gemma.py

# -*- coding: utf-8 -*-
"""Gemma3_(4B).ipynb

Automatically generated by Colab.

Original file is located at
    https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3_(4B).ipynb

To run this, press "*Runtime*" and press "*Run all*" on a **free** Tesla T4 Google Colab instance!
<div class="align-center">
<a href="https://unsloth.ai/"><img src="https://github.com/unslothai/unsloth/raw/main/images/unsloth%20new%20logo.png" width="115"></a>
<a href="https://discord.gg/unsloth"><img src="https://github.com/unslothai/unsloth/raw/main/images/Discord button.png" width="145"></a>
<a href="https://docs.unsloth.ai/"><img src="https://github.com/unslothai/unsloth/blob/main/images/documentation%20green%20button.png?raw=true" width="125"></a></a> Join Discord if you need help + ⭐ <i>Star us on <a href="https://github.com/unslothai/unsloth">Github</a> </i> ⭐
</div>

To install Unsloth on your own computer, follow the installation instructions on our Github page [here](https://docs.unsloth.ai/get-started/installing-+-updating).

You will learn how to do [data prep](#Data), how to [train](#Train), how to [run the model](#Inference), & [how to save it](#Save)

### News

**Read our [Gemma 3 blog](https://unsloth.ai/blog/gemma3) for what's new in Unsloth and our [Reasoning blog](https://unsloth.ai/blog/r1-reasoning) on how to train reasoning models.**

Visit our docs for all our [model uploads](https://docs.unsloth.ai/get-started/all-our-models) and [notebooks](https://docs.unsloth.ai/get-started/unsloth-notebooks).

### Installation
"""

# Commented out IPython magic to ensure Python compatibility.
# %%capture
# import os
# if "COLAB_" not in "".join(os.environ.keys()):
#     !pip install unsloth
# else:
#     # Do this only in Colab notebooks! Otherwise use pip install unsloth
#     !pip install --no-deps bitsandbytes accelerate xformers==0.0.29.post3 peft trl triton cut_cross_entropy unsloth_zoo
#     !pip install sentencepiece protobuf datasets huggingface_hub hf_transfer
#     !pip install --no-deps unsloth
# # Install latest Hugging Face for Gemma-3!
# !pip install --no-deps git+https://github.com/huggingface/transformers@v4.49.0-Gemma-3

"""### Unsloth

`FastModel` supports loading nearly any model now! This includes Vision and Text models!
"""

from unsloth import FastModel
import torch

fourbit_models = [
    # 4bit dynamic quants for superior accuracy and low memory use
    "unsloth/gemma-3-1b-it-unsloth-bnb-4bit",
    "unsloth/gemma-3-4b-it-unsloth-bnb-4bit",
    "unsloth/gemma-3-12b-it-unsloth-bnb-4bit",
    "unsloth/gemma-3-27b-it-unsloth-bnb-4bit",

    # Other popular models!
    "unsloth/Llama-3.1-8B",
    "unsloth/Llama-3.2-3B",
    "unsloth/Llama-3.3-70B",
    "unsloth/mistral-7b-instruct-v0.3",
    "unsloth/Phi-4",
] # More models at https://huggingface.co/unsloth

model, tokenizer = FastModel.from_pretrained(
    model_name = "unsloth/gemma-3-4b-it",
    max_seq_length = 8192, # Choose any for long context!
    load_in_4bit = False,  # 4 bit quantization to reduce memory
    load_in_8bit = False, # [NEW!] A bit more accurate, uses 2x memory
    full_finetuning = True, # [NEW!] We have full finetuning now!
    # token = "hf_...", # use one if using gated models
)

"""We now add LoRA adapters so we only need to update a small amount of parameters!"""

model = FastModel.get_peft_model(
    model,
    finetune_vision_layers     = False, # Turn off for just text!
    finetune_language_layers   = True,  # Should leave on!
    finetune_attention_modules = True,  # Attention good for GRPO
    finetune_mlp_modules       = True,  # SHould leave on always!

    r = 64,           # Larger = higher accuracy, but might overfit
    lora_alpha = 32,  # Recommended alpha == r at least
    lora_dropout = 0.1,
    bias = "none",
    random_state = 3407,
)

