wassname/ray
1
0
Fork 0
mirror of https://github.com/wassname/ray.git synced 2026-07-13 17:45:08 +08:00
Code Issues Packages Projects Releases Wiki Activity

Partially Use f string (#10218)

Browse Source 
* flynt. trial 1.

* Trial 1.

* Addressed code review.
This commit is contained in:
SangBin Cho
2020-08-20 18:21:16 -07:00
committed by GitHub
parent 07cd815e5a
commit 92664249e8
41 changed files with 195 additions and 238 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/ray/util/iter.py
+18 -22
View File
@@ -53,8 +53,8 @@ def from_range(n: int, num_shards: int = 2,
else:
end = (i + 1) * shard_size
generators.append(range(start, end))
name = "from_range[{}, shards={}{}]".format(
n, num_shards, ", repeat=True" if repeat else "")
name = (f"from_range[{n}, shards={num_shards}"
f"{', repeat=True' if repeat else ''}]")
return from_iterators(
generators,
repeat=repeat,
@@ -111,7 +111,7 @@ def from_actors(actors: List["ray.actor.ActorHandle"],
name (str): Optional name to give the iterator.
"""
if not name:
name = "from_actors[shards={}]".format(len(actors))
name = f"from_actors[shards={len(actors)}]"
return ParallelIterator([_ActorSet(actors, [])], name, parent_iterators=[])
@@ -184,7 +184,7 @@ class ParallelIterator(Generic[T]):
return repr(self)
def __repr__(self):
return "ParallelIterator[{}]".format(self.name)
return f"ParallelIterator[{self.name}]"
def _with_transform(self, local_it_fn, name):
"""Helper function to create new Parallel Iterator"""
@@ -290,7 +290,7 @@ class ParallelIterator(Generic[T]):
... [0, 1, 2, 3]
"""
return self._with_transform(lambda local_it: local_it.batch(n),
".batch({})".format(n))
f".batch({n})")
def flatten(self) -> "ParallelIterator[T[0]]":
"""Flatten batches of items into individual items.
@@ -421,8 +421,7 @@ class ParallelIterator(Generic[T]):
def make_gen_i(i):
return lambda: base_iterator(num_partitions, i)
name = self.name + ".repartition[num_partitions={}]".format(
num_partitions)
name = self.name + f".repartition[num_partitions={num_partitions}]"
generators = [make_gen_i(s) for s in range(num_partitions)]
worker_cls = ray.remote(ParallelIteratorWorker)
@@ -449,7 +448,7 @@ class ParallelIterator(Generic[T]):
... 2
"""
it = self.batch_across_shards().flatten()
it.name = "{}.gather_sync()".format(self)
it.name = f"{self}.gather_sync()"
return it
def batch_across_shards(self) -> "LocalIterator[List[T]]":
@@ -488,7 +487,7 @@ class ParallelIterator(Generic[T]):
yield results
futures = [a.par_iter_next.remote() for a in active]
name = "{}.batch_across_shards()".format(self)
name = f"{self}.batch_across_shards()"
return LocalIterator(base_iterator, SharedMetrics(), name=name)
def gather_async(self, batch_ms=0, num_async=1) -> "LocalIterator[T]":
@@ -560,7 +559,7 @@ class ParallelIterator(Generic[T]):
if timeout is not None:
yield _NextValueNotReady()
name = "{}.gather_async()".format(self)
name = f"{self}.gather_async()"
local_iter = LocalIterator(base_iterator, SharedMetrics(), name=name)
return local_iter
@@ -576,8 +575,7 @@ class ParallelIterator(Generic[T]):
"""Return an iterator that is the union of this and the other."""
