max_t=81):
assert max_t > 0, "Max (time_attr) not valid!"
FIFOScheduler.__init__(self)
python/ray/tune/schedulers/median_stopping_rule.py
+2 -2
View File
@@ -104,9 +104,9 @@ class MedianStoppingRule(FIFOScheduler):
if len(scores) >= self._min_samples_required:
return np.median(scores)
else:
return float('-inf')
return float("-inf")
def _running_result(self, trial, t_max=float('inf')):
def _running_result(self, trial, t_max=float("inf")):
results = self._results[trial]
# TODO(ekl) we could do interpolation to be more precise, but for now
# assume len(results) is large and the time diffs are roughly equal
python/ray/tune/scripts.py
+3 -3
View File
@@ -43,9 +43,9 @@ def list_trials(experiment_path, sort, output, filter_op, columns,
result_columns):
"""Lists trials in the directory subtree starting at the given path."""
if columns:
columns = columns.split(',')
columns = columns.split(",")
if result_columns:
result_columns = result_columns.split(',')
result_columns = result_columns.split(",")
commands.list_trials(experiment_path, sort, output, filter_op, columns,
result_columns)
@@ -75,7 +75,7 @@ def list_trials(experiment_path, sort, output, filter_op, columns,
def list_experiments(project_path, sort, output, filter_op, columns):
"""Lists experiments in the directory subtree."""
if columns:
columns = columns.split(',')
columns = columns.split(",")
commands.list_experiments(project_path, sort, output, filter_op, columns)
python/ray/tune/suggest/hyperopt.py
+9 -9
View File
@@ -120,8 +120,8 @@ class HyperOptSearch(SuggestionAlgorithm):
if ho_trial is None:
return
now = hpo.utils.coarse_utcnow()
ho_trial['book_time'] = now
ho_trial['refresh_time'] = now
ho_trial["book_time"] = now
ho_trial["refresh_time"] = now
def on_trial_complete(self,
trial_id,
@@ -136,17 +136,17 @@ class HyperOptSearch(SuggestionAlgorithm):
ho_trial = self._get_hyperopt_trial(trial_id)
if ho_trial is None:
return
ho_trial['refresh_time'] = hpo.utils.coarse_utcnow()
ho_trial["refresh_time"] = hpo.utils.coarse_utcnow()
if error:
ho_trial['state'] = hpo.base.JOB_STATE_ERROR
ho_trial['misc']['error'] = (str(TuneError), "Tune Error")
ho_trial["state"] = hpo.base.JOB_STATE_ERROR
ho_trial["misc"]["error"] = (str(TuneError), "Tune Error")
elif early_terminated:
ho_trial['state'] = hpo.base.JOB_STATE_ERROR
ho_trial['misc']['error'] = (str(TuneError), "Tune Removed")
ho_trial["state"] = hpo.base.JOB_STATE_ERROR
ho_trial["misc"]["error"] = (str(TuneError), "Tune Removed")
else:
ho_trial['state'] = hpo.base.JOB_STATE_DONE
ho_trial["state"] = hpo.base.JOB_STATE_DONE
hp_result = self._to_hyperopt_result(result)
ho_trial['result'] = hp_result
ho_trial["result"] = hp_result
self._hpopt_trials.refresh()
del self._live_trial_mapping[trial_id]
python/ray/tune/suggest/sigopt.py
+1 -1
View File
@@ -67,7 +67,7 @@ class SigOptSearch(SuggestionAlgorithm):
self._live_trial_mapping = {}
# Create a connection with SigOpt API, requires API key
self.conn = sgo.Connection(client_token=os.environ['SIGOPT_KEY'])
self.conn = sgo.Connection(client_token=os.environ["SIGOPT_KEY"])
self.experiment = self.conn.experiments().create(
name=name,
python/ray/tune/tests/test_automl_searcher.py
+15 -15
View File
@@ -17,26 +17,26 @@ class AutoMLSearcherTest(unittest.TestCase):
register_trainable("f1", dummy_train)
def testExpandSearchSpace(self):
exp = {"test-exp": {"run": "f1", "config": {"a": {'d': 'dummy'}}}}
exp = {"test-exp": {"run": "f1", "config": {"a": {"d": "dummy"}}}}
space = SearchSpace([
