2025-03

# March Release 🦥 
Get the latest stable Unsloth via:
```
pip install --upgrade --force-reinstall --no-cache-dir unsloth unsloth_zoo
```
The March release should be stable - you can force the version via:
```
pip install "unsloth==2025.3.18" "unsloth_zoo==2025.3.16"
```

## New Features
* Read all details here: https://unsloth.ai/blog/gemma3
* **Gemma 3 1B, 4B, 12B and 27B** finetuning all work now! [**Colab Notebook**](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3_(4B).ipynb) We fixed some issues which caused **Gemma 3 training loss to be very high**. This includes some tokenization issues so fine-tuning Gemma 3 will now work correctly if you use Unsloth.
![image](https://github.com/user-attachments/assets/fe42432c-cbdd-4b3c-bd2a-46a10afe5209)

* We also encountered many **infinite gradients during Gemma 3 (1B to 27B) finetuning**. We found float16 mixed precision (Tesla T4, RTX 2080 series) to not function well, and we defaulted to float32 precision. Float16 also failed on A100, so this is a hardware agnostic issue. Bfloat16 is fine though! Unsloth auto selects the best data-type! You do not have to do anything! [**Colab Notebook to finetune Gemma 3**](https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/Gemma3_(4B).ipynb)

* Preliminary support for **full-finetuning** and **8bit finetuning** - set `full_finetuning = True` or `load_in_8bit = True` Both will be optimized further in the future! A reminder you will need more powerful GPUs!
```python
model, tokenizer = FastModel.from_pretrained(
    model_name = "unsloth/gemma-3-4B-it",
    max_seq_length = 2048, # Choose any for long context!
    load_in_4bit = True,  # 4 bit quantization to reduce memory
    load_in_8bit = False, # [NEW!] A bit more accurate, uses 2x memory
    full_finetuning = False, # [NEW!] We have full finetuning now!
    # token = "hf_...", # use one if using gated models
)
```
* New Unsloth Auto Model support - nearly **all models are now supported**! We now supports vision and text models out of the box, without the need for custom implementations (and all are optimized!)
* Mixtral (yes finally!), Gemma 3, Granite 3.2, Cohere, OLMo, Reka, and generally any vision or language model! There might be some occasional models which don't work!
```python
model, tokenizer = FastModel.from_pretrained(
    model_name = "mistralai/Mixtral-8x7B-Instruct-v0.1",
)
```
* **Windows support** via pip install unsloth should function now! Utilizes https://pypi.org/project/triton-windows/ which provides a pip installable path for Triton. Use:
```
pip install unsloth
```
* Train on completions / responses only for vision models supported! Use it like below:
```python
data_collator = UnslothVisionDataCollator(
    model,
    tokenizer,
    train_on_responses_only = False,
    instruction_part = "<|start_header_id|>user<|end_header_id|>\n\n",
    response_part = "<|start_header_id|>assistant<|end_header_id|>\n\n",
)
SFTTrainer(..., data_collator = data_collator)
```
* Conversions to llama.cpp GGUFs for 16bit and 8bit now **DO NOT need compiling**! This solves many many issues, and this means no need to install GCC, Microsoft Visual Studio etc!
```python
model.save_pretrained_merged("gemma-3-finetune", tokenizer)
model.save_pretrained_gguf(
    "gemma-3-finetune",
    quantization_type = "Q8_0", # For now only Q8_0, BF16, F16 supported
)
```
* **Vision models now auto resize images** which stops OOMs and also allows truncating sequence lengths!
* Many multiple optimizations in Unsloth allowing a further **+10% less VRAM usage**, and **>10% speedup boost** for 4bit (on top of our original 2x faster, 70% less memory usage). 8bit and full finetuning also benefit!
* GRPO in Unsloth now allows non Unsloth uploaded models to be in 4bit as well - reduces VRAM usage a lot! (ie pretend your own finetune of Llama)
* New training logs and infos - training parameter counts, total batch size
![image](https://github.com/user-attachments/assets/db4a4a1c-483b-4722-9a14-499c396efd7d)

* Vision models now also work for normal text training! This means non vision notebooks can work with vision models!

* **Complete gradient accumulation bug fix coverage** for all models!
* GRPO notebook for Gemma 3 coming soon with Hugging Face's reasoning course!
* DoRA, Dropout, and other PEFT methods should just work!

## Bug fixes
* Faster and less error prone streamlined finetuning experience! Apologies for the recent issues with constant releases and breaking breaks - the March release should be stable! Ie `pip install "unsloth==2025.3.14" "unsloth_zoo==2025.3.12"`
* Pixtral and Llava finetuning are now fixed! In fact nearly all vision models are supported out of the box! Please update transformers for Pixtral: `pip install --no-deps git+https://github.com/huggingface/transformers.git`
* Fixed all Colabs not working - cloud instances like Runpod should just work now!
* Fixed many many bugs - will reply to each issue with updates!

