Make variable names more consistent (#45)

* Make variable names more consistent

* update

Author: rasbt

ch02/01_main-chapter-code/ch02_exercise-solutions.ipynb

@@ -112,8 +112,8 @@
     "\n",
     "download_qwen3_small(kind=\"base\", tokenizer_only=True, out_dir=\"qwen3\")\n",
     "\n",
-    "tokenizer_file_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
-    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)\n",
+    "tokenizer_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
+    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)\n",
     "\n",
     "prompt = \"Hello, Ardwarklethyrx. Haus und Garten.\"\n",
     "input_token_ids_list = tokenizer.encode(prompt)\n",
@@ -174,12 +174,12 @@
     "\n",
     "download_qwen3_small(kind=\"base\", tokenizer_only=False, out_dir=\"qwen3\")\n",
     "\n",
-    "tokenizer_file_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
-    "model_file = Path(\"qwen3\") / \"qwen3-0.6B-base.pth\"\n",
+    "tokenizer_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
+    "model_path = Path(\"qwen3\") / \"qwen3-0.6B-base.pth\"\n",
     "\n",
-    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)\n",
+    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)\n",
     "model = Qwen3Model(QWEN_CONFIG_06_B)\n",
-    "model.load_state_dict(torch.load(model_file))\n",
+    "model.load_state_dict(torch.load(model_path))\n",
     "\n",
     "model.to(device)\n",
     "\n",
@@ -372,12 +372,12 @@
     "\n",
     "download_qwen3_small(kind=\"base\", tokenizer_only=False, out_dir=\"qwen3\")\n",
     "\n",
-    "tokenizer_file_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
-    "model_file = Path(\"qwen3\") / \"qwen3-0.6B-base.pth\"\n",
+    "tokenizer_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
+    "model_path = Path(\"qwen3\") / \"qwen3-0.6B-base.pth\"\n",
     "\n",
-    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)\n",
+    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)\n",
     "model = Qwen3Model(QWEN_CONFIG_06_B)\n",
-    "model.load_state_dict(torch.load(model_file))\n",
+    "model.load_state_dict(torch.load(model_path))\n",
     "\n",
     "model.to(device);"
    ]

ch02/01_main-chapter-code/ch02_main.ipynb

@@ -367,8 +367,8 @@
     "from pathlib import Path\n",
     "from reasoning_from_scratch.qwen3 import Qwen3Tokenizer\n",
     "\n",
-    "tokenizer_file_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
-    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)"
+    "tokenizer_path = Path(\"qwen3\") / \"tokenizer-base.json\"\n",
+    "tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)"
    ]
   },
   {
@@ -640,10 +640,10 @@
    "source": [
     "from reasoning_from_scratch.qwen3 import Qwen3Model, QWEN_CONFIG_06_B\n",
     "\n",
-    "model_file = Path(\"qwen3\") / \"qwen3-0.6B-base.pth\"\n",
+    "model_path = Path(\"qwen3\") / \"qwen3-0.6B-base.pth\"\n",
     "\n",
     "model = Qwen3Model(QWEN_CONFIG_06_B)\n",
-    "model.load_state_dict(torch.load(model_file))\n",
+    "model.load_state_dict(torch.load(model_path))\n",
     "\n",
     "model.to(device)"
    ]

ch02/03_optimized-LLM/compare_inference.py

@@ -80,23 +80,23 @@
 
 if args.reasoning:
     download_qwen3_small(kind="reasoning", tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file_path = Path("qwen3") / "tokenizer-reasoning.json"
-    model_file = Path("qwen3") / "qwen3-0.6B-reasoning.pth"
+    tokenizer_path = Path("qwen3") / "tokenizer-reasoning.json"
+    model_path = Path("qwen3") / "qwen3-0.6B-reasoning.pth"
     tokenizer = Qwen3Tokenizer(
-        tokenizer_file_path=tokenizer_file_path,
+        tokenizer_file_path=tokenizer_path,
         apply_chat_template=True,
         add_generation_prompt=True,
         add_thinking=True
     )
 
 else:
     download_qwen3_small(kind="base", tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file_path = Path("qwen3") / "tokenizer-base.json"
-    model_file = Path("qwen3") / "qwen3-0.6B-base.pth"
-    tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)
+    tokenizer_path = Path("qwen3") / "tokenizer-base.json"
+    model_path = Path("qwen3") / "qwen3-0.6B-base.pth"
+    tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)
 
 model = Qwen3Model(QWEN_CONFIG_06_B)
-model.load_state_dict(torch.load(model_file, map_location=device))
+model.load_state_dict(torch.load(model_path, map_location=device))
 
 model.to(device)
 