DiscreteSpace('a.b.c', [1, 2]),
DiscreteSpace('a.d', ['a', 'b']),
DiscreteSpace("a.b.c", [1, 2]),
DiscreteSpace("a.d", ["a", "b"]),
])
searcher = GridSearch(space, 'reward')
searcher = GridSearch(space, "reward")
searcher.add_configurations(exp)
trials = searcher.next_trials()
self.assertEqual(len(trials), 4)
self.assertTrue(trials[0].config['a']['b']['c'] in [1, 2])
self.assertTrue(trials[1].config['a']['d'] in ['a', 'b'])
self.assertTrue(trials[0].config["a"]["b"]["c"] in [1, 2])
self.assertTrue(trials[1].config["a"]["d"] in ["a", "b"])
def testSearchRound(self):
exp = {"test-exp": {"run": "f1", "config": {"a": {'d': 'dummy'}}}}
exp = {"test-exp": {"run": "f1", "config": {"a": {"d": "dummy"}}}}
space = SearchSpace([
DiscreteSpace('a.b.c', [1, 2]),
DiscreteSpace('a.d', ['a', 'b']),
DiscreteSpace("a.b.c", [1, 2]),
DiscreteSpace("a.d", ["a", "b"]),
])
searcher = GridSearch(space, 'reward')
searcher = GridSearch(space, "reward")
searcher.add_configurations(exp)
trials = searcher.next_trials()
@@ -48,12 +48,12 @@ class AutoMLSearcherTest(unittest.TestCase):
self.assertTrue(searcher.is_finished())
def testBestTrial(self):
exp = {"test-exp": {"run": "f1", "config": {"a": {'d': 'dummy'}}}}
exp = {"test-exp": {"run": "f1", "config": {"a": {"d": "dummy"}}}}
space = SearchSpace([
DiscreteSpace('a.b.c', [1, 2]),
DiscreteSpace('a.d', ['a', 'b']),
DiscreteSpace("a.b.c", [1, 2]),
DiscreteSpace("a.d", ["a", "b"]),
])
searcher = GridSearch(space, 'reward')
searcher = GridSearch(space, "reward")
searcher.add_configurations(exp)
trials = searcher.next_trials()
@@ -66,4 +66,4 @@ class AutoMLSearcherTest(unittest.TestCase):
best_trial = searcher.get_best_trial()
self.assertEqual(best_trial, trials[-1])
self.assertEqual(best_trial.best_result['reward'], 3 + 10 - 1)
self.assertEqual(best_trial.best_result["reward"], 3 + 10 - 1)
python/ray/tune/tests/test_commands.py
+2 -2
View File
@@ -90,7 +90,7 @@ def test_ls(start_ray, tmpdir):
assert sum("TERMINATED" in line for line in lines) == num_samples
columns = ["status", "episode_reward_mean", "training_iteration"]
assert all(col in lines[1] for col in columns)
assert lines[1].count('|') == 4
assert lines[1].count("|") == 4
with Capturing() as output:
commands.list_trials(
@@ -123,7 +123,7 @@ def test_lsx(start_ray, tmpdir):
lines = output.captured
assert sum("1" in line for line in lines) >= num_experiments
assert "total_trials" in lines[1]
assert lines[1].count('|') == 2
assert lines[1].count("|") == 2
with Capturing() as output:
commands.list_experiments(
python/ray/tune/tests/test_dependency.py
+1 -1
View File
@@ -25,4 +25,4 @@ if __name__ == "__main__":
}
}
})
assert 'ray.rllib' not in sys.modules, "RLlib should not be imported"
assert "ray.rllib" not in sys.modules, "RLlib should not be imported"
python/ray/tune/tests/test_trial_runner.py
+15 -15
View File
@@ -331,21 +331,21 @@ class TrainableFunctionApiTest(unittest.TestCase):
self.assertFalse(trial.upload_dir)
def testLogdirStartingWithTilde(self):
local_dir = '~/ray_results/local_dir'
local_dir = "~/ray_results/local_dir"
def train(config, reporter):
cwd = os.getcwd()
assert cwd.startswith(os.path.expanduser(local_dir)), cwd
assert not cwd.startswith('~'), cwd
assert not cwd.startswith("~"), cwd
reporter(timesteps_total=1)
register_trainable('f1', train)
register_trainable("f1", train)
run_experiments({
'foo': {
'run': 'f1',
'local_dir': local_dir,
'config': {
'a': 'b'
"foo": {
"run": "f1",
"local_dir": local_dir,
