update README; remove print; make style

finetrainers/args.py

@@ -60,7 +60,9 @@ class Args:
     # Training arguments
     training_type: str = None
     seed: int = 42
-    mixed_precision: str = None # TODO: consider removing later https://github.com/a-r-r-o-w/finetrainers/pull/139#discussion_r1897438414
+    mixed_precision: str = (
+        None  # TODO: consider removing later https://github.com/a-r-r-o-w/finetrainers/pull/139#discussion_r1897438414
+    )
     batch_size: int = 1
     train_epochs: int = 1
     train_steps: int = None

finetrainers/trainer.py

@@ -520,7 +520,7 @@ class Trainer:
         )
         total_training_steps = self.state.train_steps * self.state.accelerator.num_processes
         num_warmup_steps = self.args.lr_warmup_steps * self.state.accelerator.num_processes
-        
+
         if use_deepspeed_lr_scheduler:
             from accelerate.utils import DummyScheduler
 
@@ -714,7 +714,6 @@ class Trainer:
                     loss = loss.mean(list(range(1, loss.ndim)))
                     # Average loss across batch dimension
                     loss = loss.mean()
-                    print(f"{self.transformer.dtype=}, {accelerator.mixed_precision=}")
                     accelerator.backward(loss)
 
                     if accelerator.sync_gradients: