# /// script # requires-python = ">=3.10,<3.14" # dependencies = [ # "bitsandbytes>=0.43.0", # "marimo", # "timm", # "tokenizers>=0.22.0,<=0.23.0", # "torch>=2.8.0", # "torchvision", # "transformers>=5.5.0", # "triton>=3.2.0", # "triton_kernels @ git+https://github.com/triton-lang/triton.git@0add68262ab0a2e33b84524346cb27cbb2787356#subdirectory=python/triton_kernels", # "trl>=0.28.0", # "unsloth @ git+https://github.com/unslothai/unsloth", # "unsloth_zoo @ git+https://github.com/unslothai/unsloth-zoo", # "uv", # ] # # [tool.uv] # no-build-package = [ # "bitsandbytes", # "triton", # "vllm", # "xformers", # ] # /// import marimo __generated_with = "0.23.8" app = marimo.App() @app.cell def _(): import marimo as mo return (mo,) @app.cell(hide_code=True) def _(mo): mo.md(r""" To run this notebook, hit the **▶ Run all** button in the bottom-right corner - or use `Ctrl/Cmd + Shift + R`.

Join Discord if you need help + ⭐ Star us on Github ⭐

To install Unsloth on your local device, follow [our guide](https://unsloth.ai/docs/get-started/install). This notebook is licensed [LGPL-3.0](https://github.com/unslothai/notebooks?tab=LGPL-3.0-1-ov-file#readme). You will learn how to do [data prep](#Data), how to [train](#Train), how to [run the model](#Inference), & how to save it """) return @app.cell(hide_code=True) def _(mo): mo.md(r""" # Goal: Make Gemma 4 play games with Reinforcement Learning Our goal is to make Gemma 4 play the 2048 game with reinforcement learning, or a variant of it called [GRPO](https://arxiv.org/abs/2501.12948). We want the model to devise a strategy to play 2048, and we will run this strategy until we win or lose. We then reward the model if it created a good strategy (winning the game), and we'll penalize it (negative reward) if the strategy was a bad one.

""") return @app.cell(hide_code=True) def _(mo): mo.md(r""" # Installation We'll be using [Unsloth](https://github.com/unslothai/unsloth) to do RL on Gemma 4. Unsloth saves 70% VRAM usage and makes reinforcement learning 2 to 6x faster! """) return @app.cell def _(): # For Gemma 4 vision/audio return @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Unsloth """) return @app.cell def _(): from unsloth import FastVisionModel import torch max_seq_length = 4096 # Can increase for longer reasoning traces lora_rank = 32 # Larger rank = smarter, but slower gemma4_models = [ # Gemma-4 instruct models: "unsloth/gemma-4-E2B-it", "unsloth/gemma-4-E4B-it", "unsloth/gemma-4-31B-it", "unsloth/gemma-4-26B-A4B-it", # Gemma-4 base models: "unsloth/gemma-4-E2B", "unsloth/gemma-4-E4B", "unsloth/gemma-4-31B", "unsloth/gemma-4-26B-A4B", ] # More models at https://huggingface.co/unsloth model, tokenizer = FastVisionModel.from_pretrained( model_name="unsloth/gemma-4-E2B-it", max_seq_length=max_seq_length, # Can increase for longer reasoning traces load_in_4bit=False, # False for LoRA 16bit fast_inference=False, # Enable vllm fast inference ) return FastVisionModel, lora_rank, max_seq_length, model, tokenizer @app.cell(hide_code=True) def _(mo): mo.md(r""" To do efficient RL, we will use [LoRA](https://arxiv.org/abs/2106.09685), which allows us to only add 1 to 5% of extra weights to the model for finetuning purposes. This allows us to save memory usage by over 60%, and yet it retains good accuracy. """) return @app.cell def _(FastVisionModel, lora_rank, model): model_1 = FastVisionModel.get_peft_model( model, r=lora_rank, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128 target_modules=[ "q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj", ], lora_alpha=lora_rank * 2, # *2 speeds up training use_gradient_checkpointing="unsloth", # Reduces memory usage random_state=3407, ) return (model_1,) @app.cell(hide_code=True) def _(mo): mo.md(r""" # 