diff --git a/config_files/training/config_lorem_ipsum.yaml b/config_files/training/config_lorem_ipsum.yaml
index fc1d8926..e4251a2f 100644
--- a/config_files/training/config_lorem_ipsum.yaml
+++ b/config_files/training/config_lorem_ipsum.yaml
@@ -42,7 +42,7 @@ train_dataloader:
     num_workers: 2
     pin_memory: true
     shuffle: false
-    dataloader_tag: "train"
+    dataloader_tag: train
     dataset:
       instance_key: train_dataset
       pass_type: BY_REFERENCE
@@ -264,7 +264,7 @@ optimizer:
     betas: [0.9, 0.95]
     eps: 1e-8
     weight_decay: 1e-1
-    weight_decay_groups_excluded: ["embedding", "layernorm"]
+    weight_decay_groups_excluded: [embedding, layernorm]
     wrapped_model: 
       instance_key: wrapped_model
       pass_type: BY_REFERENCE
diff --git a/src/modalities/batch.py b/src/modalities/batch.py
index 5a47511c..633dc269 100644
--- a/src/modalities/batch.py
+++ b/src/modalities/batch.py
@@ -105,11 +105,9 @@ class EvaluationResultBatch(Batch):
     throughput_metrics: Dict[str, torch.Tensor] = field(default_factory=lambda: dict())
 
     def __str__(self) -> str:
-        eval_str = f"Evaluation result on dataset tag {self.dataloader_tag} after {self.num_train_steps_done} steps:"
-        eval_str += "\n\nlosses: " + "\n\t".join([f"{k}: {v.mean().item()}" for k, v in self.losses.items()])
-        eval_str += "\n\nmetrics: " + "\n\t".join([f"{k}: {v.mean().item()}" for k, v in self.metrics.items()])
-        eval_str += "\n\nthroughput metrics: " + "\n\t".join(
-            [f"{k}: {v.mean().item()}" for k, v in self.throughput_metrics.items()]
-        )
-        eval_str += "\n==============================================="
+        eval_str = f"Dataloader: {self.dataloader_tag} | "
+        eval_str = f"step: {self.num_train_steps_done} | "
+        eval_str += " | ".join([f"{k}: {v.mean().item()}" for k, v in self.throughput_metrics.items()]) + " | "
+        eval_str += " | ".join([f"{k}: {v.mean().item()}" for k, v in self.losses.items()]) + " | "
+        eval_str += " | ".join([f"{k}: {v.mean().item()}" for k, v in self.metrics.items()]) + " | "
         return eval_str
diff --git a/src/modalities/trainer.py b/src/modalities/trainer.py
index 51aaa273..2f8a72ad 100644
--- a/src/modalities/trainer.py
+++ b/src/modalities/trainer.py
@@ -141,10 +141,6 @@ def train(
                 num_train_steps_done=num_train_steps_done,
                 dataloader_tag=train_loader.dataloader_tag,
             )
-            print(
-                f"num_train_steps_done: {num_train_steps_done}, micro_batch_id: {micro_batch_id}",
-                f" (micro_batch_id +1) % GAS: {(micro_batch_id +1) % self.gradient_acc_steps}",
-            )
             # Check if model performance should be logged
             if num_train_steps_done % training_log_interval_in_steps == 0 and step_performed:
                 forward_backward_time = torch.tensor(forward_backward_time_recorder.delta_t).to(device)
@@ -175,15 +171,15 @@ def train(
                     reduced_losses[1],
                 )
                 losses = {
-                    f"{loss_fun.tag} average": train_loss_avg,
-                    f"{loss_fun.tag} last step": train_loss_last_batch,
+                    "train loss avg": train_loss_avg,
+                    "train loss last": train_loss_last_batch,
                 }
 
                 consumed_tokens = torch.Tensor([num_train_steps_done * self.global_num_tokens_per_train_step])
                 metrics = {
-                    "consumed_tokens": consumed_tokens,
-                    "grad_norm_avg": torch.mean(torch.Tensor(gradient_norm_scores)),
-                    "grad_norm_last_batch": gradient_norm_scores[-1],
+                    "consumed tokens": consumed_tokens,
+                    "grad norm avg": torch.mean(torch.Tensor(gradient_norm_scores)),
+                    "grad norm last": torch.tensor(gradient_norm_scores[-1]),
                 }
                 gradient_norm_scores = []
 
@@ -192,13 +188,14 @@ def train(
                     metrics=metrics,
                     # TODO: hardcoded metric key
                     throughput_metrics={
-                        "training_synced_num_samples_per_second": synced_num_samples_per_second,
-                        "lr_mean": torch.tensor(scheduler.get_last_lr()).mean(),
-                        "lr_first": torch.tensor(scheduler.get_last_lr())[0],
+                        "train samples/s": synced_num_samples_per_second,
+                        "lr mean": torch.tensor(scheduler.get_last_lr()).mean(),
                     },
                     dataloader_tag=train_loader.dataloader_tag,
                     num_train_steps_done=num_train_steps_done,
                 )
+                if self.local_rank == 0:
+                    print(training_metrics)
                 self._publish_evaluation_result(
                     evaluation_result_publisher=self.evaluation_result_publisher,
                     evaluation_result=training_metrics,