maint: Implement comprehensive .gitignore rule suite

.github/workflows/release.yml

@@ -32,8 +32,6 @@ jobs:
             tag_name="v${raw_tag}"
           fi
           echo "tag_name=${tag_name}" >> "$GITHUB_OUTPUT"
-
-  # macOS Jobs
   macos-arm64-cpu:
     name: macOS Apple Silicon (arm64) CPU
     needs: release-metadata
@@ -134,8 +132,6 @@ jobs:
           path: llm.cpp-${{ needs.release-metadata.outputs.tag_name }}-ios-xcframework.tar.gz
           if-no-files-found: error
           retention-days: 30
-
-  # Ubuntu Jobs
   ubuntu-x64-cpu:
     name: Ubuntu x64 (CPU)
     needs: release-metadata
@@ -457,8 +453,6 @@ jobs:
           path: llm.cpp-${{ needs.release-metadata.outputs.tag_name }}-bin-ubuntu-x64-sycl-fp16.tar.gz
           if-no-files-found: error
           retention-days: 30
-
-  # Android Job
   android-arm64-cpu:
     name: Android arm64 (CPU)
     needs: release-metadata
@@ -516,8 +510,6 @@ jobs:
           path: llm.cpp-${{ needs.release-metadata.outputs.tag_name }}-bin-android-arm64-cpu.tar.gz
           if-no-files-found: error
           retention-days: 30
-
-  # Windows Jobs
   windows-x64-cpu:
     name: Windows x64 (CPU)
     needs: release-metadata

.gitignore

@@ -57,3 +57,6 @@ Thumbs.db
 *.swo
 .idea
 docker-compose.override.yml
+libs/lmgnu/build/
+libs/lmgnu/dist/
+libs/lmgnu/*.egg-info/

.gitmodules

@@ -0,0 +1,4 @@
+[submodule "libs/lmgnu"]
+    path = libs/lmgnu
+    url = https://github.com/LMGNU/LMGNU.git
+    branch = master

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/)
 
 

config/config.h

@@ -5,15 +5,15 @@ static const std::string DEFAULT_CLEANED_PATH = "data/input.txt";
 static const std::string DATA_PATH_ENV_VAR = "GPT_DATA_PATH";
 static const unsigned int SEED = 1337;
 static const double TRAIN_SPLIT = 0.9; // 90% train, 10% val
-static const int BATCH_SIZE = 16;      
-static const int BLOCK_SIZE = 64;       // Context length
-static const int MAX_ITERS = 50;     
-static const int EVAL_INTERVAL = 250; 
-static const float LEARNING_RATE = 5e-4f; 
-static const int EVAL_ITERS = 100;      
+static const int BATCH_SIZE = 16;
+static const int BLOCK_SIZE = 64; // Context length
+static const int MAX_ITERS = 50;
+static const int EVAL_INTERVAL = 250;
+static const float LEARNING_RATE = 1e-4f;
+static const int EVAL_ITERS = 100;
 static const int N_EMBD = 128;
 static const int N_HEAD = 4;
 static const int N_LAYER = 4;
-static const float DROPOUT = 0.05f; 
+static const float DROPOUT = 0.05f;
 static const std::string BEST_MODEL_PATH = "best_model.bin";
 static const std::string MODEL_PATH_ENV_VAR = "GPT_MODEL_PATH";

data/data_set.py

@@ -2,9 +2,9 @@
 import os
 
 # Settings
-target_size_mb = 30
+target_size_mb = 10_000
 target_size_bytes = target_size_mb * 1024 * 1024
-output_file = "cleaned.txt"
+output_file = "input.txt"
 
 print(f"Streaming TinyStories until {target_size_mb} MB is reached...")
 
@@ -15,14 +15,14 @@
 with open(output_file, "w", encoding="utf-8") as f:
     for entry in dataset:
         story_text = entry["text"] + "\n\n"
-        
+
         # Calculate size of this story in bytes
         story_bytes = len(story_text.encode('utf-8'))
-        
+
         if current_size + story_bytes > target_size_bytes:
             break
-            
+
         f.write(story_text)
         current_size += story_bytes
 
-print(f"Done! Created '{output_file}' ({current_size / (1024*1024):.2f} MB)")
+print(f"Done! Created '{output_file}' ({current_size / (1024*1024):.2f} MB)")