## Other items
* GRPO Bug fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/1623
* Fixes Triton url in README.md by @DiogoNeves in https://github.com/unslothai/unsloth/pull/1607
* Update README.md by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/1654
* Update README.md by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/1688
* Fix bugs by @danielhanchen in https://github.com/unslothai/unsloth/pull/1701
* Fix bugs by @danielhanchen in https://github.com/unslothai/unsloth/pull/1706
* Memory efficient GRPO, DPO etc by @danielhanchen in https://github.com/unslothai/unsloth/pull/1716
* Add GRPO metrics by @danielhanchen in https://github.com/unslothai/unsloth/pull/1718
* llama-quantize on WINDOWS WSL error fix - edit save.py (gguf saving breaks) by @everythingisc00l in https://github.com/unslothai/unsloth/pull/1649
* Update rl_replacements.py by @SethHWeidman in https://github.com/unslothai/unsloth/pull/1754
* Update README.md by @danielhanchen in https://github.com/unslothai/unsloth/pull/1768
* fix an import error by @NinoRisteski in https://github.com/unslothai/unsloth/pull/1767
* Gemma Mask convert to float by @Erland366 in https://github.com/unslothai/unsloth/pull/1762
* [Windows Support] Add latest `xformers` wheels to pyproject.toml by @versipellis in https://github.com/unslothai/unsloth/pull/1753
* Memory Efficient GRPO by @danielhanchen in https://github.com/unslothai/unsloth/pull/1773
* Bug Fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/1774
* Export Model to ollama.com  by @gjyotin305 in https://github.com/unslothai/unsloth/pull/1648
* Fix: GRPO with Mistral and importing by @oKatanaaa in https://github.com/unslothai/unsloth/pull/1831
* Fix key error in GRPOTrainer by @le-big-mac in https://github.com/unslothai/unsloth/pull/1818
* fixed syntax warnings by @KareemMusleh in https://github.com/unslothai/unsloth/pull/1522
* Direct windows support for unsloth by @adityaghai07 in https://github.com/unslothai/unsloth/pull/1841
* Fix Layernorm when num_cols not a power of 2 by @MekkCyber in https://github.com/unslothai/unsloth/pull/1867
* Added Python version warning to Windows Install Section by @areebuzair in https://github.com/unslothai/unsloth/pull/1872
* Update README.md by @shimmyshimmer in https://github.com/unslothai/unsloth/pull/1885
* Bug fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/1891
* Many bug fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/1900
* Logits fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/1916
* Bug fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/1920
* Bug fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/1951
* move use_modelscope to _utils by @KareemMusleh in https://github.com/unslothai/unsloth/pull/1938
* Don't use revision when loading model_config and is_peft=True by @wiwu2390 in https://github.com/unslothai/unsloth/pull/1949
* More syntax warnings by @KareemMusleh in https://github.com/unslothai/unsloth/pull/1944
* Gemma 3 by @danielhanchen in https://github.com/unslothai/unsloth/pull/1986
* Gemma 3 bug fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/2005
* Triton windows update by @Captain-T2004 in https://github.com/unslothai/unsloth/pull/1976
* Update RMS LayerNorm implementation, and list compr. change in chat templates by @NinoRisteski in https://github.com/unslothai/unsloth/pull/1974
* Gemma 3, bug fixes by @danielhanchen in https://github.com/unslothai/unsloth/pull/2014

## New Contributors
* @DiogoNeves made their first contribution in https://github.com/unslothai/unsloth/pull/1607
* @everythingisc00l made their first contribution in https://github.com/unslothai/unsloth/pull/1649
* @SethHWeidman made their first contribution in https://github.com/unslothai/unsloth/pull/1754
* @versipellis made their first contribution in https://github.com/unslothai/unsloth/pull/1753
* @gjyotin305 made their first contribution in https://github.com/unslothai/unsloth/pull/1648
* @le-big-mac made their first contribution in https://github.com/unslothai/unsloth/pull/1818
* @MekkCyber made their first contribution in https://github.com/unslothai/unsloth/pull/1867
* @areebuzair made their first contribution in https://github.com/unslothai/unsloth/pull/1872
* @wiwu2390 made their first contribution in https://github.com/unslothai/unsloth/pull/1949
* @Captain-T2004 made their first contribution in https://github.com/unslothai/unsloth/pull/1976

**Full Changelog**: https://github.com/unslothai/unsloth/compare/2025-02...2025-03