2048 game We used GPT-5 to create a variant of the 2048 game. It should output the current game board state, and allow us to advance the game board state with 1 action (up, down, left, right). """) return @app.cell def _(): # @title (Collapsible) 2048 Game Implementation from dataclasses import dataclass, field from typing import List, Tuple, Optional import random import copy def _compress_and_merge_row_left(row: List[int]) -> Tuple[List[int], int, bool]: n = len(row) tiles = [x for x in row if x != 0] gained = 0 i = 0 merged = [] while i < len(tiles): if i + 1 < len(tiles) and tiles[i] == tiles[i + 1]: v = tiles[i] * 2 gained = gained + v merged.append(v) i = i + 2 else: merged.append(tiles[i]) i = i + 1 merged = merged + [0] * (n - len(merged)) changed = merged != row return (merged, gained, changed) def _move_left(board: List[List[int]]) -> Tuple[List[List[int]], int, bool]: changed_any = False total_gain = 0 new_board = [] for row in board: new_row, gained, changed = _compress_and_merge_row_left(row) new_board.append(new_row) total_gain = total_gain + gained changed_any = changed_any or changed return (new_board, total_gain, changed_any) def _move_right(board: List[List[int]]) -> Tuple[List[List[int]], int, bool]: changed_any = False total_gain = 0 new_board = [] for row in board: rev = list(reversed(row)) new_rev, gained, changed = _compress_and_merge_row_left(rev) new_row = list(reversed(new_rev)) new_board.append(new_row) total_gain = total_gain + gained changed_any = changed_any or changed return (new_board, total_gain, changed_any) def _transpose(board: List[List[int]]) -> List[List[int]]: return [list(row) for row in zip(*board)] def _move_up(board: List[List[int]]) -> Tuple[List[List[int]], int, bool]: t = _transpose(board) # safety; avoid log2(1) moved, gain, changed = _move_left(t) return (_transpose(moved), gain, changed) def _move_down(board: List[List[int]]) -> Tuple[List[List[int]], int, bool]: t = _transpose(board) # safety; avoid log2(1) moved, gain, changed = _move_right(t) return (_transpose(moved), gain, changed) def _empty_cells(board: List[List[int]]) -> List[Tuple[int, int]]: size = len(board) return [(r, c) for r in range(size) for c in range(size) if board[r][c] == 0] def _can_move(board: List[List[int]]) -> bool: if _empty_cells(board): return True size = len(board) for r in range(size): for c in range(size - 1): if board[r][c] == board[r][c + 1]: return True for r in range(size - 1): for c in range(size): if board[r][c] == board[r + 1][c]: return True return False @dataclass class GameBoard: size: int seed: Optional[int] = None target: int = 2048 probability_fours: float = 0.1 # originally spawns (4) 10% of the time! _rng: random.Random = field(init=False, repr=False) _board: List[List[int]] = field(init=False, repr=False) _score: int = field(default=0, init=False, repr=False) _state: str = field(default="ongoing", init=False, repr=False) def __post_init__(self): if self.size < 2: raise ValueError("Board size must be at least 2.") self._rng = random.Random(self.seed) self._board = [[0 for _ in range(self.size)] for _ in range(self.size)] self._add_random_tile() self._add_random_tile() self._update_state_after_change() class _BoardView: def __init__(self, game: "GameBoard"): self._game = game def __iter__(self): return iter(self._game._board) def __len__(self): return len(self._game._board) def __getitem__(self, idx): return self._game._board[idx] def __repr__(self) -> str: return repr(self._game._board) __str__ = __repr__ def do_action(self, key: str) -> None: self._game.do_action(key) def state(self) -> str: return self._game.state() def pretty( self, colors: bool = True, border: bool = True, dot_for_zero: bool = True, ) -> str: return