First good level of progress

.gitignore

@@ -0,0 +1,208 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+/data/books

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 [fullname]
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

catlin_chatlog.txt

@@ -0,0 +1,6 @@
+[2025-06-29T12:13:41.661917] You: Hi, what's your name?
+[2025-06-29T12:13:41.661917] Catlin: what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what
+
+[2025-06-29T12:16:56.771441] You: Hi, what is your name?
+[2025-06-29T12:16:56.771441] Catlin: what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what what
+

config.py

@@ -0,0 +1,11 @@
+# config.py
+
+VOCAB_SIZE = 50000
+CONTEXT_SIZE = 128
+EMBED_DIM = 256
+NUM_HEADS = 8
+NUM_LAYERS = 6
+BATCH_SIZE = 16
+LEARNING_RATE = 3e-4
+DEVICE = "cuda"  # fallback handled in trainer
+MAX_TOKENS = 100_000  # Used to cap input corpus size

main.py

@@ -0,0 +1,84 @@
+# main.py
+
+import os
+import torch
+from config import *
+from tokenizers.word_tokenizer import WordTokenizer
+from models.gpt import GPT
+from training.trainer import TextDataset, train
+
+
+def load_texts():
+    text = ""
+
+    # --- Books ---
+    book_dir = os.path.join("data", "books")
+    os.makedirs(book_dir, exist_ok=True)
+
+    print(f"[INFO] Scanning books from: {book_dir}")
+    for file in os.listdir(book_dir):
+        path = os.path.join(book_dir, file)
+        if file.endswith(".txt"):
+            print(f"  📚 Loading {file}")
+            try:
+                with open(path, encoding="utf-8") as f:
+                    text += f.read() + "\n"
+            except Exception as e:
+                print(f"  ❌ Failed to read {file}: {e}")
+
+    # --- OpenWebText ---
+    owt_path = os.path.join("data/openwebtext", "owt_20000.jsonl")
+    print(f"[INFO] Scanning OpenWebText: {owt_path}")
+
+    if os.path.exists(owt_path):
+        with open(owt_path, encoding="utf-8") as f:
+            for i, line in enumerate(f):
+                if i % 1000 == 0:
+                    print(f"    ⏳ {i} lines read...")
+                try:
+                    text += line.strip() + "\n"
+                except Exception as e:
+                    print(f"    ❌ Line {i} decode error: {e}")
+    else:
+        print(f"[WARN] OpenWebText file not found: {owt_path}")
+
+    # --- Chat logs ---
+    if os.path.exists("catlin_chatlog.txt"):
+        print(f"[INFO] Appending chat log...")
+        with open("catlin_chatlog.txt", encoding="utf-8") as f:
+            text += "\n" + f.read()
+
+    print(f"[INFO] Raw text loaded: {len(text)} characters")
+    return text[:MAX_TOKENS * 10]
+
+
+def main():
+    print("[INFO] Starting main()")
+    raw_text = load_texts()
+    print(f"[INFO] Loaded text: {len(raw_text)} characters")
+
+    tokenizer = WordTokenizer(VOCAB_SIZE)
+    tokenizer.fit(raw_text)
+    tokenizer.save("catlin_tokenizer.pkl")
+    print("[INFO] Tokenizer built and saved")
+
+    tokens = tokenizer.encode(raw_text)
+    print(f"[INFO] Total tokens: {len(tokens)}")
+
+    dataset = TextDataset(tokens, CONTEXT_SIZE)
+    if len(dataset) == 0:
+        print("❌ ERROR: Dataset is empty. Aborting.")
+        return
+
+    model = GPT(VOCAB_SIZE, CONTEXT_SIZE, EMBED_DIM, NUM_HEADS, NUM_LAYERS)
+    print("[INFO] Model initialized")
+
+    train(model, dataset, DEVICE if torch.cuda.is_available() else "cpu", LEARNING_RATE, BATCH_SIZE, epochs=10)
+    print("[INFO] Training complete")
+
+    torch.save(model.state_dict(), "catlin_model.pt")
+    print("[INFO] Model saved to catlin_model.pt")
+
+
+if __name__ == "__main__":
+    main()

