diff --git a/pyproject.toml b/pyproject.toml index 2f812c769f..6409d74c05 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -39,11 +39,10 @@ triton = [ "triton>=3.0.0 ; ('linux' in sys_platform)", "triton-windows ; (sys_platform == 'win32') and (platform_machine == 'AMD64' or platform_machine == 'x86_64')", ] -huggingface = [ - "unsloth_zoo>=2025.10.4", +huggingfacenotorch = [ + "unsloth_zoo>=2025.10.5", "wheel>=0.42.0", "packaging", - "torchvision", "numpy", "tqdm", "psutil", @@ -58,6 +57,10 @@ huggingface = [ "diffusers", "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,<=4.56.2", "trl>=0.7.9,!=0.9.0,!=0.9.1,!=0.9.2,!=0.9.3,!=0.15.0,!=0.19.0,<=0.23.0", +] +huggingface = [ + "unsloth[huggingfacenotorch]", + "torchvision", "unsloth[triton]", ] windows = [ @@ -458,7 +461,7 @@ colab-ampere-torch220 = [ "flash-attn>=2.6.3 ; ('linux' in sys_platform)", ] colab-new = [ - "unsloth_zoo>=2025.10.4", + "unsloth_zoo>=2025.10.5", "packaging", "tyro", "transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,<=4.56.2", @@ -740,7 +743,12 @@ intel-gpu-torch270 = [ "torch @ https://download.pytorch.org/whl/xpu/torch-2.7.0%2Bxpu-cp312-cp312-linux_x86_64.whl#sha256=c806d44aa2ca5d225629f6fbc6c994d5deaac2d2cde449195bc8e3522ddd219a ; ('linux' in sys_platform) and python_version == '3.12' and (platform_machine == 'AMD64' or platform_machine == 'x86_64')", "torch @ https://download.pytorch.org/whl/xpu/torch-2.7.0%2Bxpu-cp313-cp313-linux_x86_64.whl#sha256=25d8277b7f01d42e2e014ccbab57a2692b6ec4eff8dcf894eda1b297407cf97a ; ('linux' in sys_platform) and python_version == '3.13' and (platform_machine == 'AMD64' or platform_machine == 'x86_64')", ] - +amd = [ + "unsloth[huggingfacenotorch]", + "bitsandbytes @ https://github.com/bitsandbytes-foundation/bitsandbytes/releases/download/continuous-release_main/bitsandbytes-1.33.7.preview-py3-none-manylinux_2_24_x86_64.whl ; ('linux' in sys_platform) and (platform_machine == 'AMD64' or platform_machine == 'x86_64')", + "bitsandbytes @ https://github.com/bitsandbytes-foundation/bitsandbytes/releases/download/continuous-release_main/bitsandbytes-1.33.7.preview-py3-none-win_amd64.whl ; (sys_platform == 'win32') and (platform_machine == 'AMD64' or platform_machine == 'x86_64')", + "bitsandbytes @ https://github.com/bitsandbytes-foundation/bitsandbytes/releases/download/continuous-release_main/bitsandbytes-1.33.7.preview-py3-none-manylinux_2_24_aarch64.whl ; ('linux' in sys_platform) and (platform_machine == 'aarch64')", +] [project.urls] homepage = "http://www.unsloth.ai" documentation = "https://github.com/unslothai/unsloth" diff --git a/unsloth/__init__.py b/unsloth/__init__.py index 99d651ae5f..b336ad03fe 100644 --- a/unsloth/__init__.py +++ b/unsloth/__init__.py @@ -69,49 +69,14 @@ except Exception as exception: raise exception pass -@functools.cache -def is_hip(): - return bool(getattr(getattr(torch, "version", None), "hip", None)) -pass - -@functools.cache -def get_device_type(): - if hasattr(torch, "cuda") and torch.cuda.is_available(): - if is_hip(): - return "hip" - return "cuda" - elif hasattr(torch, "xpu") and torch.xpu.is_available(): - return "xpu" - # Check torch.accelerator - if hasattr(torch, "accelerator"): - if not torch.accelerator.is_available(): - raise NotImplementedError("Unsloth cannot find any torch accelerator? You need a GPU.") - accelerator = str(torch.accelerator.current_accelerator()) - if accelerator in ("cuda", "xpu", "hip"): - raise RuntimeError( - f"Unsloth: Weirdly `torch.cuda.is_available()`, `torch.xpu.is_available()` and `is_hip` all failed.