self._game._render_pretty( colors=colors, border=border, dot_for_zero=dot_for_zero ) def board(self) -> "_BoardView": return GameBoard._BoardView(self) def state(self) -> str: return self._state def score(self) -> int: return self._score def do_action(self, key: str) -> None: if self._state != "ongoing": return if not isinstance(key, str) or len(key) == 0: self._state = "failed" return k = key.strip().lower() if k == "q": self._state = "failed" return move_map = { "a": _move_left, "d": _move_right, "w": _move_up, "s": _move_down, } if k not in move_map: self._state = "failed" return mover = move_map[k] new_board, gain, changed = mover(self._board) if changed: self._board = new_board self._score = self._score + gain self._add_random_tile() self._update_state_after_change() def _add_random_tile(self) -> bool: empties = _empty_cells(self._board) if not empties: return False r, c = self._rng.choice(empties) self._board[r][c] = 4 if self._rng.random() < self.probability_fours else 2 return True def _update_state_after_change(self) -> None: if any((self.target in row for row in self._board)): self._state = "success" return if not _can_move(self._board): self._state = "failed" return self._state = "ongoing" # A smooth-ish gradient from cool → warm def _render_pretty( self, colors: bool = True, border: bool = True, dot_for_zero: bool = True ) -> str: # (blue/cyan/green → yellow/orange/red). Tweak or expand as you like. """ Pretty-print the board with colors that scale from 0 up to self.target. # dim gray Uses ANSI 256-color codes (works in most terminals). Set colors = False to disable. """ import math b = self._board mx = max((max(row) for row in b), default=0) cell_w = max(3, len(str(mx))) RESET = "\x1b[0m" # Normalize by exponent relative to target: r in [0,1] GRAD = [ 33, 39, 45, 51, 50, 49, 48, 47, 46, 82, 118, 154, 190, 226, 220, 214, 208, 202, 196, ] # safety; avoid log2(1) ZERO_FG = ( # dim gray 239 # Guard: if v is not a power of two or is <1, handle gracefully ) def color_code(v: int) -> str: if not colors: return "" if v == 0: return f"\x1b[38;5;{ZERO_FG}m" t = max(2, self.target) # safety; avoid log2(1) try: r = max(0.0, min(1.0, math.log2(v) / math.log2(t))) # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128 except ValueError: r = 0.0 # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128 idx = int(round(r * (len(GRAD) - 1))) return f"\x1b[38;5;{GRAD[idx]}m" def fmt(v: int) -> str: s = "." if v == 0 and dot_for_zero else str(v) s = s.rjust(cell_w) return color_code(v) + s + (RESET if colors else "") def hline(left: str, mid: str, right: str) -> str: return left + mid.join(("─" * cell_w for _ in range(self.size))) + right rows = [] if border: rows.append(hline("┌", "┬", "┐")) for r in range(self.size): content = "│".join((fmt(v) for v in b[r])) rows.append("│" + content + "│" if border else content) if border: rows.append( hline( "└" if r == self.size - 1 else "├", "┴" if r == self.size - 1 else "┼", "┘" if r == self.size - 1 else "┤", ) ) return "\n".join(rows) return (GameBoard,) @app.cell(hide_code=True) def _(mo): mo.md(r""" For example let's create a board of size 5 X 5 and set the target to 8 instead of 2048. **[NOTE]** 2048 originally spawns a (4) 10% of the time! We can disable this for harder games. See [Wikipedia page](https://en.wikipedia.org/wiki/2048_(video_game)) for more details. """) return @app.cell def _(GameBoard): game = GameBoard(size=5, seed=42, target=8, probability_fours=0.10) print(game.board().pretty(), game.state()) return (game,) @app.cell def _(game): game return @app.cell(hide_code=True) def _(mo): mo.md(r""" We'll use WASD for the action space: ``` W A S D ``` Also `game.state()` will say `success` if we succeeded in getting the target! """) return @app.cell def _(game): game.do_action("A") print(game.board().pretty(), game.state()) return @app.cell def _(game): game.do_action("W") print(game.board().pretty(), game.state()) return @app.cell def _(game): game.do_action("D") print(game.board().pretty(), game.state()) return @app.cell def _(game): game.do_action("W") print(game.board().pretty(), game.state()) return @app.cell def _(game): game.do_action("D") print(game.board().pretty(), game.state()) return @app.cell(hide_code=True) def _(mo): mo.md(r""" If we do some other action that's not part of the action space, we will get an error, and the game will not accept anymore actions. """) return @app.cell def _(GameBoard): game_1 = GameBoard(size=3, seed=42, target=8, probability_fours=0.1) game_1.do_action("AA") # Not in WASD game_1.do_action("W") # Doesn't do anything game_1.do_action("A") # Doesn't do anything print(game_1.board().pretty(), game_1.state()) return @app.cell(hide_code=True) def _(mo): mo.md(r""" # RL Environment Setup We'll set up a function to accept some strategy that'll emit an action within `WASD` and check the game state. We'll also add a timer to only execute the strategy for 2 seconds maximum, otherwise it might never terminate! """) return @app.cell def _(GameBoard): from typing import Callable from unsloth import execute_with_time_limit def _execute_strategy(strategy: Callable, game: GameBoard): assert callable(strategy) steps = 0 while game.state() == "ongoing": action = strategy(list(game.board())) steps = steps + 1 if type(action) is not str: return (steps, "failed") game.do_action(action) return (steps, game.state()) @execute_with_time_limit(2) def execute_strategy(strategy: Callable, game: GameBoard): return _execute_strategy(strategy, game) return Callable, execute_strategy, execute_with_time_limit @app.cell(hide_code=True) def _(mo): mo.md(r""" Let's make a generic strategy to just hit `W`. We should expect this generic strategy to fail: """) return @app.cell def _(GameBoard, execute_strategy): def always_move_left(board): return "W" game_2 = GameBoard(size=8, seed=42, target=2048, probability_fours=0.1) try: execute_strategy(always_move_left, game_2) except TimeoutError as e: print(f"Timed out with error = {str(e)}") return @app.cell(hide_code=True) def _(mo): mo.md(r""" To allow longer strategies for Gemma 4 Reinforcement Learning, we shall allow a 5 second timer. """) return @app.cell def _(Callable, GameBoard, execute_with_time_limit): @execute_with_time_limit(5) def execute_strategy_1(strategy: Callable, game: GameBoard): return _execute_strategy(strategy, game) return (execute_strategy_1,) @app.cell(hide_code=True) def _(mo): mo.md(r""" # Code Execution To execute and create a new Python function, we first have to check if the function does not call other global variables or cheat. This is called `countering reward hacking` since we don't want the function to cheat. For example the below piece of code is fine, since it only imports Python level functions. We use `check_python_modules`: """) return @app.cell def _(): from unsloth import check_python_modules sample = """ def strategy(board): import math from typing import Callable return "W" """ ok, info = check_python_modules(sample) print("Only Python imports?", ok) print(info) return (check_python_modules,) @app.cell(hide_code=True) def _(mo): mo.md(r""" For the below piece of code, since we import `numpy`, we should not allow the execution: """) return @app.cell def _(check_python_modules): sample_1 = '\ndef strategy(board):\n from numpy import matmul\n return "W"\n' ok_1, info_1 = check_python_modules(sample_1) print("Only Python imports?", ok_1) print(info_1) return @app.cell(hide_code=True) def _(mo): mo.md(r""" We also disallow global variable access. We'll use Unsloth's `create_locked_down_function` function """) return @app.cell def _(): from unsloth import create_locked_down_function function = """ def import_numpy(): np.matmul print("Success") """ f = create_locked_down_function(function) try: f() except Exception as e: print(str(e)) return (create_locked_down_function,) @app.cell def _(create_locked_down_function): function_1 = "\ndef add(a, b):\n def adder(a):\n return a + b\n return adder(b) + b\n" f_1 = create_locked_down_function(function_1) try: print(f_1(10, 20)) except Exception as e: print(str(e)) return @app.cell(hide_code=True) def _(mo): mo.md(r""" # Data & RL task setup We now have to create a prompt to tell the model to create a strategy for the 2048 game. You can customize this to some other task for another RL task. """) return @app.cell def _(): prompt = """ Create a new short 2048 strategy using only native Python code. You are given a list of list of numbers for the current board state. Output one action for "W", "A", "S", "D" on what is the optimal next step. Output your new short function in backticks using the format below: ```python def strategy(board): return "W" # Example ``` All helper functions should be inside def strategy. Only output the short function `strategy`. """.strip() print(prompt) return (prompt,) @app.cell(hide_code=True) def _(mo): mo.md(r""" First, let's prompt Gemma 4 without RL and see how it goes: """) return @app.cell def _(model_1, prompt, tokenizer): text = tokenizer.apply_chat_template( [{"role": "user", "content": prompt.strip()}], tokenize=False, add_generation_prompt=True, ) from transformers import TextStreamer print("=" * 50) print("BASE MODEL OUTPUT (before RL training):") print("=" * 50) inputs = tokenizer(text=text, add_special_tokens=False, return_tensors="pt").to( "cuda" ) text_streamer = TextStreamer(tokenizer, skip_prompt=True) result = model_1.generate( **inputs, streamer=text_streamer, max_new_tokens=512, use_cache=True, temperature=1.0, top_p=0.95, top_k=64, ) return (TextStreamer,) @app.cell(hide_code=True) def _(mo): mo.md(r""" # Reward functions We now design a `extract_function` function which simply extracts the function wrapped in 3 back ticks. And 3 reward functions: 1. `function_works` which rewards the model if the strategy is a valid Python function. 2. `no_cheating` which checks if the function imported other modules, and if it did, we penalize it. 3. `strategy_succeeds` which checks if the game strategy actually succeeds in attaining 2048 after running the auto-generated strategy. """) return @app.cell def _(prompt): def extract_function(text): if text.count("```") >= 2: first = text.find("```") + 3 second = text.find("```", first) fx = text[first:second].strip() fx = fx.removeprefix("python\n") fx = fx[fx.find("def") :] if fx.startswith("def strategy(board):"): return fx return None print(extract_function(prompt)) return (extract_function,) @app.cell(hide_code=True) def _(mo): mo.md(r""" Below is our `function_works` reward function which uses Python's `exec` but guarded by not allowing leakage of local and global variables. We can also use `check_python_modules` first to check if there are errors before even executing the function: """) return @app.cell def _(check_python_modules): ok_2, info_2 = check_python_modules("def a") (ok_2, info_2) return @app.cell def _(check_python_modules, create_locked_down_function, extract_function): def function_works(completions, **kwargs): scores = [] for completion in completions: score = 0 response = completion[0]["content"] function = extract_function(response) if function is not None: ok, info = check_python_modules(function) if function is None or "error" in info: score = -2.0 else: try: new_strategy = create_locked_down_function(function) score = 1.0 except: score = -0.5 scores.append(score) return scores return (function_works,) @app.cell(hide_code=True) def _(mo): mo.md(r""" `no_cheating` checks if the function cheated since it might have imported Numpy or other functions: """) return @app.cell def _(check_python_modules, extract_function): def no_cheating(completions, **kwargs): scores = [] for completion in completions: score = 0 response = completion[0]["content"] function = extract_function(response) if function is not None: ok, info = check_python_modules(function) scores.append(1.0 if ok else -20.0) # Penalize heavily! else: scores.append(-1.0) # Failed creating function return scores return (no_cheating,) @app.cell(hide_code=True) def _(mo): mo.md(r""" Next `strategy_succeeds` checks if the strategy actually allows the game to terminate. Imagine if the strategy simply returned "W" which would fail after a time limit of 10 seconds. We also add a global `PRINTER` to print out the strategy and board state. """) return @app.cell def _( GameBoard, check_python_modules, create_locked_down_function, execute_strategy_1, extract_function, ): import numpy as np global PRINTER PRINTER = 0 def strategy_succeeds(completions, **kwargs): global PRINTER scores = [] # Generate a random game board with seed seed = np.random.randint(10000) for completion in completions: printed = False score = 0 response = completion[0]["content"] function = extract_function(response) if PRINTER % 5 == 0: printed = True print(function) PRINTER = PRINTER + 1 if function is not None: ok, info = check_python_modules(function) if function is None or "error" in info: scores.append(0) continue try: new_strategy = create_locked_down_function(function) except: scores.append(0) continue try: game = GameBoard(size=6, seed=seed, target=2048, probability_fours=0.1) steps, game_state = execute_strategy_1(new_strategy, game) print(f"Steps = {steps} State = {game_state}") if printed is False: print(function) print(game.board().pretty()) if game_state == "success": scores.append(20.0) # Success - massively reward! else: scores.append(2.0) # Failed but function works! except TimeoutError as e: print("Timeout") scores.append(-1.0) # Failed with timeout except Exception as e: print(f"Exception = {str(e)}") scores.append(-3.0) # Failed return scores return PRINTER, strategy_succeeds @app.cell(hide_code=True) def _(mo): mo.md(r""" We'll now create the dataset which includes a replica of our prompt. """) return @app.cell def _(prompt, tokenizer): from datasets import Dataset dataset = Dataset.from_list( [{"prompt": [{"role": "user", "content": prompt.strip()}], "answer": 0}] * 1000 ) maximum_length = len( tokenizer.apply_chat_template( [{"role": "user", "content": prompt.strip()}], add_generation_prompt=True, tokenize=True, ) ) print(maximum_length) dataset[0] return dataset, maximum_length @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Train the model Now set up GRPO Trainer and all configurations! We also support GSPO, GAPO, Dr GRPO and more! Go the Unsloth [Reinforcement Learning Docs](https://unsloth.ai/docs/get-started/reinforcement-learning-rl-guide) for more options. """) return @app.cell def _(max_seq_length, maximum_length): # Leave room for the prompt (plus 1 token safety margin) max_completion_length = max_seq_length - (maximum_length + 1) from trl import GRPOConfig, GRPOTrainer training_args = GRPOConfig( temperature=1.0, top_p=0.95, top_k=64, learning_rate=5e-5, weight_decay=0.001, warmup_ratio=0.1, lr_scheduler_type="linear", optim="adamw_8bit", logging_steps=1, per_device_train_batch_size=1, gradient_accumulation_steps=2, # Increase to 4 for smoother training num_generations=2, # Decrease if out of memory max_completion_length=max_completion_length, # num_train_epochs = 1, # Set