chG/01_main-chapter-code/qwen3_chat_interface.py

@@ -32,23 +32,23 @@
 def get_model_and_tokenizer(qwen3_config, local_dir, device, use_compile, use_reasoning):
     if use_reasoning:
         download_qwen3_small(kind="reasoning", tokenizer_only=False, out_dir=local_dir)
-        tokenizer_file_path = Path(local_dir) / "tokenizer-reasoning.json"
-        model_file = Path(local_dir) / "qwen3-0.6B-reasoning.pth"
+        tokenizer_path = Path(local_dir) / "tokenizer-reasoning.json"
+        model_path = Path(local_dir) / "qwen3-0.6B-reasoning.pth"
         tokenizer = Qwen3Tokenizer(
-            tokenizer_file_path=tokenizer_file_path,
+            tokenizer_file_path=tokenizer_path,
             apply_chat_template=True,
             add_generation_prompt=True,
             add_thinking=True
         )
 
     else:
         download_qwen3_small(kind="base", tokenizer_only=False, out_dir=local_dir)
-        tokenizer_file_path = Path(local_dir) / "tokenizer-base.json"
-        model_file = Path(local_dir) / "qwen3-0.6B-base.pth"
-        tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)
+        tokenizer_path = Path(local_dir) / "tokenizer-base.json"
+        model_path = Path(local_dir) / "qwen3-0.6B-base.pth"
+        tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)
 
     model = Qwen3Model(qwen3_config)
-    model.load_state_dict(torch.load(model_file, map_location=device))
+    model.load_state_dict(torch.load(model_path, map_location=device))
     model.to(device)
     if use_compile:
         model = torch.compile(model)

chG/01_main-chapter-code/qwen3_chat_interface_multiturn.py

@@ -43,23 +43,23 @@ def trim_input_tensor(input_ids_tensor, context_len, max_new_tokens):
 def get_model_and_tokenizer(qwen3_config, local_dir, device, use_compile, use_reasoning):
     if use_reasoning:
         download_qwen3_small(kind="reasoning", tokenizer_only=False, out_dir=local_dir)
-        tokenizer_file_path = Path(local_dir) / "tokenizer-reasoning.json"
-        model_file = Path(local_dir) / "qwen3-0.6B-reasoning.pth"
+        tokenizer_path = Path(local_dir) / "tokenizer-reasoning.json"
+        model_path = Path(local_dir) / "qwen3-0.6B-reasoning.pth"
         tokenizer = Qwen3Tokenizer(
-            tokenizer_file_path=tokenizer_file_path,
+            tokenizer_file_path=tokenizer_path,
             apply_chat_template=True,
             add_generation_prompt=True,
             add_thinking=True
         )
 
     else:
         download_qwen3_small(kind="base", tokenizer_only=False, out_dir=local_dir)
-        tokenizer_file_path = Path(local_dir) / "tokenizer-base.json"
-        model_file = Path(local_dir) / "qwen3-0.6B-base.pth"
-        tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)
+        tokenizer_path = Path(local_dir) / "tokenizer-base.json"
+        model_path = Path(local_dir) / "qwen3-0.6B-base.pth"
+        tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)
 
     model = Qwen3Model(qwen3_config)
-    model.load_state_dict(torch.load(model_file, map_location=device))
+    model.load_state_dict(torch.load(model_path, map_location=device))
     model.to(device)
     if use_compile:
         model = torch.compile(model)

tests/examples/01_quick-example.py

@@ -76,23 +76,23 @@
 
 if args.reasoning:
     download_qwen3_small(kind="reasoning", tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file_path = Path("qwen3") / "tokenizer-reasoning.json"
-    model_file = Path("qwen3") / "qwen3-0.6B-reasoning.pth"
+    tokenizer_path = Path("qwen3") / "tokenizer-reasoning.json"
+    model_path = Path("qwen3") / "qwen3-0.6B-reasoning.pth"
     tokenizer = Qwen3Tokenizer(
-        tokenizer_file_path=tokenizer_file_path,
+        tokenizer_file_path=tokenizer_path,
         apply_chat_template=True,
         add_generation_prompt=True,
         add_thinking=True
     )
 
 else:
     download_qwen3_small(kind="base", tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file_path = Path("qwen3") / "tokenizer-base.json"
-    model_file = Path("qwen3") / "qwen3-0.6B-base.pth"
-    tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_file_path)
+    tokenizer_path = Path("qwen3") / "tokenizer-base.json"
+    model_path = Path("qwen3") / "qwen3-0.6B-base.pth"
+    tokenizer = Qwen3Tokenizer(tokenizer_file_path=tokenizer_path)
 
 model = Qwen3Model(QWEN_CONFIG_06_B)
-model.load_state_dict(torch.load(model_file, map_location=device))
+model.load_state_dict(torch.load(model_path, map_location=device))
 
 model.to(device)
 

tests/test_qwen3_batched.py

@@ -37,22 +37,22 @@ def test_batched_vs_nonbatched_equivalence_with_batched_model(reasoning):
     # Download and init tokenizer
     kind = "reasoning" if reasoning else "base"
     download_qwen3_small(kind=kind, tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file = Path("qwen3") / (
+    tokenizer_path = Path("qwen3") / (
         "tokenizer-reasoning.json" if reasoning else "tokenizer-base.json"
     )
-    model_file = Path("qwen3") / (
+    model_path = Path("qwen3") / (
         "qwen3-0.6B-reasoning.pth" if reasoning else "qwen3-0.6B-base.pth"
     )
     tokenizer = Qwen3Tokenizer(
-        tokenizer_file_path=tokenizer_file,
+        tokenizer_file_path=tokenizer_path,
         apply_chat_template=True if reasoning else False,
         add_generation_prompt=True if reasoning else False,
         add_thinking=True if reasoning else False,
     )
 