models/gpt.py

@@ -0,0 +1,44 @@
+# models/gpt.py
+
+import torch
+import torch.nn as nn
+
+
+class GPTBlock(nn.Module):
+    def __init__(self, embed_dim, num_heads):
+        super().__init__()
+        self.attn = nn.MultiheadAttention(embed_dim, num_heads, batch_first=True)
+        self.ln1 = nn.LayerNorm(embed_dim)
+        self.ff = nn.Sequential(
+            nn.Linear(embed_dim, 4 * embed_dim),
+            nn.GELU(),
+            nn.Linear(4 * embed_dim, embed_dim)
+        )
+        self.ln2 = nn.LayerNorm(embed_dim)
+
+    def forward(self, x):
+        attn_out, _ = self.attn(x, x, x, need_weights=False)
+        x = self.ln1(x + attn_out)
+        x = self.ln2(x + self.ff(x))
+        return x
+
+
+class GPT(nn.Module):
+    def __init__(self, vocab_size, context_size, embed_dim, num_heads, num_layers, dropout=0.1):
+        super().__init__()
+        self.token_emb = nn.Embedding(vocab_size, embed_dim)
+        self.pos_emb = nn.Parameter(torch.zeros(1, context_size, embed_dim))
+        self.dropout = nn.Dropout(dropout)
+        self.blocks = nn.ModuleList([GPTBlock(embed_dim, num_heads) for _ in range(num_layers)])
+        self.ln_f = nn.LayerNorm(embed_dim)
+        self.head = nn.Linear(embed_dim, vocab_size)
+
+    def forward(self, x):
+        B, T = x.size()
+        tok_emb = self.token_emb(x)
+        x = tok_emb + self.pos_emb[:, :T, :]
+        x = self.dropout(x)
+        for block in self.blocks:
+            x = block(x)
+        x = self.ln_f(x)
+        return self.head(x)