if not isinstance(other, ParallelIterator):
raise TypeError(
"other must be of type ParallelIterator, got {}".format(
type(other)))
f"other must be of type ParallelIterator, got {type(other)}")
actor_sets = []
actor_sets.extend(self.actor_sets)
actor_sets.extend(other.actor_sets)
@@ -585,7 +583,7 @@ class ParallelIterator(Generic[T]):
# keep an explicit reference to its parent iterator
return ParallelIterator(
actor_sets,
"ParallelUnion[{}, {}]".format(self, other),
f"ParallelUnion[{self}, {other}]",
parent_iterators=self.parent_iterators + other.parent_iterators)
def select_shards(self,
@@ -608,7 +606,7 @@ class ParallelIterator(Generic[T]):
new_actor_set = _ActorSet(new_actors, old_actor_set.transforms)
return ParallelIterator(
[new_actor_set],
"{}.select_shards({} total)".format(self, len(shards_to_keep)),
f"{self}.select_shards({len(shards_to_keep)} total)",
parent_iterators=self.parent_iterators)
def num_shards(self) -> int:
@@ -674,7 +672,7 @@ class ParallelIterator(Generic[T]):
except StopIteration:
break
name = self.name + ".shard[{}]".format(shard_index)
name = self.name + f".shard[{shard_index}]"
return LocalIterator(base_iterator, SharedMetrics(), name=name)
@@ -761,7 +759,7 @@ class LocalIterator(Generic[T]):
return repr(self)
def __repr__(self):
return "LocalIterator[{}]".format(self.name)
return f"LocalIterator[{self.name}]"
def transform(self, fn: Callable[[Iterable[T]], Iterable[U]]
) -> "LocalIterator[U]":
@@ -871,7 +869,7 @@ class LocalIterator(Generic[T]):
self.base_iterator,
self.shared_metrics,
self.local_transforms + [apply_batch],
name=self.name + ".batch({})".format(n))
name=self.name + f".batch({n})")
def flatten(self) -> "LocalIterator[T[0]]":
def apply_flatten(it):
@@ -1013,7 +1011,7 @@ class LocalIterator(Generic[T]):
LocalIterator(
make_next(i),
self.shared_metrics, [],
name=self.name + ".duplicate[{}]".format(i)))
name=self.name + f".duplicate[{i}]"))
return iterators
@@ -1038,8 +1036,7 @@ class LocalIterator(Generic[T]):
for it in others:
if not isinstance(it, LocalIterator):
raise ValueError(
"other must be of type LocalIterator, got {}".format(
type(it)))
f"other must be of type LocalIterator, got {type(it)}")
active = []
parent_iters = [self] + list(others)
@@ -1090,8 +1087,7 @@ class LocalIterator(Generic[T]):
return LocalIterator(
build_union,
shared_metrics, [],
name="LocalUnion[{}, {}]".format(self, ", ".join(map(str,
others))))
name=f"LocalUnion[{self}, {', '.join(map(str, others))}]")
class ParallelIteratorWorker(object):
python/ray/util/multiprocessing/pool.py
+3 -3
View File
@@ -168,7 +168,7 @@ class AsyncResult:
"""
if not self.ready():
raise ValueError("{0!r} not ready".format(self))
raise ValueError(f"{self!r} not ready")
return not self._result_thread.got_error()
@@ -355,8 +355,8 @@ class Pool:
os.environ[RAY_ADDRESS_ENV]))
ray.init()
elif ray_address is not None:
logger.info("Connecting to ray cluster at address='{}'".format(
ray_address))
logger.info(
f"Connecting to ray cluster at address='{ray_address}'")
ray.init(address=ray_address)
# Local mode.
else:
python/ray/util/sgd/tests/test_torch.py
+1 -1
View File
@@ -158,7 +158,7 @@ def test_multi_model(ray_start_2_cpus, num_workers):
data = list(iterator)
for i, (model, optimizer) in enumerate(
zip(self.models, self.optimizers)):
result["model_{}".format(i)] = train(
result[f"model_{i}"] = train(
model=model,
criterion=self.criterion,
optimizer=optimizer,
python/ray/util/sgd/tf/examples/cifar_tf_example.py
+1 -1
View File
@@ -208,7 +208,7 @@ if __name__ == "__main__":
# Trains num epochs
train_stats1 = trainer.train()
train_stats1.update(trainer.validate())
print("iter {}:".format(i), train_stats1)
print(f"iter {i}:", train_stats1)
dt = (time.time() - training_start) / 3