to 1 for a full training run max_steps=60, save_steps=100, report_to="none", # Can use Weights & Biases, TrackIO output_dir="outputs", epsilon=0.2, epsilon_high=0.28, # one sided delta=1.5, # two sided loss_type="bnpo", mask_truncated_completions=True, # For optional training + evaluation # fp16_full_eval = True, # per_device_eval_batch_size = 4, # eval_accumulation_steps = 1, # eval_strategy = "steps", # eval_steps = 1, ) return GRPOTrainer, training_args @app.cell(hide_code=True) def _(mo): mo.md(r""" And let's run the trainer! If you scroll up, you'll see a table of rewards. The goal is to see the `reward` column increase! You might have to wait 150 to 200 steps for any action. You'll probably get 0 reward for the first 100 steps. Please be patient! | Step | Training Loss | reward | reward_std | completion_length | kl | |------|---------------|-----------|------------|-------------------|----------| | 1 | 0.000000 | 0.125000 | 0.000000 | 200.000000 | 0.000000 | | 2 | 0.000000 | 0.072375 | 0.248112 | 200.000000 | 0.000000 | | 3 | 0.000000 | -0.079000 | 0.163776 | 182.500000 | 0.000005 | """) return @app.cell def _( GRPOTrainer, dataset, function_works, model_1, no_cheating, strategy_succeeds, tokenizer, training_args, ): # For optional training + evaluation # new_dataset = dataset.train_test_split(test_size = 0.01) trainer = GRPOTrainer( model=model_1, processing_class=tokenizer, reward_funcs=[function_works, no_cheating, strategy_succeeds], args=training_args, train_dataset=dataset, ) return (trainer,) @app.cell(hide_code=True) def _(mo): mo.md(r""" And let's train the model! **NOTE** A T4 free GPU might take 5 minutes for one generation sadly since it's an old GPU - A100 or H100 will be much faster! """) return @app.cell def _(trainer): trainer.train() return @app.cell(hide_code=True) def _(mo): mo.md(r""" And now with the LoRA we just trained with GRPO - we first save the LoRA first! """) return @app.cell def _(model_1, tokenizer): model_1.save_pretrained("gemma_4_lora") # Local saving tokenizer.save_pretrained("gemma_4_lora") return @app.cell(hide_code=True) def _(mo): mo.md(r""" Verify LoRA is actually trained! """) return @app.cell def _(): from safetensors import safe_open tensors = {} with safe_open("grpo_saved_lora/adapter_model.safetensors", framework="pt") as f_2: for key in f_2.keys(): tensor = f_2.get_tensor(key) # Verify both A and B are non zero n_zeros = (tensor == 0).sum() / tensor.numel() assert n_zeros.item() != tensor.numel() return @app.cell(hide_code=True) def _(mo): mo.md(r""" # Inference Now let's try the model we just trained! """) return @app.cell def _(TextStreamer, model_1, prompt, tokenizer): text_1 = tokenizer.apply_chat_template( [{"role": "user", "content": prompt.strip()}], tokenize=False, add_generation_prompt=True, ) _ = model_1.generate( **tokenizer(images=None, text=text_1, return_tensors="pt").to("cuda"), temperature=1.0, top_p=0.95, top_k=64, max_new_tokens=1024, streamer=TextStreamer(tokenizer, skip_prompt=False), ) return @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Saving to float16 for VLLM We also support saving to `float16` directly. Select `merged_16bit` for float16 or `merged_4bit` for int4. We also allow `lora` adapters as a fallback. Use `push_to_hub_merged` to upload to your Hugging Face account! You can go to https://huggingface.co/settings/tokens for your personal tokens. See [our docs](https://unsloth.ai/docs/basics/inference-and-deployment) for more deployment options. """) return @app.cell def _(model_1, tokenizer): # Merge to 16bit if False: model_1.save_pretrained_merged( "gemma_4_finetune_16bit", tokenizer, save_method="merged_16bit" ) if False: # Merge to 4bit model_1.push_to_hub_merged( "HF_USERNAME/gemma_4_finetune_16bit", tokenizer, save_method="merged_16bit", token="YOUR_HF_TOKEN", ) if