talk_to_catlin.py

@@ -0,0 +1,95 @@
+# talk_to_catlin.py
+
+import os
+import torch
+from config import *
+from tokenizers.word_tokenizer import WordTokenizer
+from models.gpt import GPT
+from datetime import datetime
+
+CHAT_LOG_PATH = "catlin_chatlog.txt"
+
+
+def load_model(device):
+    model = GPT(VOCAB_SIZE, CONTEXT_SIZE, EMBED_DIM, NUM_HEADS, NUM_LAYERS)
+    model.load_state_dict(torch.load("catlin_model.pt", map_location=device))
+    model.to(device)
+    model.eval()
+    return model
+
+
+def load_tokenizer(path="catlin_tokenizer.pkl"):
+    if not os.path.exists(path):
+        print(f"[ERROR] Tokenizer file '{path}' not found. Please train first.")
+        exit(1)
+    return WordTokenizer.load(path)
+
+
+def top_p_sampling(logits, p=0.9):
+    sorted_logits, sorted_indices = torch.sort(logits, descending=True)
+    cumulative_probs = torch.cumsum(torch.nn.functional.softmax(sorted_logits, dim=-1), dim=-1)
+    sorted_indices_to_keep = cumulative_probs <= p
+    sorted_indices_to_keep[1:] = sorted_indices_to_keep[:-1].clone()
+    sorted_indices_to_keep[0] = True
+
+    filtered_logits = sorted_logits[sorted_indices_to_keep]
+    filtered_indices = sorted_indices[sorted_indices_to_keep]
+
+    probs = torch.nn.functional.softmax(filtered_logits, dim=-1)
+    next_token = filtered_indices[torch.multinomial(probs, 1)]
+    return next_token.item()
+
+
+def generate_response(model, tokenizer, prompt, device, max_tokens=50, k=10, temperature=1.0):
+    tokens = tokenizer.encode(prompt)[-CONTEXT_SIZE:]
+    input_ids = torch.tensor([tokens], dtype=torch.long).to(device)
+
+    for _ in range(max_tokens):
+        with torch.no_grad():
+            logits = model(input_ids)
+        next_token_logits = logits[0, -1, :]
+        next_token_logits = next_token_logits / temperature
+        next_token = top_p_sampling(next_token_logits, p=0.9)
+
+        input_ids = torch.cat([input_ids, torch.tensor([[next_token]], device=device)], dim=1)
+        if input_ids.shape[1] > CONTEXT_SIZE:
+            input_ids = input_ids[:, -CONTEXT_SIZE:]
+
+        word = tokenizer.id_to_word.get(next_token, "")
+        if word in {".", "!", "?"}:
+            break
+
+    return tokenizer.decode(input_ids[0].tolist()[len(tokens):])
+
+
+def log_chat(user_input, catlin_response):
+    with open(CHAT_LOG_PATH, "a", encoding="utf-8") as f:
+        f.write(f"[{datetime.now().isoformat()}] You: {user_input}\n")
+        f.write(f"[{datetime.now().isoformat()}] Catlin: {catlin_response}\n\n")
+
+
+def main():
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    model = load_model(device)
+    tokenizer = load_tokenizer()
+
+    print("🤖 Catlin is online. Type 'exit' to end the conversation.\n")
+    history = []
+
+    while True:
+        user_input = input("You: ")
+        if user_input.strip().lower() == "exit":
+            break
+
+        history.append(f"You: {user_input}")
+        # Build memory window
+        memory_text = " ".join(history[-10:])  # Keep last 10 lines
+        catlin_response = generate_response(model, tokenizer, memory_text, device)
+
+        print(f"Catlin: {catlin_response}\n")
+        history.append(f"Catlin: {catlin_response}")
+        log_chat(user_input, catlin_response)
+
+
+if __name__ == "__main__":
+    main()

tokenizers/word_tokenizer.py

@@ -0,0 +1,42 @@
+# tokenizers/word_tokenizer.py
+
+import re
+from collections import Counter
+import pickle
+
+
+class WordTokenizer:
+    def __init__(self, vocab_size=50000):
+        self.vocab_size = vocab_size
+        self.word_to_id = {"<PAD>": 0, "<UNK>": 1}
+        self.id_to_word = {0: "<PAD>", 1: "<UNK>"}
+
+    def fit(self, texts):
+        words = re.findall(r"\b\w+\b", texts.lower())
+        freq = Counter(words).most_common(self.vocab_size - 2)
+        for idx, (word, _) in enumerate(freq, start=2):
+            self.word_to_id[word] = idx
+            self.id_to_word[idx] = word
+
+    def encode(self, text):
+        return [self.word_to_id.get(word, 1) for word in re.findall(r"\b\w+\b", text.lower())]
+
+    def decode(self, tokens):
+        return " ".join([self.id_to_word.get(token, "<UNK>") for token in tokens])
+
+    def save(self, path):
+        with open(path, "wb") as f:
+            pickle.dump({
+                "vocab_size": self.vocab_size,
+                "word_to_id": self.word_to_id,
+                "id_to_word": self.id_to_word
+            }, f)
+
+    @classmethod
+    def load(cls, path):
+        with open(path, "rb") as f:
+            data = pickle.load(f)
+        obj = cls(data["vocab_size"])
+        obj.word_to_id = data["word_to_id"]
+        obj.id_to_word = data["id_to_word"]
+        return obj