print(f"Training on workers takes: {dt:.3f} seconds/epoch")
python/ray/util/sgd/tf/tf_trainer.py
+1 -4
View File
@@ -85,10 +85,7 @@ class TFTrainer:
ports = ray.get(
[worker.find_free_port.remote() for worker in self.workers])
urls = [
"{ip}:{port}".format(ip=ips[i], port=ports[i])
for i in range(len(self.workers))
]
urls = [f"{ips[i]}:{ports[i]}" for i in range(len(self.workers))]
# Get setup tasks in order to throw errors on failure
ray.get([
python/ray/util/sgd/torch/distributed_torch_runner.py
+1 -1
View File
@@ -230,7 +230,7 @@ def reserve_resources(num_cpus, num_gpus, retries=20):
cuda_device_set = {}
match_devices = bool(cuda_devices)
if match_devices:
logger.debug("Found set devices: {}".format(cuda_devices))
logger.debug(f"Found set devices: {cuda_devices}")
assert isinstance(cuda_devices, str)
cuda_device_set = set(cuda_devices.split(","))
python/ray/util/sgd/torch/examples/benchmarks/benchmark.py
+2 -2
View File
@@ -102,7 +102,7 @@ if __name__ == "__main__":
num_workers = 2 if args.local else int(ray.cluster_resources().get(device))
from ray.util.sgd.torch.examples.train_example import LinearDataset
print("Model: %s" % args.model)
print(f"Model: {args.model}")
print("Batch size: %d" % args.batch_size)
print("Number of %ss: %d" % (device, num_workers))
@@ -131,7 +131,7 @@ if __name__ == "__main__":
# Results
img_sec_mean = np.mean(img_secs)
img_sec_conf = 1.96 * np.std(img_secs)
print("Img/sec per %s: %.1f +-%.1f" % (device, img_sec_mean, img_sec_conf))
print(f"Img/sec per {device}: {img_sec_mean:.1f} +-{img_sec_conf:.1f}")
print("Total img/sec on %d %s(s): %.1f +-%.1f" %
(num_workers, device, num_workers * img_sec_mean,
num_workers * img_sec_conf))
python/ray/util/sgd/torch/examples/benchmarks/dp_benchmark.py
+2 -2
View File
@@ -77,7 +77,7 @@ def benchmark_step():
optimizer.step()
print("Model: %s" % args.model)
print(f"Model: {args.model}")
print("Batch size: %d" % args.batch_size)
device = "GPU"
print("Number of %ss: %d" % (device, args.num_gpus))
@@ -98,7 +98,7 @@ for x in range(args.num_iters):
# Results
img_sec_mean = np.mean(img_secs)
img_sec_conf = 1.96 * np.std(img_secs)
print("Img/sec per %s: %.1f +-%.1f" % (device, img_sec_mean, img_sec_conf))
print(f"Img/sec per {device}: {img_sec_mean:.1f} +-{img_sec_conf:.1f}")
print("Total img/sec on %d %s(s): %.1f +-%.1f" % (
args.num_gpus,
device,
python/ray/util/sgd/torch/examples/benchmarks/horovod_benchmark_apex.py
+2 -2
View File
@@ -118,7 +118,7 @@ def log(s, nl=True):
print(s, end="\n" if nl else "")
log("Model: %s" % args.model)
log(f"Model: {args.model}")
log("Batch size: %d" % args.batch_size)
device = "GPU" if args.cuda else "CPU"
log("Number of %ss: %d" % (device, hvd.size()))
@@ -139,6 +139,6 @@ for x in range(args.num_iters):
# Results
img_sec_mean = np.mean(img_secs)
img_sec_conf = 1.96 * np.std(img_secs)
log("Img/sec per %s: %.1f +-%.1f" % (device, img_sec_mean, img_sec_conf))
log(f"Img/sec per {device}: {img_sec_mean:.1f} +-{img_sec_conf:.1f}")
log("Total img/sec on %d %s(s): %.1f +-%.1f" %
(hvd.size(), device, hvd.size() * img_sec_mean, hvd.size() * img_sec_conf))
python/ray/util/sgd/torch/examples/image_models/util.py
+2 -3
View File
@@ -20,13 +20,12 @@ def mock_data(train_dir, val_dir):
def mock_class(base, n):
random_cls = random.randint(per_cls * n, per_cls * n + per_cls)
sub_dir = join(base, "n{:08d}".format(random_cls))
sub_dir = join(base, f"n{random_cls:08d}")
os.makedirs(sub_dir, exist_ok=True)
for i in range(total_imgs):
random_img = random.randint(per_img * i, per_img * i + per_img)
file = join(sub_dir,
"ILSVRC2012_val_{:08d}.JPEG".format(random_img))
file = join(sub_dir, f"ILSVRC2012_val_{random_img:08d}.JPEG")
PIL.Image.fromarray(np.zeros((375, 500, 3),
dtype=np.uint8)).save(file)