\n"\ - f"But `torch.accelerator.current_accelerator()` works with it being = `{accelerator}`\n"\ - f"Please reinstall torch - it's most likely broken :(" - ) - raise NotImplementedError("Unsloth currently only works on NVIDIA, AMD and Intel GPUs.") -pass -DEVICE_TYPE : str = get_device_type() -# HIP fails for autocast and other torch functions. Use CUDA instead -DEVICE_TYPE_TORCH = DEVICE_TYPE -if DEVICE_TYPE_TORCH == "hip": DEVICE_TYPE_TORCH = "cuda" - -@functools.cache -def get_device_count(): - if DEVICE_TYPE in ("cuda", "hip"): - return torch.cuda.device_count() - elif DEVICE_TYPE == "xpu": - return torch.xpu.device_count() - else: - return 1 -pass - -DEVICE_COUNT : int = get_device_count() - +from .device_type import ( + is_hip, + get_device_type, + DEVICE_TYPE, + DEVICE_TYPE_TORCH, + DEVICE_COUNT, + ALLOW_PREQUANTIZED_MODELS, +) # Reduce VRAM usage by reducing fragmentation # And optimize pinning of memory # TODO(billishyahao): need to add hip related optimization... @@ -201,7 +166,10 @@ if DEVICE_TYPE == "cuda": else: from triton.common.build import libcuda_dirs # Try loading bitsandbytes and triton - import bitsandbytes as bnb + try: + import bitsandbytes as bnb + except: + print("Unsloth: `bitsandbytes` is not installed - 4bit QLoRA unallowed, but 16bit and full finetuning works!") try: cdequantize_blockwise_fp32 = bnb.functional.lib.cdequantize_blockwise_fp32 libcuda_dirs() diff --git a/unsloth/_amd_install.sh b/unsloth/_amd_install.sh new file mode 100644 index 0000000000..b83dbd2ad0 --- /dev/null +++ b/unsloth/_amd_install.sh @@ -0,0 +1,32 @@ +#!/usr/bin/env bash +# _amd_install.sh +# Non-interactive installer: build tools, PyTorch (ROCm 6.4), bitsandbytes (HIP), and Unsloth from source. +# Usage: +# bash _amd_install.sh +# + +set -euo pipefail +export DEBIAN_FRONTEND=noninteractive + +apt-get update +apt-get install -y --no-install-recommends build-essential cmake git + +pip install \ + torch==2.8.0 torchvision torchaudio torchao==0.13.0 xformers \ + --index-url https://download.pytorch.org/whl/rocm6.4 + +WORKDIR="$(pwd)" +TMPDIR="$(mktemp -d)" +cd "$TMPDIR" +git clone https://github.com/bitsandbytes-foundation/bitsandbytes.git +cd bitsandbytes +arch +cmake -DCOMPUTE_BACKEND=hip -S . +make -j"$(nproc)" +pip install . +cd "$WORKDIR" +rm -rf "$TMPDIR" + +pip install --no-deps unsloth unsloth-zoo +pip install "unsloth_zoo[base] @ git+https://github.com/unslothai/unsloth-zoo" +pip install "unsloth[base] @ git+https://github.com/unslothai/unsloth" diff --git a/unsloth/device_type.py b/unsloth/device_type.py new file mode 100644 index 0000000000..547750019a --- /dev/null +++ b/unsloth/device_type.py @@ -0,0 +1,80 @@ +# Copyright 2023-present Daniel Han-Chen & the Unsloth team. All rights reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +__all__ = [ + "is_hip", + "get_device_type", + "DEVICE_TYPE", + "DEVICE_TYPE_TORCH", + "DEVICE_COUNT", + "ALLOW_PREQUANTIZED_MODELS", +] + +import torch +import functools + +@functools.cache +def is_hip(): + return bool(getattr(getattr(torch, "version", None), "hip", None)) +pass + +@functools.cache +def get_device_type(): + if hasattr(torch, "cuda") and torch.cuda.is_available(): + if is_hip(): + return "hip" + return "cuda" + elif hasattr(torch, "xpu") and torch.xpu.is_available(): + return "xpu" + # Check torch.accelerator + if hasattr(torch, "accelerator"): + if not torch.accelerator.is_available(): + raise NotImplementedError("Unsloth cannot find any torch accelerator? You need a GPU.") + accelerator = str(torch.accelerator.current_accelerator()) + if accelerator in ("cuda", "xpu", "hip"): + raise RuntimeError( + f"Unsloth: Weirdly `torch.cuda.is_available()`, `torch.xpu.is_available()` and `is_hip` all failed.