Partially Use f string (#10218)

* flynt. trial 1.

* Trial 1.

* Addressed code review.

python/ray/util/sgd/tests/test_torch.py

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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

@@ -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