diff --git a/appendix-D/01_main-chapter-code/appendix-D.ipynb b/appendix-D/01_main-chapter-code/appendix-D.ipynb
index 64c6d67..f8a5aca 100644
--- a/appendix-D/01_main-chapter-code/appendix-D.ipynb
+++ b/appendix-D/01_main-chapter-code/appendix-D.ipynb
@@ -615,7 +615,7 @@
     "                if global_step > warmup_steps:\n",
     "                    torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)  \n",
     "            else:\n",
-    "                if global_step >= warmup_steps:  # the book originally used global_step > warmup_steps, which lead to a skipped clipping step after warmup\n",
+    "                if global_step >= warmup_steps:  # the book originally used global_step > warmup_steps, which led to a skipped clipping step after warmup\n",
     "                    torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)\n",
     "                \n",
     "            optimizer.step()\n",
diff --git a/pkg/llms_from_scratch/appendix_d.py b/pkg/llms_from_scratch/appendix_d.py
index aa3cda4..e6b9923 100644
--- a/pkg/llms_from_scratch/appendix_d.py
+++ b/pkg/llms_from_scratch/appendix_d.py
@@ -66,7 +66,7 @@ def train_model(model, train_loader, val_loader, optimizer, device,
                 if global_step > warmup_steps:
                     torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)
             else:
-                if global_step >= warmup_steps:  # the book originally used global_step > warmup_steps, which lead to a skipped clipping step after warmup
+                if global_step >= warmup_steps:  # the book originally used global_step > warmup_steps, which led to a skipped clipping step after warmup
                     torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)
 
             optimizer.step()
diff --git a/pkg/llms_from_scratch/ch03.py b/pkg/llms_from_scratch/ch03.py
index 99ff4a7..e6ca561 100644
--- a/pkg/llms_from_scratch/ch03.py
+++ b/pkg/llms_from_scratch/ch03.py
@@ -160,7 +160,7 @@ class PyTorchMultiHeadAttention(nn.Module):
     def __init__(self, d_in, d_out, num_heads, dropout=0.0, qkv_bias=False):
         super().__init__()
 
-        assert d_out % num_heads == 0, "embed_dim is indivisible by num_heads"
+        assert d_out % num_heads == 0, "d_out is indivisible by num_heads"
 
         self.num_heads = num_heads
         self.head_dim = d_out // num_heads