False: model_1.save_pretrained_merged( "gemma_4_finetune_4bit", tokenizer, save_method="merged_4bit" ) # Just LoRA adapters if False: model_1.push_to_hub_merged( "HF_USERNAME/gemma_4_finetune_4bit", tokenizer, save_method="merged_4bit", token="YOUR_HF_TOKEN", ) if False: model_1.save_pretrained("gemma_4_lora") tokenizer.save_pretrained("gemma_4_lora") if False: model_1.push_to_hub("HF_USERNAME/gemma_4_lora", token="YOUR_HF_TOKEN") tokenizer.push_to_hub("HF_USERNAME/gemma_4_lora", token="YOUR_HF_TOKEN") return @app.cell(hide_code=True) def _(mo): mo.md(r""" ### GGUF / llama.cpp Conversion To save to `GGUF` / `llama.cpp`, we support it natively now! We clone `llama.cpp` and we default save it to `q8_0`. We allow all methods like `q4_k_m`. Use `save_pretrained_gguf` for local saving and `push_to_hub_gguf` for uploading to HF. Some supported quant methods (full list on our [docs page](https://unsloth.ai/docs/basics/inference-and-deployment/saving-to-gguf)): * `q8_0` - Fast conversion. High resource use, but generally acceptable. * `q4_k_m` - Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q4_K. * `q5_k_m` - Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q5_K. [**NEW**] To finetune and auto export to Ollama, try our [Ollama notebook](https://github.com/unslothai/notebooks/blob/main/nb/Llama3_(8B)-Ollama.ipynb) """) return @app.cell def _(model_1, tokenizer): # Save to 8bit Q8_0 if False: # Remember to go to https://huggingface.co/settings/tokens for a token! # And change hf to your username! model_1.save_pretrained_gguf("gemma_4_finetune", tokenizer) if False: # Save to 16bit GGUF model_1.push_to_hub_gguf( "HF_USERNAME/gemma_4_finetune", tokenizer, token="YOUR_HF_TOKEN" ) if False: model_1.save_pretrained_gguf( "gemma_4_finetune", tokenizer, quantization_method="f16" ) # Save to q4_k_m GGUF if False: model_1.push_to_hub_gguf( "HF_USERNAME/gemma_4_finetune", tokenizer, quantization_method="f16", token="YOUR_HF_TOKEN", ) if False: # Save to multiple GGUF options - much faster if you want multiple! model_1.save_pretrained_gguf( "gemma_4_finetune", tokenizer, quantization_method="q4_k_m" ) if False: model_1.push_to_hub_gguf( "HF_USERNAME/gemma_4_finetune", tokenizer, quantization_method="q4_k_m", token="YOUR_HF_TOKEN", ) # Change hf to your username! if False: model_1.push_to_hub_gguf( "HF_USERNAME/gemma_4_finetune", tokenizer, quantization_method=["q4_k_m", "q8_0", "q5_k_m"], token="YOUR_HF_TOKEN", ) return @app.cell(hide_code=True) def _(mo): mo.md(r""" Now, use the `gemma_4_finetune.Q8_0.gguf` file or `gemma_4_finetune.Q4_K_M.gguf` file in llama.cpp. And we're done! If you have any questions on Unsloth, we have a [Discord](https://discord.gg/unsloth) channel! If you find any bugs or want to keep updated with the latest LLM stuff, or need help, join projects etc, feel free to join our Discord! Some other resources: 1. Train your own reasoning model - Llama GRPO notebook [Open in molab](https://github.com/unslothai/notebooks/blob/main/nb/Llama3.1_(8B)-GRPO.ipynb) 2. Saving finetunes to Ollama. [Free notebook](https://github.com/unslothai/notebooks/blob/main/nb/Llama3_(8B)-Ollama.ipynb) 3. Llama 3.2 Vision finetuning - Radiography use case. [Open in molab](https://github.com/unslothai/notebooks/blob/main/nb/Llama3.2_(11B)-Vision.ipynb) 4. See notebooks for DPO, ORPO, Continued pretraining, conversational finetuning and more on our [documentation](https://unsloth.ai/docs/get-started/unsloth-notebooks)!

Join Discord if you need help + ⭐️ Star us on Github ⭐️

This notebook and all Unsloth notebooks are licensed [LGPL-3.0](https://github.com/unslothai/notebooks?tab=LGPL-3.0-1-ov-file#readme). """) return if __name__ == "__main__": app.run()