diff --git a/README.md b/README.md
index 7b941c5..453e07e 100644
--- a/README.md
+++ b/README.md
@@ -14,6 +14,12 @@ Alongside it sits a parallel PyTorch implementation in [engine/main.py](engine/m
 
 The point of this repo is the C++ core. The PyTorch, FastAPI, and frontend layers exist to make the model usable, but if you're here to learn how a GPT is actually built and trained without a framework doing the work for you, [include/backward.h](include/backward.h) is where to start reading.
 
+<h1 align="center">
+<img width="824" height="250" alt="image" src="https://github.com/user-attachments/assets/5c65daa2-903a-4392-85e3-56442bb82cac" />
+
+
+</h1>
+
 ***technical notes***: [docs](https://eamon2009.github.io/LLMs/)
 
 
diff --git a/assets/run_2026-07-16 165731.png b/assets/run_2026-07-16 165731.png
new file mode 100644
index 0000000..3372892
Binary files /dev/null and b/assets/run_2026-07-16 165731.png differ
diff --git a/engine/main.py b/engine/main.py
index 5c15109..0725493 100644
--- a/engine/main.py
+++ b/engine/main.py
@@ -66,23 +66,23 @@ def success(msg): log(f"  ok  {msg}")
 
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
 dropout = 0.1
-block_size = 256     
-n_embd = 192         
-n_head = 6           
-n_layer = 6          
-batch_size = 64      
-max_iters = 5000     
-eval_interval = 250  
-learning_rate = 6e-4 
-eval_iters = 200     
-dropout = 0.1        
+block_size = 256
+n_embd = 192
+n_head = 6
+n_layer = 6
+batch_size = 64
+max_iters = 5000
+eval_interval = 250
+learning_rate = 6e-4
+eval_iters = 200
+dropout = 0.1
 
 torch.manual_seed(seed)
 
 
 # tokenizer
 
-def get_tokenizer(encoding_name="o200k"):
+def get_tokenizer(encoding_name="o200k_base"):
     tokenizer = tiktoken.get_encoding(encoding_name)
     vocab_size = tokenizer.n_vocab
     return tokenizer, vocab_size
@@ -96,7 +96,7 @@ def decode(tokens, tokenizer): return tokenizer.decode(tokens)
 with open(cleaned_path, 'r', encoding='utf-8') as f:
     text = f.read()
 
-tokenizer, vocab_size = get_tokenizer("o200k")
+tokenizer, vocab_size = get_tokenizer("o200k_base")
 encoded_data = encode(text, tokenizer)
 
 data = torch.tensor(encoded_data, dtype=torch.long)