python/ray/util/sgd/torch/examples/segmentation/train_segmentation.py
+2 -2
View File
@@ -193,11 +193,11 @@ def main(args):
state_dict = trainer.state_dict()
state_dict.update(epoch=epoch, args=args)
torch.save(state_dict,
os.path.join(args.output_dir, "model_{}.pth".format(epoch)))
os.path.join(args.output_dir, f"model_{epoch}.pth"))
total_time = time.time() - start_time
total_time_str = str(datetime.timedelta(seconds=int(total_time)))
print("Training time {}".format(total_time_str))
print(f"Training time {total_time_str}")
def parse_args():
python/ray/util/sgd/torch/examples/segmentation/utils.py
+2 -2
View File
@@ -49,8 +49,8 @@ class ConfusionMatrix(object):
'IoU: {}\n'
'mean IoU: {:.1f}').format(
acc_global.item() * 100,
['{:.1f}'.format(i) for i in (acc * 100).tolist()],
['{:.1f}'.format(i) for i in (iu * 100).tolist()],
[f'{i:.1f}' for i in (acc * 100).tolist()],
[f'{i:.1f}' for i in (iu * 100).tolist()],
iu.mean().item() * 100)
python/ray/util/sgd/torch/examples/transformers/transformers_example.py
+2 -2
View File
@@ -147,7 +147,7 @@ def data_creator(config):
if args.tokenizer_name else args.model_name_or_path,
cache_dir=args.cache_dir if args.cache_dir else None,
)
logger.info("tokenizer instantiation time: {}".format(time.time() - start))
logger.info(f"tokenizer instantiation time: {time.time() - start}")
train_dataset = load_and_cache_examples(
args, args.task_name, tokenizer, evaluate=False)
@@ -322,7 +322,7 @@ def main():
# Prepare GLUE task
args.task_name = args.task_name.lower()
if args.task_name not in processors:
raise ValueError("Task not found: %s" % (args.task_name))
raise ValueError(f"Task not found: {args.task_name}")
args.output_mode = output_modes[args.task_name]
logging.basicConfig(
python/ray/util/sgd/torch/examples/transformers/utils.py
+2 -2
View File
@@ -133,7 +133,7 @@ def save_and_evaluate_checkpoints(args, model, tokenizer):
model.to(args.device)
result = evaluate(args, model, tokenizer, prefix=prefix)
result = dict(
(k + "_{}".format(global_step), v) for k, v in result.items())
(k + f"_{global_step}", v) for k, v in result.items())
results.update(result)
return results
@@ -163,7 +163,7 @@ def evaluate(args, model, tokenizer, prefix=""):
batch_size=args.eval_batch_size)
# Eval!
logger.info("***** Running evaluation {} *****".format(prefix))
logger.info(f"***** Running evaluation {prefix} *****")
logger.info(" Num examples = %d", len(eval_dataset))
logger.info(" Batch size = %d", args.eval_batch_size)
eval_loss = 0.0
python/ray/util/sgd/torch/torch_runner.py
+3 -3
View File
@@ -82,7 +82,7 @@ class TorchRunner:
return loaders
else:
raise ValueError(
"Number of loaders must be <= 2. Got {}".format(loaders))
f"Number of loaders must be <= 2. Got {loaders}")
# No great way of checking type otherwise
return loaders, None
@@ -146,7 +146,7 @@ class TorchRunner:
if not isinstance(self.models, Iterable):
self.models = [self.models]
assert all(isinstance(model, nn.Module) for model in self.models), (
"All models must be PyTorch models: {}.".format(self.models))
f"All models must be PyTorch models: {self.models}.")
if self.use_gpu and torch.cuda.is_available():
self.models = [model.cuda() for model in self.models]
@@ -189,7 +189,7 @@ class TorchRunner:
info=None,
iterator=None):
"""Runs a training epoch and updates the model parameters."""
logger.debug("Begin Training Step {}".format(self.epochs + 1))
logger.debug(f"Begin Training Step {self.epochs + 1}")
info = info or {}
self._toggle_profiling(profile=profile)
python/ray/util/sgd/torch/torch_trainer.py
+4 -4
View File
@@ -245,7 +245,7 @@ class TorchTrainer:
if backend == "auto":
backend = "nccl" if use_gpu else "gloo"
logger.debug("Using {} as backend.".format(backend))
logger.debug(f"Using {backend} as backend.")
