Simplify track grad norm condition (#9992)

Author: carmocca

pytorch_lightning/loops/optimization/optimizer_loop.py

@@ -478,9 +478,7 @@ def _track_and_norm_grad(self, optimizer: torch.optim.Optimizer, opt_idx: int) -
         """
         # track gradient norms
         grad_norm_dict = {}
-        can_log = (self.trainer.global_step + 1) % self.trainer.log_every_n_steps == 0
-        should_track = float(self.trainer.track_grad_norm) > 0
-        if should_track and can_log:
+        if self.trainer.track_grad_norm != -1:
             grad_norm_dict = grad_norm(self.trainer.lightning_module, self.trainer.track_grad_norm)
 
         # clip gradients