"""<a name="Data"></a>
### Data Prep
We now use the `Gemma-3` format for conversation style finetunes. We use [Maxime Labonne's FineTome-100k](https://huggingface.co/datasets/mlabonne/FineTome-100k) dataset in ShareGPT style. Gemma-3 renders multi turn conversations like below:

```
<bos><start_of_turn>user
Hello!<end_of_turn>
<start_of_turn>model
Hey there!<end_of_turn>
```

We use our `get_chat_template` function to get the correct chat template. We support `zephyr, chatml, mistral, llama, alpaca, vicuna, vicuna_old, phi3, llama3, phi4, qwen2.5, gemma3` and more.
"""

from unsloth.chat_templates import get_chat_template
tokenizer = get_chat_template(
    tokenizer,
    chat_template = "gemma-3",
)
from datasets import load_dataset
dataset = load_dataset("FourOhFour/RP_Phase", split = "train")

"""We now use `standardize_data_formats` to try converting datasets to the correct format for finetuning purposes!"""
from unsloth.chat_templates import standardize_data_formats
dataset = standardize_data_formats(dataset)

"""Let's see how row 100 looks like!"""
dataset[100]

"""We validate and fix conversations to ensure proper role alternation"""
def validate_and_fix_conversations(examples):
    valid_convs = []
    for conv in examples["conversations"]:
        # Check if roles alternate properly
        prev_role = None
        
        # Clean up the conversation to ensure proper alternation
        fixed_conv = []
        for turn in conv:
            role = turn.get("role", "").lower()
            # Skip if same role appears consecutively
            if role == prev_role:
                continue
            # Normalize roles to expected format
            if role in ["assistant", "bot", "chatbot"]:
                role = "model"
            elif role in ["human", "usr"]:
                role = "user"
            
            fixed_conv.append({"role": role, "content": turn.get("content", "")})
            prev_role = role
            
        # Ensure it starts with user and alternates correctly
        if fixed_conv and fixed_conv[0]["role"] == "user":
            valid_convs.append(fixed_conv)
    
    return {"conversations": valid_convs}

# Apply the validation and fixing step
dataset = dataset.map(validate_and_fix_conversations, batched=True)

"""We now have to apply the chat template for `Gemma-3` onto the conversations, and save it to `text`"""
def apply_chat_template(examples):
    texts = tokenizer.apply_chat_template(examples["conversations"])
    return { "text" : texts }

dataset = dataset.map(apply_chat_template, batched = True)

"""Let's see how the chat template did! Notice `Gemma-3` default adds a `<bos>`!"""
dataset[100]["text"]

"""<a name="Train"></a>
### Train the model
Now let's use Huggingface TRL's `SFTTrainer`! More docs here: [TRL SFT docs](https://huggingface.co/docs/trl/sft_trainer). 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`.
"""

from trl import SFTTrainer, SFTConfig
trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    eval_dataset = None, # Can set up evaluation!
    args = SFTConfig(
        dataset_text_field = "text",
        per_device_train_batch_size = 2,
        gradient_accumulation_steps = 4, # Use GA to mimic batch size!
        warmup_steps = 35,
        num_train_epochs = 2, # Set this for 1 full training run.
        learning_rate = 1e-5, # Reduce to 2e-5 for long training runs
        logging_steps = 1,
        optim = "paged_adamw_8bit",
        weight_decay = 0.02,
        lr_scheduler_type = "linear",
        seed = 3407,
        report_to = "wandb", # Use this for WandB etc
    ),
)

"""We also use Unsloth's `train_on_completions` method to only train on the assistant outputs and ignore the loss on the user's inputs. This helps increase accuracy of finetunes!"""

from unsloth.chat_templates import train_on_responses_only
trainer = train_on_responses_only(
    trainer,
    instruction_part = "<start_of_turn>user\n",
    response_part = "<start_of_turn>model\n",
)

"""Let's verify masking the instruction part is done! Let's print the 100th row again:"""

tokenizer.decode(trainer.train_dataset[100]["input_ids"])

"""Now let's print the masked out example - you should see only the answer is present:"""

tokenizer.decode([tokenizer.pad_token_id if x == -100 else x for x in trainer.train_dataset[100]["labels"]]).replace(tokenizer.pad_token, " ")

# @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.")

"""Let's train the model! To resume a training run, set `trainer.train(resume_from_checkpoint = True)`"""

trainer_stats = trainer.train()

# @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} %.")