# /// script # requires-python = ">=3.10,<3.14" # dependencies = [ # "causal-conv1d @ git+https://github.com/Dao-AILab/causal-conv1d.git@main", # "mamba @ git+https://github.com/state-spaces/mamba.git@main", # "marimo", # "torchao>=0.16.0", # "transformers @ git+https://github.com/huggingface/transformers.git", # "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""" # Unsloth Training for Falcon H1 This Notebook has been authored by TII Falcon Team. For more details on Falcon H1 series of models : 1. [Official Page](https://tiiuae.github.io/Falcon-H1/) 2. [blogpost](https://falcon-lm.github.io/blog/falcon-h1/) 3. [Official github page ](https://github.com/tiiuae/Falcon-H1) 4. [hf collection](https://huggingface.co/collections/tiiuae/falcon-h1-6819f2795bc406da60fab8df) """) return @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 own computer, follow the installation instructions on our Github page [here](https://unsloth.ai/docs/get-started/install). 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 def _(): import subprocess import subprocess # Installs Unsloth, Xformers (Flash Attention) and all other packages! # Get latest Unsloth #! pip uninstall unsloth -y subprocess.call(["pip", "uninstall", "unsloth", "-y"]) return @app.cell def _(): import unsloth from unsloth import FastLanguageModel import torch import os os.environ["TRITON_JIT_DISABLE_OPT"] = "1" # Likely the most critical change max_seq_length = 2048 # 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. model, tokenizer = FastLanguageModel.from_pretrained( model_name="tiiuae/Falcon-H1-0.5B-Instruct", # Choose any model from https://huggingface.co/collections/tiiuae/falcon-h1-6819f2795bc406da60fab8df 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, max_seq_length, model, os, tokenizer, torch @app.cell def _(FastLanguageModel, model): # Configure PEFT model model_1 = FastLanguageModel.get_peft_model( model, r=64, target_modules=[ "q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj", ], lora_alpha=32, lora_dropout=0.1, use_gradient_checkpointing=False, random_state=3407, ) # Mamba out_proj and conv1d layers should not be included here see https://github.com/huggingface/peft/pull/2562 return (model_1,) @app.cell def _(tokenizer): alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request. ### Instruction: {} ### Input: {} ### Response: {}""" EOS_TOKEN = tokenizer.eos_token def formatting_prompts_func(examples): instructions = examples["instruction"] inputs = examples["input"] outputs = examples["output"] texts = [] for instruction, input, output in zip(instructions, inputs, outputs): # Must add EOS_TOKEN, otherwise your generation will go on forever! text = alpaca_prompt.format(instruction, input, output) + EOS_TOKEN texts.append(text) return { "text": texts, } from datasets import load_dataset dataset = load_dataset("yahma/alpaca-cleaned", split="train") dataset = dataset.map( formatting_prompts_func, batched=True, ) return (dataset,) @app.cell def _(dataset, max_seq_length, model_1, tokenizer, torch): from trl import SFTTrainer from transformers import TrainingArguments trainer = SFTTrainer( model=model_1, tokenizer=tokenizer, train_dataset=dataset, dataset_text_field="text", max_seq_length=max_seq_length, # Choose any! We auto support RoPE Scaling internally! packing=False, # Can make training 5x faster for short sequences. args=TrainingArguments( per_device_train_batch_size=1, gradient_accumulation_steps=8, warmup_steps=5, max_steps=60, learning_rate=0.0002, fp16=not torch.cuda.is_bf16_supported(), bf16=torch.cuda.is_bf16_supported(), logging_steps=1, optim="adamw_8bit", weight_decay=0.001, lr_scheduler_type="linear", seed=3407, output_dir="outputs", ), ) # Can make training 5x faster for short sequences. return (trainer,) @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Show current memory stats """) return @app.cell def _(torch): 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(hide_code=True) def _(mo): mo.md(r""" # Training """) return @app.cell def _(os, trainer): os.environ["TRITON_DISABLE_LINE_INFO"] = "1" trainer_stats = trainer.train() return (trainer_stats,) @app.cell(hide_code=True) def _(mo): mo.md(r""" ### Show final memory and time stats """) return @app.cell def _(max_memory, start_gpu_memory, torch, trainer_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""" 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()