# /// script # requires-python = ">=3.10,<3.14" # dependencies = [ # "accelerate", # "bitsandbytes>=0.43.0", # "datasets==4.3.0", # "hf_transfer", # "huggingface_hub>=0.34.0", # "marimo", # "peft", # "protobuf", # "sentencepiece", # "torchao>=0.16.0", # "transformers==4.56.2", # "triton>=3.2.0", # "trl==0.22.2", # "unsloth @ git+https://github.com/unslothai/unsloth", # "unsloth_zoo @ git+https://github.com/unslothai/unsloth-zoo", # ] # # [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 on your A100 molab Pro instance, 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""" ### News """) return @app.cell(hide_code=True) def _(mo): mo.md(r""" Introducing **Unsloth Studio** - a new open source, no-code web UI to train and run LLMs. [Blog](https://unsloth.ai/docs/new/studio) • [Notebook](https://github.com/unslothai/unsloth/blob/main/studio/Unsloth_Studio_Colab.ipynb)

_{Train models — no code needed}

_{Run GGUF models on Mac, Windows & Linux}

Train MoEs - DeepSeek, GLM, Qwen and gpt-oss 12x faster with 35% less VRAM. [Blog](https://unsloth.ai/docs/new/faster-moe) Ultra Long-Context Reinforcement Learning is here with 7x more context windows! [Blog](https://unsloth.ai/docs/new/grpo-long-context) New in Reinforcement Learning: [FP8 RL](https://unsloth.ai/docs/new/fp8-reinforcement-learning) • [Vision RL](https://unsloth.ai/docs/new/vision-reinforcement-learning-vlm-rl) • [Standby](https://unsloth.ai/docs/basics/memory-efficient-rl) • [gpt-oss RL](https://unsloth.ai/docs/new/gpt-oss-reinforcement-learning) Visit our docs for all our [model uploads](https://unsloth.ai/docs/get-started/unsloth-model-catalog) and [notebooks](https://unsloth.ai/docs/get-started/unsloth-notebooks). """) return @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Unsloth """) return @app.cell def _(): from unsloth import FastLanguageModel import torch max_seq_length = 32768 # Choose any! We auto support RoPE Scaling internally! dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+ load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False. # 4bit pre quantized models we support for 4x faster downloading + no OOMs. fourbit_models = [ "unsloth/Qwen2.5-Coder-0.5B-Instruct-bnb-4bit", "unsloth/Qwen2.5-Coder-0.5B-Instruct", "unsloth/Qwen2.5-Coder-1.5B-Instruct-bnb-4bit", "unsloth/Qwen2.5-Coder-1.5B-Instruct", "unsloth/Qwen2.5-Coder-3B-Instruct", "unsloth/Qwen2.5-Coder-3B-Instruct-bnb-4bit", "unsloth/Qwen2.5-Coder-14B-Instruct", "unsloth/Qwen2.5-Coder-14B-Instruct-bnb-4bit", "unsloth/Qwen2.5-Coder-32B-Instruct-bnb-4bit", ] # More models at https://huggingface.co/unsloth model, tokenizer = FastLanguageModel.from_pretrained( model_name="unsloth/Qwen2.5-Coder-7B-Instruct-bnb-4bit", # YOUR MODEL YOU USED FOR TRAINING max_seq_length=max_seq_length, # Choose any! We auto support RoPE Scaling internally! dtype=dtype, # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+ load_in_4bit=load_in_4bit, # Use 4bit quantization to reduce memory usage. Can be False. ) return ( FastLanguageModel, dtype, load_in_4bit, max_seq_length, model, tokenizer, torch, ) @app.cell(hide_code=True) def _(mo): mo.md(r""" We now add LoRA adapters so we only need to update 1 to 10% of all parameters! """) return @app.cell def _(FastLanguageModel, model): model_1 = FastLanguageModel.get_peft_model( model, r=16, # 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=16, lora_dropout=0, # Supports any, but = 0 is optimized bias="none", # Supports any, but = "none" is optimized use_gradient_checkpointing="unsloth", # True or "unsloth" for very long context random_state=3407, use_rslora=False, # We support rank stabilized LoRA loftq_config=None, # And LoftQ ) return (model_1,) @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Dataset We are using the [`open-r1/codeforces-cots`](https://huggingface.co/datasets/open-r1/codeforces-cots/viewer/solutions_decontaminated?views%5B%5D=solutions_decontaminated) by the Hugging Face Open R1 project. This is a dataset of problems for a competitive coding challenge. The Open R1 team have found that models trained to reason on competitive coding tasks like this outperform other models on coding benchmarks. ![image](https://huggingface.co/open-r1/OlympicCoder-7B/resolve/main/lcb-evals.png) """) return @app.cell def _(): from datasets import load_dataset # Login using e.g. `huggingface-cli login` to access this dataset ds = load_dataset("open-r1/codeforces-cots", "solutions_decontaminated") return (ds,) @app.cell def _(ds, tokenizer): tokenizer.apply_chat_template(ds["train"]["messages"][0], tokenize=False) return @app.function def formatting_prompts_func(example): output_texts = [] for i in range(len(example["problem"])): text = f"### Problem: {example['problem'][i]}\n ### Response: {example['generation'][i]}" output_texts.append(text) return output_texts @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Train the model Now let's train our model. We do 60 steps to speed things up, but you can set `num_train_epochs=1` for a full run, and turn off `max_steps=None`. We also support `DPOTrainer` and `GRPOTrainer` for reinforcement learning!! """) return @app.cell def _(ds, max_seq_length, model_1, tokenizer): from trl import SFTTrainer, SFTConfig from transformers import TrainingArguments trainer = SFTTrainer( model=model_1, tokenizer=tokenizer, train_dataset=ds["train"], max_seq_length=max_seq_length, # Choose any! We auto support RoPE Scaling internally! formatting_func=formatting_prompts_func, packing=False, # Can make training 5x faster for short sequences. args=SFTConfig( per_device_train_batch_size=2, gradient_accumulation_steps=4, warmup_steps=5, max_steps=60, learning_rate=0.0002, logging_steps=1, optim="adamw_8bit", weight_decay=0.001, lr_scheduler_type="linear", seed=3407, output_dir="outputs", report_to="none", # Use TrackIO/WandB etc ), ) return (trainer,) @app.cell def _(torch): # @title Show current memory stats gpu_stats = torch.cuda.get_device_properties(0) start_gpu_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3) max_memory = round(gpu_stats.total_memory / 1024 / 1024 / 1024, 3) print(f"GPU = {gpu_stats.name}. Max memory = {max_memory} GB.") print(f"{start_gpu_memory} GB of memory reserved.") return max_memory, start_gpu_memory @app.cell def _(trainer): trainer_stats = trainer.train() return (trainer_stats,) @app.cell def _(max_memory, start_gpu_memory, torch, trainer_stats): # @title Show final memory and time stats used_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3) used_memory_for_lora = round(used_memory - start_gpu_memory, 3) used_percentage = round(used_memory / max_memory * 100, 3) lora_percentage = round(used_memory_for_lora / max_memory * 100, 3) print(f"{trainer_stats.metrics['train_runtime']} seconds used for training.") print( f"{round(trainer_stats.metrics['train_runtime'] / 60, 2)} minutes used for training." ) print(f"Peak reserved memory = {used_memory} GB.") print(f"Peak reserved memory for training = {used_memory_for_lora} GB.") print(f"Peak reserved memory % of max memory = {used_percentage} %.") print(f"Peak reserved memory for training % of max memory = {lora_percentage} %.") return @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Inference Let's run the model! You can change the instruction and input - leave the output blank! """) return @app.cell def _(FastLanguageModel, model_1, tokenizer): example_problem = "A tech company, \"DataGoSlow,\" produces high-performance memory modules. They sell these modules in 'n' different retail outlets. Each outlet sells the modules at a unique price, represented by 'xi' dollars.\n\nA large data center plans to purchase these memory modules over 'q' consecutive procurement cycles. For each cycle, the data center has a budget constraint 'mi' dollars. They want to know how many outlets sell the memory modules at a price *strictly less than* their budget.\n\n**Constraints:**\n\n* Time limit per test: 2.0 seconds\n* Memory limit per test: 256.0 megabytes\n\n**Input Format:**\n\n* The first line of the input contains a single integer 'n' (1 ≤ n ≤ 100 000) — the number of retail outlets.\n* The second line contains 'n' integers 'xi' (1 ≤ xi ≤ 100 000) — prices of the memory modules in each outlet.\n* The third line contains a single integer 'q' (1 ≤ q ≤ 100 000) — the number of procurement cycles.\n* Then follow 'q' lines, each containing one integer 'mi' (1 ≤ mi ≤ 109) — the budget for the i-th procurement cycle.\n\n**Output Format:**\n\n* Print 'q' integers. The i-th of them should be equal to the number of outlets where the memory modules can be purchased at a price *strictly less than* the budget for the i-th cycle.\n\n**Example:**\n\n```input\n5\n3 10 8 6 11\n4\n1\n10\n3\n11" FastLanguageModel.for_inference(model_1) inputs = tokenizer([example_problem], return_tensors="pt").to("cuda") outputs = model_1.generate(**inputs, max_new_tokens=64, use_cache=True) tokenizer.batch_decode(outputs) # Enable native 2x faster inference return (inputs,) @app.cell(hide_code=True) def _(mo): mo.md(r""" You can also use a `TextStreamer` for continuous inference - so you can see the generation token by token, instead of waiting the whole time! """) return @app.cell def _(FastLanguageModel, inputs, model_1, tokenizer): FastLanguageModel.for_inference(model_1) # Enable native 2x faster inference from transformers import TextStreamer text_streamer = TextStreamer(tokenizer) _ = model_1.generate(**inputs, streamer=text_streamer, max_new_tokens=128) return (TextStreamer,) @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Saving, loading finetuned models To save the final model as LoRA adapters, either use Hugging Face's `push_to_hub` for an online save or `save_pretrained` for a local save. **[NOTE]** This ONLY saves the LoRA adapters, and not the full model. To save to 16bit or GGUF, scroll down! """) return @app.cell def _(model_1, tokenizer): model_1.save_pretrained("codeforces_cot_finetune_lora") tokenizer.save_pretrained("codeforces_cot_finetune_lora") return @app.cell(hide_code=True) def _(mo): mo.md(r""" Now if you want to load the LoRA adapters we just saved for inference, set `False` to `True`: """) return @app.cell def _( TextStreamer, alpaca_prompt, dtype, load_in_4bit, max_seq_length, model_1, tokenizer, ): if False: from unsloth import FastLanguageModel as _FastLanguageModel _model, _tokenizer = _FastLanguageModel.from_pretrained( model_name="codeforces_cot_finetune_lora", # YOUR MODEL YOU USED FOR TRAINING max_seq_length=max_seq_length, # Choose any! We auto support RoPE Scaling internally! dtype=dtype, # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+ load_in_4bit=load_in_4bit, # Use 4bit quantization to reduce memory usage. Can be False. ) _FastLanguageModel.for_inference(_model) inputs_1 = tokenizer( [alpaca_prompt.format("What is a famous tall tower in Paris?", "", "")], return_tensors="pt", ).to("cuda") # alpaca_prompt = You MUST copy from above! text_streamer_1 = TextStreamer(tokenizer) _ = model_1.generate( **inputs_1, streamer=text_streamer_1, max_new_tokens=128 ) return @app.cell(hide_code=True) def _(mo): mo.md(r""" You can also use Hugging Face's `AutoPeftModelForCausalLM`. Only use this if you do not have `unsloth` installed. It can be hopelessly slow, since `4bit` model downloading is not supported, and Unsloth's **inference is 2x faster**. """) return @app.cell def _(load_in_4bit): if False: from peft import AutoPeftModelForCausalLM as _AutoPeftModelForCausalLM from transformers import AutoTokenizer as _AutoTokenizer _model = _AutoPeftModelForCausalLM.from_pretrained( "codeforces_cot_finetune_lora", load_in_4bit=load_in_4bit ) _tokenizer = _AutoTokenizer.from_pretrained("codeforces_cot_finetune_lora") 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( "codeforces_cot_finetune_16bit", tokenizer, save_method="merged_16bit" ) if False: # Merge to 4bit model_1.push_to_hub_merged( "HF_USERNAME/codeforces_cot_finetune_16bit", tokenizer, save_method="merged_16bit", token="YOUR_HF_TOKEN", # Get a token at https://huggingface.co/settings/tokens ) if False: model_1.save_pretrained_merged( "codeforces_cot_finetune_4bit", tokenizer, save_method="merged_4bit" ) # Just LoRA adapters if False: model_1.push_to_hub_merged( "HF_USERNAME/codeforces_cot_finetune_4bit", tokenizer, save_method="merged_4bit", token="YOUR_HF_TOKEN", # Get a token at https://huggingface.co/settings/tokens ) if False: model_1.save_pretrained("codeforces_cot_finetune_lora") tokenizer.save_pretrained("codeforces_cot_finetune_lora") if False: model_1.push_to_hub( "HF_USERNAME/codeforces_cot_finetune_lora", token="YOUR_HF_TOKEN" ) tokenizer.push_to_hub( "HF_USERNAME/codeforces_cot_finetune_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("codeforces_cot_finetune", tokenizer) if False: # Save to 16bit GGUF model_1.push_to_hub_gguf( "HF_USERNAME/codeforces_cot_finetune", tokenizer, token="YOUR_HF_TOKEN" ) if False: model_1.save_pretrained_gguf( "codeforces_cot_finetune", tokenizer, quantization_method="f16" ) # Save to q4_k_m GGUF if False: model_1.push_to_hub_gguf( "HF_USERNAME/codeforces_cot_finetune", tokenizer, quantization_method="f16", token="YOUR_HF_TOKEN", # Get a token at https://huggingface.co/settings/tokens ) if False: # Save to multiple GGUF options - much faster if you want multiple! model_1.save_pretrained_gguf( "codeforces_cot_finetune", tokenizer, quantization_method="q4_k_m" ) if False: model_1.push_to_hub_gguf( "HF_USERNAME/codeforces_cot_finetune", tokenizer, quantization_method="q4_k_m", token="YOUR_HF_TOKEN", # Get a token at https://huggingface.co/settings/tokens ) # Change hf to your username! if False: model_1.push_to_hub_gguf( "HF_USERNAME/codeforces_cot_finetune", tokenizer, quantization_method=["q4_k_m", "q8_0", "q5_k_m"], token="YOUR_HF_TOKEN", # Get a token at https://huggingface.co/settings/tokens ) # Get a token at https://huggingface.co/settings/tokens return @app.cell(hide_code=True) def _(mo): mo.md(r""" Now, use the `codeforces_cot_finetune.Q8_0.gguf` file or `codeforces_cot_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()