ultralytics/docs/reference/yolo/utils/torch_utils.md

description
Optimize your PyTorch models with Ultralytics YOLO's torch_utils functions such as ModelEMA, select_device, and is_parallel.

ModelEMA

---

:::ultralytics.yolo.utils.torch_utils.ModelEMA

EarlyStopping

---

:::ultralytics.yolo.utils.torch_utils.EarlyStopping

torch_distributed_zero_first

---

:::ultralytics.yolo.utils.torch_utils.torch_distributed_zero_first

smart_inference_mode

---

:::ultralytics.yolo.utils.torch_utils.smart_inference_mode

select_device

---

:::ultralytics.yolo.utils.torch_utils.select_device

time_sync

---

:::ultralytics.yolo.utils.torch_utils.time_sync

fuse_conv_and_bn

---

:::ultralytics.yolo.utils.torch_utils.fuse_conv_and_bn

fuse_deconv_and_bn

---

:::ultralytics.yolo.utils.torch_utils.fuse_deconv_and_bn

model_info

---

:::ultralytics.yolo.utils.torch_utils.model_info

get_num_params

---

:::ultralytics.yolo.utils.torch_utils.get_num_params

get_num_gradients

---

:::ultralytics.yolo.utils.torch_utils.get_num_gradients

get_flops

---

:::ultralytics.yolo.utils.torch_utils.get_flops

initialize_weights

---

:::ultralytics.yolo.utils.torch_utils.initialize_weights

scale_img

---

:::ultralytics.yolo.utils.torch_utils.scale_img

make_divisible

---

:::ultralytics.yolo.utils.torch_utils.make_divisible

copy_attr

---

:::ultralytics.yolo.utils.torch_utils.copy_attr

get_latest_opset

---

:::ultralytics.yolo.utils.torch_utils.get_latest_opset

intersect_dicts

---

:::ultralytics.yolo.utils.torch_utils.intersect_dicts

is_parallel

---

:::ultralytics.yolo.utils.torch_utils.is_parallel

de_parallel

---

:::ultralytics.yolo.utils.torch_utils.de_parallel

one_cycle

---

:::ultralytics.yolo.utils.torch_utils.one_cycle

init_seeds

---

:::ultralytics.yolo.utils.torch_utils.init_seeds

strip_optimizer

---

:::ultralytics.yolo.utils.torch_utils.strip_optimizer

profile

---

:::ultralytics.yolo.utils.torch_utils.profile