self.backend = backend
self.num_cpus_per_worker = num_cpus_per_worker
self.use_gpu = use_gpu
@@ -659,12 +659,12 @@ class TorchTrainer:
"Failed to shutdown gracefully, forcing a shutdown.")
for worker in self.remote_workers:
logger.warning("Killing worker {}.".format(worker))
logger.warning(f"Killing worker {worker}.")
ray.kill(worker)
else:
self.local_worker.shutdown()
for worker in self.remote_workers:
logger.debug("Killing worker {}.".format(worker))
logger.debug(f"Killing worker {worker}.")
ray.kill(worker)
self.local_worker = DeactivatedRunner()
@@ -674,7 +674,7 @@ class TorchTrainer:
"""Terminates models without giving up local resource reservation."""
self.local_worker.shutdown(cleanup=False)
for worker in self.remote_workers:
logger.debug("Killing worker {}.".format(worker))
logger.debug(f"Killing worker {worker}.")
ray.kill(worker)
self.local_worker = DeactivatedRunner()
self.remote_workers = []
python/ray/util/sgd/torch/training_operator.py
+7 -13
View File
@@ -72,23 +72,18 @@ class TrainingOperator:
self._models = models # List of models
assert isinstance(
models,
Iterable), ("Components need to be iterable. Got: {}".format(
type(models)))
Iterable), (f"Components need to be iterable. Got: {type(models)}")
self._optimizers = optimizers # List of optimizers
assert isinstance(
optimizers,
Iterable), ("Components need to be iterable. Got: {}".format(
type(optimizers)))
assert isinstance(optimizers, Iterable), (
f"Components need to be iterable. Got: {type(optimizers)}")
self._train_loader = train_loader
self._validation_loader = validation_loader
self._world_rank = world_rank
self._criterion = criterion
self._schedulers = schedulers
if schedulers:
assert isinstance(
schedulers,
Iterable), ("Components need to be iterable. Got: {}".format(
type(schedulers)))
assert isinstance(schedulers, Iterable), (
f"Components need to be iterable. Got: {type(schedulers)}")
self._config = config
self._use_fp16 = use_fp16
self._device_ids = device_ids
@@ -165,10 +160,9 @@ class TrainingOperator:
desc = ""
if info is not None and "epoch_idx" in info:
if "num_epochs" in info:
desc = "{}/{}e".format(info["epoch_idx"] + 1,
info["num_epochs"])
desc = f"{info['epoch_idx'] + 1}/{info['num_epochs']}e"
else:
desc = "{}e".format(info["epoch_idx"] + 1)
desc = f"{info['epoch_idx'] + 1}e"
_progress_bar = tqdm(
total=info[NUM_STEPS] or len(self.train_loader),
desc=desc,
python/ray/util/sgd/torch/utils.py
+2 -2
View File
@@ -11,7 +11,7 @@ logger = logging.getLogger(__name__)
def setup_address():
ip = ray.services.get_node_ip_address()
port = find_free_port()
return "tcp://{ip}:{port}".format(ip=ip, port=port)
return f"tcp://{ip}:{port}"
def setup_process_group(url, world_rank, world_size, timeout, backend="gloo"):
@@ -28,7 +28,7 @@ def setup_process_group(url, world_rank, world_size, timeout, backend="gloo"):
"""
logger.debug("Connecting to {} world_rank: {} world_size: {}".format(
url, world_rank, world_size))
logger.debug("using {}".format(backend))
logger.debug(f"using {backend}")
if backend == "nccl" and "NCCL_BLOCKING_WAIT" not in os.environ:
logger.debug(
"Setting NCCL_BLOCKING_WAIT for detecting node failure. "
python/ray/util/sgd/utils.py
+2 -2
View File
@@ -142,9 +142,9 @@ class TimerCollection:
for k, t in self._timers.items():
if t.count > 0:
if mean:
aggregates["mean_%s_s" % k] = t.mean
aggregates[f"mean_{k}_s"] = t.mean
if last:
aggregates["last_%s_s" % k] = t.last
aggregates[f"last_{k}_s"] = t.last
return aggregates