Inheritance diagram for fastreid.engine.hooks.LRScheduler:
Public Member Functions | |
| __init__ (self, optimizer, scheduler) | |
| after_step (self) | |
 Public Member Functions inherited from fastreid.engine.train_loop.HookBase | |
| before_train (self) | |
| after_train (self) | |
| before_step (self) | |
Protected Attributes | |
| _optimizer | |
| _scheduler | |
| _best_param_group_id | |
Detailed Description
A hook which executes a torch builtin LR scheduler and summarizes the LR. It is executed after every iteration.
Constructor & Destructor Documentation
◆ __init__()
| fastreid.engine.hooks.LRScheduler.__init__ | ( | self, | |
| optimizer, | |||
| scheduler ) |
Args:
optimizer (torch.optim.Optimizer):
scheduler (torch.optim._LRScheduler)
Definition at line 198 of file hooks.py.
198 def __init__(self, optimizer, scheduler):
199 """
200 Args:
201 optimizer (torch.optim.Optimizer):
202 scheduler (torch.optim._LRScheduler)
203 """
204 self._optimizer = optimizer
205 self._scheduler = scheduler
206
207 # NOTE: some heuristics on what LR to summarize
208 # summarize the param group with most parameters
209 largest_group = max(len(g["params"]) for g in optimizer.param_groups)
210
211 if largest_group == 1:
212 # If all groups have one parameter,
213 # then find the most common initial LR, and use it for summary
214 lr_count = Counter([g["lr"] for g in optimizer.param_groups])
215 lr = lr_count.most_common()[0][0]
216 for i, g in enumerate(optimizer.param_groups):
217 if g["lr"] == lr:
218 self._best_param_group_id = i
219 break
220 else:
221 for i, g in enumerate(optimizer.param_groups):
222 if len(g["params"]) == largest_group:
223 self._best_param_group_id = i
224 break
225
Member Function Documentation
◆ after_step()
| fastreid.engine.hooks.LRScheduler.after_step | ( | self | ) |
Called after each iteration.
Reimplemented from fastreid.engine.train_loop.HookBase.
Definition at line 226 of file hooks.py.
226 def after_step(self):
227 lr = self._optimizer.param_groups[self._best_param_group_id]["lr"]
228 self.trainer.storage.put_scalar("lr", lr, smoothing_hint=False)
229 self._scheduler.step()
230
231
Member Data Documentation
◆ _best_param_group_id
|
protected |
◆ _optimizer
◆ _scheduler
The documentation for this class was generated from the following file:
- smreid/fastreid/engine/hooks.py
