Fix code comment: embed_dim -> d_out (#698)

ch03/02_bonus_efficient-multihead-attention/mha-implementations.ipynb

@@ -352,7 +352,7 @@
     "    def __init__(self, d_in, d_out, num_heads, context_length, dropout=0.0, qkv_bias=False):\n",
     "        super().__init__()\n",
     "\n",
-    "        assert d_out % num_heads == 0, \"embed_dim is indivisible by num_heads\"\n",
+    "        assert d_out % num_heads == 0, \"d_out is indivisible by num_heads\"\n",
     "\n",
     "        self.num_heads = num_heads\n",
     "        self.context_length = context_length\n",
@@ -588,7 +588,7 @@
     "    def __init__(self, d_in, d_out, num_heads, context_length, dropout=0.0, qkv_bias=False):\n",
     "        super().__init__()\n",
     "\n",
-    "        assert d_out % num_heads == 0, \"embed_dim is indivisible by num_heads\"\n",
+    "        assert d_out % num_heads == 0, \"d_out is indivisible by num_heads\"\n",
     "\n",
     "        self.num_heads = num_heads\n",
     "        self.context_length = context_length\n",

ch05/10_llm-training-speed/01_opt_single_gpu.py

@@ -65,7 +65,7 @@ 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

ch05/10_llm-training-speed/02_opt_multi_gpu_ddp.py

@@ -108,7 +108,7 @@ 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