"config": {
"a": "b"
},
}
})
@@ -501,7 +501,7 @@ class TrainableFunctionApiTest(unittest.TestCase):
def testReportInfinity(self):
def train(config, reporter):
for i in range(100):
reporter(mean_accuracy=float('inf'))
reporter(mean_accuracy=float("inf"))
register_trainable("f1", train)
[trial] = run_experiments({
@@ -510,7 +510,7 @@ class TrainableFunctionApiTest(unittest.TestCase):
}
})
self.assertEqual(trial.status, Trial.TERMINATED)
self.assertEqual(trial.last_result['mean_accuracy'], float('inf'))
self.assertEqual(trial.last_result["mean_accuracy"], float("inf"))
def testReportTimeStep(self):
# Test that no timestep count are logged if never the Trainable never
@@ -1532,7 +1532,7 @@ class TrialRunnerTest(unittest.TestCase):
runner.add_trial(Trial("__fake", **kwargs))
trials = runner.get_trials()
with patch('ray.global_state.cluster_resources') as resource_mock:
with patch("ray.global_state.cluster_resources") as resource_mock:
resource_mock.return_value = {"CPU": 1, "GPU": 1}
runner.step()
self.assertEqual(trials[0].status, Trial.RUNNING)
@@ -1717,12 +1717,12 @@ class TrialRunnerTest(unittest.TestCase):
def on_step_begin(self):
self._update_avail_resources()
cnt = self.pre_step if hasattr(self, 'pre_step') else 0
setattr(self, 'pre_step', cnt + 1)
cnt = self.pre_step if hasattr(self, "pre_step") else 0
setattr(self, "pre_step", cnt + 1)
def on_step_end(self):
cnt = self.pre_step if hasattr(self, 'post_step') else 0
setattr(self, 'post_step', 1 + cnt)
cnt = self.pre_step if hasattr(self, "post_step") else 0
setattr(self, "post_step", 1 + cnt)
import types
runner.trial_executor.on_step_begin = types.MethodType(
python/ray/tune/tests/test_trial_scheduler.py
+5 -5
View File
@@ -135,8 +135,8 @@ class EarlyStoppingSuite(unittest.TestCase):
rule = MedianStoppingRule(
grace_period=0,
min_samples_required=1,
time_attr='training_iteration',
reward_attr='neg_mean_loss')
time_attr="training_iteration",
reward_attr="neg_mean_loss")
t1 = Trial("PPO") # mean is 450, max 900, t_max=10
t2 = Trial("PPO") # mean is 450, max 450, t_max=5
for i in range(10):
@@ -495,7 +495,7 @@ class HyperbandSuite(unittest.TestCase):
return dict(time_total_s=t, neg_mean_loss=rew)
sched = HyperBandScheduler(
time_attr='time_total_s', reward_attr='neg_mean_loss')
time_attr="time_total_s", reward_attr="neg_mean_loss")
stats = self.default_statistics()
for i in range(stats["max_trials"]):
@@ -987,8 +987,8 @@ class AsyncHyperBandSuite(unittest.TestCase):
scheduler = AsyncHyperBandScheduler(
grace_period=1,
time_attr='training_iteration',
reward_attr='neg_mean_loss',
time_attr="training_iteration",
reward_attr="neg_mean_loss",
brackets=1)
t1 = Trial("PPO") # mean is 450, max 900, t_max=10
t2 = Trial("PPO") # mean is 450, max 450, t_max=5
python/ray/tune/tests/test_tune_server.py
+4 -4
View File
@@ -69,8 +69,8 @@ class TuneServerSuite(unittest.TestCase):
"training_iteration": 3
},
"resources_per_trial": {
'cpu': 1,
'gpu': 1
"cpu": 1,
"gpu": 1
},
}
client.add_trial("test", spec)
@@ -134,8 +134,8 @@ class TuneServerSuite(unittest.TestCase):
for i in range(2):
runner.step()
stdout = subprocess.check_output(
'curl "http://{}:{}/trials"'.format(client.server_address,
client.server_port),
"curl \"http://{}:{}/trials\"".format(client.server_address,
client.server_port),
shell=True)
self.assertNotEqual(stdout, None)
curl_trials = json.loads(stdout.decode())["trials"]
python/ray/tune/trial.py
+14 -14
View File
@@ -437,39 +437,39 @@ class Trial(object):