tools/book_downloader.py

@@ -0,0 +1,38 @@
+# tools/book_downloader.py
+
+import requests
+import os
+
+DATA_DIR = os.path.join("data", "books")
+GUTENBERG_URL = "https://www.gutenberg.org/files/{id}/{id}-0.txt"
+
+
+def download_book(gutenberg_id, title_hint="book"):
+    os.makedirs(DATA_DIR, exist_ok=True)
+    url = GUTENBERG_URL.format(id=gutenberg_id)
+    try:
+        response = requests.get(url, timeout=10)
+        if response.status_code != 200:
+            print(f"❌ Failed to download book ID {gutenberg_id}")
+            return
+
+        filename = os.path.join(DATA_DIR, f"{title_hint}_{gutenberg_id}.txt")
+        with open(filename, "w", encoding="utf-8") as f:
+            f.write(response.text)
+
+        print(f"✅ Saved: {filename}")
+    except Exception as e:
+        print(f"❌ Error: {e}")
+
+
+if __name__ == "__main__":
+    books = [
+        (1342, "PrideAndPrejudice"),  # Jane Austen
+        (11, "AliceInWonderland"),    # Lewis Carroll
+        (98, "AesopFables"),          # Aesop
+        (1661, "SherlockHolmes"),     # Doyle
+        (76, "HuckFinn")              # Mark Twain
+    ]
+
+    for gutenberg_id, name in books:
+        download_book(gutenberg_id, name)

tools/openwebtext_fetcher.py

@@ -0,0 +1,21 @@
+# tools/openwebtext_fetcher.py
+
+from datasets import load_dataset
+import os
+from tqdm import tqdm
+
+TARGET_DIR = "data/openwebtext"
+os.makedirs(TARGET_DIR, exist_ok=True)
+
+
+def fetch_subset(n=10000, split="train"):
+    ds = load_dataset("stas/openwebtext-10k", split=split)
+    with open(os.path.join(TARGET_DIR, f"owt_{n}.jsonl"), "w", encoding="utf-8") as f:
+        for i, item in tqdm(enumerate(ds), total=n, desc="Writing JSONL"):
+            f.write(f"{item['text'].replace(chr(10),' ')}\n")
+            if i + 1 >= n:
+                break
+
+
+if __name__ == "__main__":
+    fetch_subset(20000)  # fetch 20K examples (~100 MB)

training/trainer.py

@@ -0,0 +1,58 @@
+# training/trainer.py
+
+import torch
+import torch.nn as nn
+import torch.optim as optim
+from torch.utils.data import Dataset, DataLoader
+from tqdm import tqdm
+
+
+class TextDataset(Dataset):
+    def __init__(self, tokens, context_size):
+        self.tokens = tokens
+        self.context_size = context_size
+
+    def __len__(self):
+        return len(self.tokens) - self.context_size
+
+    def __getitem__(self, idx):
+        x = torch.tensor(self.tokens[idx:idx+self.context_size], dtype=torch.long)
+        y = torch.tensor(self.tokens[idx+1:idx+self.context_size+1], dtype=torch.long)
+        return x, y
+
+
+def train(model, dataset, device, lr, batch_size, epochs=1):
+    loader = DataLoader(dataset, batch_size=batch_size, shuffle=True)
+    optimizer = optim.Adam(model.parameters(), lr=lr)
+    loss_fn = nn.CrossEntropyLoss()
+
+    model.to(device)
+    model.train()
+    torch.autograd.set_detect_anomaly(True)
+
+    for epoch in range(epochs):
+        total_loss = 0.0
+        progress = tqdm(loader, desc=f"Epoch {epoch+1}/{epochs}", unit="batch", dynamic_ncols=True)
+
+        if len(loader) == 0:
+            print("❌ No data to train on. Check your token count or dataset.")
+            return
+
+        for x, y in progress:
+            x, y = x.to(device), y.to(device)
+            logits = model(x)
+            loss = loss_fn(logits.view(-1, logits.size(-1)), y.view(-1))
+
+            if torch.isnan(loss):
+                print("❌ Loss is NaN! Aborting training.")
+                exit(1)
+
+            optimizer.zero_grad()
+            loss.backward()
+            torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)
+            optimizer.step()
+
+            total_loss += loss.item()
+            progress.set_postfix(loss=loss.item())
+
+        print(f"[Epoch {epoch+1}] Avg Loss: {total_loss / len(loader):.4f}")