diff --git a/en-wordlist.txt b/en-wordlist.txt
index 56cb35ba58..d06a376751 100644
--- a/en-wordlist.txt
+++ b/en-wordlist.txt
@@ -1,6 +1,7 @@
 
 ACL
 ADI
+ALiBi
 AOT
 AOTInductor
 APIs
@@ -80,6 +81,7 @@ FX
 FX's
 FairSeq
 Fastpath
+FFN
 FloydHub
 FloydHub's
 Frobenius
@@ -128,7 +130,7 @@ Kihyuk
 Kiuk
 Kubernetes
 Kuei
-KV-Caching
+KV
 LRSchedulers
 LSTM
 LSTMs
@@ -164,6 +166,7 @@ NLP
 NTK
 NUMA
 NaN
+NaNs
 NanoGPT
 Netron
 NeurIPS
@@ -232,6 +235,7 @@ Sigmoid
 SoTA
 Sohn
 Spacy
+SwiGLU
 TCP
 THP
 TIAToolbox
@@ -467,6 +471,7 @@ nheads
 nightlies
 NJT
 NJTs
+NJT's
 num
 numericalize
 numpy
diff --git a/intermediate_source/transformer_building_blocks.py b/intermediate_source/transformer_building_blocks.py
index 95118eb0b9..c18fb22193 100644
--- a/intermediate_source/transformer_building_blocks.py
+++ b/intermediate_source/transformer_building_blocks.py
@@ -344,7 +344,7 @@ def benchmark(func, *args, **kwargs):
 torch.manual_seed(6)
 vanilla_mha_layer = nn.MultiheadAttention(E_q, nheads, dropout=dropout, batch_first=True, bias=bias, device='cuda')
 
-# nn.MultiheadAttention uses a non conventional initialization for layers, so do this for exact parity :(
+# ``nn.MultiheadAttention`` uses a non conventional initialization for layers, so do this for exact parity :(
 mha_layer.out_proj.weight = nn.Parameter(vanilla_mha_layer.out_proj.weight.clone().detach())
 mha_layer.packed_proj.weight = nn.Parameter(vanilla_mha_layer.in_proj_weight.clone().detach())
 mha_layer.out_proj.bias = nn.Parameter(vanilla_mha_layer.out_proj.bias.clone().detach())
@@ -421,7 +421,7 @@ def benchmark(func, *args, **kwargs):
 # gives equivalent results to an ``nn.TransformerEncoderLayer`` with
 # ``is_causal=True``.
 #
-# We  demonstrate examples of implementing the rest of the nn layers
+# We  demonstrate examples of implementing the rest of the ``nn`` layers
 # `here <https://github.com/mikaylagawarecki/temp>`_ but omit that from this
 # tutorial for brevity.
 
@@ -457,7 +457,7 @@ def benchmark(func, *args, **kwargs):
 # * SwiGLU activation in feed-forward network of Transformer Layer
 # 
 # Input projection for MultiheadAttention
-# ----------------------------------------
+# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 # Recall that when doing self-attention, the ``query``, ``key`` and ``value``
 # are the same tensor. Each of these tensors is projected with a 
 # ``Linear(E_q, E_total)`` layer. Instead, we can pack this into one layer,
@@ -502,8 +502,9 @@ def forward(self, query):
 (q_out, k_out, v_out), time_packed, _ = benchmark(packed_in_proj, q)
 print(f"InputProjection: {time:5f} s, PackedInputProjection: {time_packed:5f} s, speedup: {time/time_packed:.2f}x")
 
+##################################################
 # SwiGLU feed forward network of Transformer Layer
-# ------------------------------------------------
+# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 # SwiGLU is a non-linear activation function that is increasingly popular in the feed-forward
 # network of the transformer layer (e.g. Llama). A feed-forward network with SwiGLU activation is defined as
 
@@ -524,6 +525,7 @@ def __init__(self, dim, hidden_dim, multiple_of, ffn_dim_multiplier=None, device
     def forward(self, x):
         return self.w2(F.silu(self.w1(x)) * self.w3(x))
 
+########################################################################
 # An alternative way of implementing this that uses packed projection is 
 
 class PackedSwiGLUFFN(nn.Module):
@@ -543,6 +545,7 @@ def forward(self, x):
         x1, x3 = torch.chunk(self.w13(x), 2, dim=-1)
         return self.w2(F.silu(x1) * x3)
 
+################################################################################
 # We can compare the performance of the two implementations as follows
 # Depending on your hardware, you might see different results. On an A100 I see
 # 1.12x speedup for D=128.
@@ -635,20 +638,14 @@ def alibi_mod(score, b, h, q_idx, kv_idx):
 )
 out_flex2 = flex_attention(query, key, value, score_mod=alibi_score_mod)
 
-################################################################################
-# And more
-# --------
-# 
-# We intend to update this tutorial to demonstrate more examples of how to use
-# the various performant building blocks such as KV-Caching, Grouped Query Attention
-# etc.
-
 
 ################################################################################
 # Extended examples
 # -----------------
 # 
-# There are several good examples of using various performant building blocks to
+# We intend to update this tutorial to demonstrate more examples of how to use
+# the various performant building blocks such as KV-Caching, Grouped Query Attention
+# etc. Further, there are several good examples of using various performant building blocks to
 # implement various transformer architectures. Some examples include
 #
 # * `gpt-fast <https://github.com/pytorch-labs/gpt-fast>`_