diff --git a/optimum/habana/transformers/models/detr/modeling_detr.py b/optimum/habana/transformers/models/detr/modeling_detr.py
index 7eef317372..079898d2d7 100644
--- a/optimum/habana/transformers/models/detr/modeling_detr.py
+++ b/optimum/habana/transformers/models/detr/modeling_detr.py
@@ -1,7 +1,7 @@
 def gaudi_DetrConvModel_forward(self, pixel_values, pixel_mask):
     """
     Copied from modeling_detr: https://github.com/huggingface/transformers/blob/main/src/transformers/models/detr/modeling_detr.py#L398
-    The modications are:
+    The modifications are:
         - Use CPU to calculate the position_embeddings and transfer back to HPU
     """
 
diff --git a/optimum/habana/transformers/trainer.py b/optimum/habana/transformers/trainer.py
index 622e6a11db..ec7d31e3a6 100644
--- a/optimum/habana/transformers/trainer.py
+++ b/optimum/habana/transformers/trainer.py
@@ -1278,7 +1278,7 @@ def _maybe_log_save_evaluate(self, tr_loss, _grad_norm, model, trial, epoch, ign
             tr_loss -= tr_loss
             logs["loss"] = round(tr_loss_scalar / (self.state.global_step - self._globalstep_last_logged), 4)
 
-            # This grad_norm block was outside of _maybe_log_save_evaluate method causing perf degradataion.
+            # This grad_norm block was outside of _maybe_log_save_evaluate method causing perf degradation.
             # Moving it here so the grad tensor is only copied when it's needed.
             if self.accelerator.distributed_type == GaudiDistributedType.DEEPSPEED:
                 grad_norm = model.get_global_grad_norm()
diff --git a/optimum/habana/transformers/training_args.py b/optimum/habana/transformers/training_args.py
index 79e43efa72..4a2b12593f 100644
--- a/optimum/habana/transformers/training_args.py
+++ b/optimum/habana/transformers/training_args.py
@@ -119,7 +119,7 @@ class GaudiTrainingArguments(TrainingArguments):
         non_blocking_data_copy (`bool`, *optional*, defaults to `False`):
             Whether to enable async data copy when preparing inputs.
         profiling_warmup_steps (`int`, *optional*, defaults to 0):
-            Number of steps to ignore for profling.
+            Number of steps to ignore for profiling.
         profiling_steps (`int`, *optional*, defaults to 0):
             Number of steps to be captured when enabling profiling.
     """
@@ -235,7 +235,7 @@ class GaudiTrainingArguments(TrainingArguments):
 
     profiling_warmup_steps: Optional[int] = field(
         default=0,
-        metadata={"help": ("Number of steps to ignore for profling.")},
+        metadata={"help": ("Number of steps to ignore for profiling.")},
     )
 
     profiling_steps: Optional[int] = field(