def location_string(hostname, pid):
if hostname == os.uname()[1]:
return 'pid={}'.format(pid)
return "pid={}".format(pid)
else:
return '{} pid={}'.format(hostname, pid)
return "{} pid={}".format(hostname, pid)
pieces = [
'{}'.format(self._status_string()), '[{}]'.format(
self.resources.summary_string()), '[{}]'.format(
"{}".format(self._status_string()), "[{}]".format(
self.resources.summary_string()), "[{}]".format(
location_string(
self.last_result.get(HOSTNAME),
self.last_result.get(PID))), '{} s'.format(
self.last_result.get(PID))), "{} s".format(
int(self.last_result.get(TIME_TOTAL_S)))
]
if self.last_result.get(TRAINING_ITERATION) is not None:
pieces.append('{} iter'.format(
pieces.append("{} iter".format(
self.last_result[TRAINING_ITERATION]))
if self.last_result.get(TIMESTEPS_TOTAL) is not None:
pieces.append('{} ts'.format(self.last_result[TIMESTEPS_TOTAL]))
pieces.append("{} ts".format(self.last_result[TIMESTEPS_TOTAL]))
if self.last_result.get(EPISODE_REWARD_MEAN) is not None:
pieces.append('{} rew'.format(
format(self.last_result[EPISODE_REWARD_MEAN], '.3g')))
pieces.append("{} rew".format(
format(self.last_result[EPISODE_REWARD_MEAN], ".3g")))
if self.last_result.get(MEAN_LOSS) is not None:
pieces.append('{} loss'.format(
format(self.last_result[MEAN_LOSS], '.3g')))
pieces.append("{} loss".format(
format(self.last_result[MEAN_LOSS], ".3g")))
if self.last_result.get(MEAN_ACCURACY) is not None:
pieces.append('{} acc'.format(
format(self.last_result[MEAN_ACCURACY], '.3g')))
pieces.append("{} acc".format(
format(self.last_result[MEAN_ACCURACY], ".3g")))
return ', '.join(pieces)
return ", ".join(pieces)
def _status_string(self):
return "{}{}".format(
python/ray/tune/trial_runner.py
+1 -1
View File
@@ -127,7 +127,7 @@ class TrialRunner(object):
# For debugging, it may be useful to halt trials after some time has
# elapsed. TODO(ekl) consider exposing this in the API.
self._global_time_limit = float(
os.environ.get("TRIALRUNNER_WALLTIME_LIMIT", float('inf')))
os.environ.get("TRIALRUNNER_WALLTIME_LIMIT", float("inf")))
self._total_time = 0
self._iteration = 0
self._verbose = verbose
python/ray/tune/web_server.py
+5 -5
View File
@@ -110,7 +110,7 @@ def RunnerHandler(runner):
headers (list[tuples]): Standard HTTP response headers
"""
if headers is None:
headers = [('Content-type', 'application/json')]
headers = [("Content-type", "application/json")]
self.send_response(response_code)
for key, value in headers:
@@ -170,14 +170,14 @@ def RunnerHandler(runner):
"""HTTP POST handler method."""
response_code = 201
content_len = int(self.headers.get('Content-Length'), 0)
content_len = int(self.headers.get("Content-Length"), 0)
raw_body = self.rfile.read(content_len)
parsed_input = json.loads(raw_body.decode())
resource = self._add_trials(parsed_input["name"],
parsed_input["spec"])
headers = [('Content-type', 'application/json'), ('Location',
'/trials/')]
headers = [("Content-type", "application/json"), ("Location",
"/trials/")]
self._do_header(response_code=response_code, headers=headers)
self.wfile.write(json.dumps(resource).encode())
@@ -237,7 +237,7 @@ class TuneServer(threading.Thread):
"""Initialize HTTPServer and serve forever by invoking self.run()"""
threading.Thread.__init__(self)
self._port = port if port else self.DEFAULT_PORT
address = ('localhost', self._port)
address = ("localhost", self._port)
logger.info("Starting Tune Server...")
self._server = HTTPServer(address, RunnerHandler(runner))
self.daemon = True