     # Models
     model_batched = Qwen3ModelBatched(QWEN_CONFIG_06_B)
-    model_batched.load_state_dict(torch.load(model_file, map_location=device))
+    model_batched.load_state_dict(torch.load(model_path, map_location=device))
     model_batched.to(device)
     model_batched.eval()
 
@@ -197,27 +197,27 @@ def test_batched_vs_nonbatched_equivalence_with_single_versus_batched_model(reas
     # Download and init tokenizer
     kind = "reasoning" if reasoning else "base"
     download_qwen3_small(kind=kind, tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file = Path("qwen3") / (
+    tokenizer_path = Path("qwen3") / (
         "tokenizer-reasoning.json" if reasoning else "tokenizer-base.json"
     )
-    model_file = Path("qwen3") / (
+    model_path = Path("qwen3") / (
         "qwen3-0.6B-reasoning.pth" if reasoning else "qwen3-0.6B-base.pth"
     )
     tokenizer = Qwen3Tokenizer(
-        tokenizer_file_path=tokenizer_file,
+        tokenizer_file_path=tokenizer_path,
         apply_chat_template=True if reasoning else False,
         add_generation_prompt=True if reasoning else False,
         add_thinking=True if reasoning else False,
     )
 
     # Models
     model = Qwen3Model(QWEN_CONFIG_06_B)
-    model.load_state_dict(torch.load(model_file, map_location=device))
+    model.load_state_dict(torch.load(model_path, map_location=device))
     model.to(device)
     model.eval()
 
     model_batched = Qwen3ModelBatched(QWEN_CONFIG_06_B)
-    model_batched.load_state_dict(torch.load(model_file, map_location=device))
+    model_batched.load_state_dict(torch.load(model_path, map_location=device))
     model_batched.to(device)
     model_batched.eval()
 
@@ -290,27 +290,27 @@ def test_plain_vs_streaming_generation(reasoning):
     # Download and init tokenizer
     kind = "reasoning" if reasoning else "base"
     download_qwen3_small(kind=kind, tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file = Path("qwen3") / (
+    tokenizer_path = Path("qwen3") / (
         "tokenizer-reasoning.json" if reasoning else "tokenizer-base.json"
     )
-    model_file = Path("qwen3") / (
+    model_path = Path("qwen3") / (
         "qwen3-0.6B-reasoning.pth" if reasoning else "qwen3-0.6B-base.pth"
     )
     tokenizer = Qwen3Tokenizer(
-        tokenizer_file_path=tokenizer_file,
+        tokenizer_file_path=tokenizer_path,
         apply_chat_template=True if reasoning else False,
         add_generation_prompt=True if reasoning else False,
         add_thinking=True if reasoning else False,
     )
 
     # Models
     model = Qwen3Model(QWEN_CONFIG_06_B)
-    model.load_state_dict(torch.load(model_file, map_location=device))
+    model.load_state_dict(torch.load(model_path, map_location=device))
     model.to(device)
     model.eval()
 
     model_batched = Qwen3ModelBatched(QWEN_CONFIG_06_B)
-    model_batched.load_state_dict(torch.load(model_file, map_location=device))
+    model_batched.load_state_dict(torch.load(model_path, map_location=device))
     model_batched.to(device)
     model_batched.eval()
 

tests/test_qwen3_optimized.py

@@ -86,27 +86,27 @@ def test_qwen3_vs_optimized_qwen3(reasoning):
     # Download and init tokenizer
     kind = "reasoning" if reasoning else "base"
     download_qwen3_small(kind=kind, tokenizer_only=False, out_dir="qwen3")
-    tokenizer_file = Path("qwen3") / (
+    tokenizer_path = Path("qwen3") / (
         "tokenizer-reasoning.json" if reasoning else "tokenizer-base.json"
     )
-    model_file = Path("qwen3") / (
+    model_path = Path("qwen3") / (
         "qwen3-0.6B-reasoning.pth" if reasoning else "qwen3-0.6B-base.pth"
     )
     tokenizer = Qwen3Tokenizer(
-        tokenizer_file_path=tokenizer_file,
+        tokenizer_file_path=tokenizer_path,
         apply_chat_template=True if reasoning else False,
         add_generation_prompt=True if reasoning else False,
         add_thinking=True if reasoning else False,
     )
 
     # Models
     model = Qwen3Model(QWEN_CONFIG_06_B)
-    model.load_state_dict(torch.load(model_file, map_location=device))
+    model.load_state_dict(torch.load(model_path, map_location=device))
     model.to(device)
     model.eval()
 
     model_optimized = Qwen3ModelOptimized(QWEN_CONFIG_06_B, exact=True)
-    model_optimized.load_state_dict(torch.load(model_file, map_location=device))
+    model_optimized.load_state_dict(torch.load(model_path, map_location=device))
     model_optimized.to(device)
     model_optimized.eval()