|
|
|
|
|
import subprocess |
|
|
|
from ultralytics.cfg import TASK2DATA, TASK2METRIC, get_save_dir |
|
from ultralytics.utils import DEFAULT_CFG, DEFAULT_CFG_DICT, LOGGER, NUM_THREADS, checks |
|
|
|
|
|
def run_ray_tune( |
|
model, space: dict = None, grace_period: int = 10, gpu_per_trial: int = None, max_samples: int = 10, **train_args |
|
): |
|
""" |
|
Runs hyperparameter tuning using Ray Tune. |
|
|
|
Args: |
|
model (YOLO): Model to run the tuner on. |
|
space (dict, optional): The hyperparameter search space. Defaults to None. |
|
grace_period (int, optional): The grace period in epochs of the ASHA scheduler. Defaults to 10. |
|
gpu_per_trial (int, optional): The number of GPUs to allocate per trial. Defaults to None. |
|
max_samples (int, optional): The maximum number of trials to run. Defaults to 10. |
|
train_args (dict, optional): Additional arguments to pass to the `train()` method. Defaults to {}. |
|
|
|
Returns: |
|
(dict): A dictionary containing the results of the hyperparameter search. |
|
|
|
Example: |
|
```python |
|
from ultralytics import YOLO |
|
|
|
# Load a YOLOv8n model |
|
model = YOLO('yolov8n.pt') |
|
|
|
# Start tuning hyperparameters for YOLOv8n training on the COCO8 dataset |
|
result_grid = model.tune(data='coco8.yaml', use_ray=True) |
|
``` |
|
""" |
|
|
|
LOGGER.info("π‘ Learn about RayTune at https://docs.ultralytics.com/integrations/ray-tune") |
|
if train_args is None: |
|
train_args = {} |
|
|
|
try: |
|
subprocess.run("pip install ray[tune]<=2.9.3".split(), check=True) |
|
|
|
import ray |
|
from ray import tune |
|
from ray.air import RunConfig |
|
from ray.air.integrations.wandb import WandbLoggerCallback |
|
from ray.tune.schedulers import ASHAScheduler |
|
except ImportError: |
|
raise ModuleNotFoundError('Ray Tune required but not found. To install run: pip install "ray[tune]<=2.9.3"') |
|
|
|
try: |
|
import wandb |
|
|
|
assert hasattr(wandb, "__version__") |
|
except (ImportError, AssertionError): |
|
wandb = False |
|
|
|
checks.check_version(ray.__version__, "<=2.9.3", "ray") |
|
default_space = { |
|
|
|
"lr0": tune.uniform(1e-5, 1e-1), |
|
"lrf": tune.uniform(0.01, 1.0), |
|
"momentum": tune.uniform(0.6, 0.98), |
|
"weight_decay": tune.uniform(0.0, 0.001), |
|
"warmup_epochs": tune.uniform(0.0, 5.0), |
|
"warmup_momentum": tune.uniform(0.0, 0.95), |
|
"box": tune.uniform(0.02, 0.2), |
|
"cls": tune.uniform(0.2, 4.0), |
|
"hsv_h": tune.uniform(0.0, 0.1), |
|
"hsv_s": tune.uniform(0.0, 0.9), |
|
"hsv_v": tune.uniform(0.0, 0.9), |
|
"degrees": tune.uniform(0.0, 45.0), |
|
"translate": tune.uniform(0.0, 0.9), |
|
"scale": tune.uniform(0.0, 0.9), |
|
"shear": tune.uniform(0.0, 10.0), |
|
"perspective": tune.uniform(0.0, 0.001), |
|
"flipud": tune.uniform(0.0, 1.0), |
|
"fliplr": tune.uniform(0.0, 1.0), |
|
"bgr": tune.uniform(0.0, 1.0), |
|
"mosaic": tune.uniform(0.0, 1.0), |
|
"mixup": tune.uniform(0.0, 1.0), |
|
"copy_paste": tune.uniform(0.0, 1.0), |
|
} |
|
|
|
|
|
task = model.task |
|
model_in_store = ray.put(model) |
|
|
|
def _tune(config): |
|
""" |
|
Trains the YOLO model with the specified hyperparameters and additional arguments. |
|
|
|
Args: |
|
config (dict): A dictionary of hyperparameters to use for training. |
|
|
|
Returns: |
|
None |
|
""" |
|
model_to_train = ray.get(model_in_store) |
|
model_to_train.reset_callbacks() |
|
config.update(train_args) |
|
results = model_to_train.train(**config) |
|
return results.results_dict |
|
|
|
|
|
if not space: |
|
space = default_space |
|
LOGGER.warning("WARNING β οΈ search space not provided, using default search space.") |
|
|
|
|
|
data = train_args.get("data", TASK2DATA[task]) |
|
space["data"] = data |
|
if "data" not in train_args: |
|
LOGGER.warning(f'WARNING β οΈ data not provided, using default "data={data}".') |
|
|
|
|
|
trainable_with_resources = tune.with_resources(_tune, {"cpu": NUM_THREADS, "gpu": gpu_per_trial or 0}) |
|
|
|
|
|
asha_scheduler = ASHAScheduler( |
|
time_attr="epoch", |
|
metric=TASK2METRIC[task], |
|
mode="max", |
|
max_t=train_args.get("epochs") or DEFAULT_CFG_DICT["epochs"] or 100, |
|
grace_period=grace_period, |
|
reduction_factor=3, |
|
) |
|
|
|
|
|
tuner_callbacks = [WandbLoggerCallback(project="YOLOv8-tune")] if wandb else [] |
|
|
|
|
|
tune_dir = get_save_dir(DEFAULT_CFG, name="tune").resolve() |
|
tune_dir.mkdir(parents=True, exist_ok=True) |
|
tuner = tune.Tuner( |
|
trainable_with_resources, |
|
param_space=space, |
|
tune_config=tune.TuneConfig(scheduler=asha_scheduler, num_samples=max_samples), |
|
run_config=RunConfig(callbacks=tuner_callbacks, storage_path=tune_dir), |
|
) |
|
|
|
|
|
tuner.fit() |
|
|
|
|
|
return tuner.get_results() |
|
|