\n"\ + f"But `torch.accelerator.current_accelerator()` works with it being = `{accelerator}`\n"\ + f"Please reinstall torch - it's most likely broken :(" + ) + raise NotImplementedError("Unsloth currently only works on NVIDIA, AMD and Intel GPUs.") +pass +DEVICE_TYPE : str = get_device_type() +# HIP fails for autocast and other torch functions. Use CUDA instead +DEVICE_TYPE_TORCH = DEVICE_TYPE +if DEVICE_TYPE_TORCH == "hip": DEVICE_TYPE_TORCH = "cuda" + +@functools.cache +def get_device_count(): + if DEVICE_TYPE in ("cuda", "hip"): + return torch.cuda.device_count() + elif DEVICE_TYPE == "xpu": + return torch.xpu.device_count() + else: + return 1 +pass + +DEVICE_COUNT : int = get_device_count() + +# Check blocksize for 4bit -> 64 for CUDA, 128 for AMD +# If AMD, we cannot load pre-quantized models for now :( +ALLOW_PREQUANTIZED_MODELS : bool = True +if DEVICE_TYPE == "hip": + try: + from bitsandbytes.nn.modules import Params4bit + if "blocksize = 64 if not HIP_ENVIRONMENT else 128" in inspect.getsource(Params4bit): + ALLOW_PREQUANTIZED_MODELS = False + except: + pass +pass diff --git a/unsloth/kernels/utils.py b/unsloth/kernels/utils.py index 9a46d0d5d7..16fc694230 100644 --- a/unsloth/kernels/utils.py +++ b/unsloth/kernels/utils.py @@ -19,7 +19,14 @@ next_power_of_2 = triton.next_power_of_2 import functools from typing import Optional -from .. import DEVICE_TYPE, DEVICE_COUNT +from ..device_type import ( + is_hip, + get_device_type, + DEVICE_TYPE, + DEVICE_TYPE_TORCH, + DEVICE_COUNT, + ALLOW_PREQUANTIZED_MODELS, +) from .fp8 import weight_dequant, fp8_linear # torch.cuda.amp.custom_fwd is deprecated >= 2.4 diff --git a/unsloth/models/_utils.py b/unsloth/models/_utils.py index beefada7bf..e713c8efff 100644 --- a/unsloth/models/_utils.py +++ b/unsloth/models/_utils.py @@ -12,7 +12,7 @@ # See the License for the specific language governing permissions and # limitations under the License. -__version__ = "2025.10.4" +__version__ = "2025.10.5" __all__ = [ "SUPPORTS_BFLOAT16", @@ -87,7 +87,14 @@ import functools import warnings, subprocess, re, inspect, psutil, os, math from unsloth_zoo.utils import Version from importlib.metadata import version as importlib_version -from unsloth import DEVICE_TYPE, DEVICE_COUNT, DEVICE_TYPE_TORCH +from ..device_type import ( + is_hip, + get_device_type, + DEVICE_TYPE, + DEVICE_TYPE_TORCH, + DEVICE_COUNT, + ALLOW_PREQUANTIZED_MODELS, +) from unsloth_zoo.log import logger from unsloth_zoo.tokenizer_utils import ( patch_tokenizer as _patch_tokenizer, @@ -1331,6 +1338,7 @@ pass def patch_gradient_accumulation_fix(Trainer): # Fixes gradient accumulation + # Fixes Output 0 of UnslothFusedLossBackward is a view and is being modified inplace. import inspect if hasattr(Trainer, "get_batch_samples"): if Trainer.get_batch_samples.__name__ == "_unsloth_get_batch_samples": return @@ -1346,6 +1354,32 @@ def patch_gradient_accumulation_fix(Trainer): # Also fix passing in num_items_in_batch if not hasattr(Trainer, "_old_compute_loss"): + + # Fix transformers 4.57.0 causing `Output 0 of UnslothFusedLossBackward is a view and is being modified inplace.` + function = inspect.getsource(Trainer.compute_loss) + if "loss *=" in function or "loss*=" in function: + where = function.find("def") + function = function.split("\n") + function = "\n".join(x[where:] for x in function) + + # Import all variables that need importing + import transformers.trainer + items_in_trainer = dir(transformers.trainer) + good_items = [] + for item in items_in_trainer: + if item in function: good_items.append(item) + pass + exec("from transformers.trainer import (" + ", ".join(x for x in good_items) + ")", globals()) + + # Replace loss*= with loss = loss * + function = re.sub( + r"loss[\s]{0,}\*\=", + "loss = loss *", + function, + ) + exec(function, globals()) + Trainer.compute_loss = compute_loss + pass Trainer._old_compute_loss = Trainer.compute_loss Trainer.compute_loss = _unsloth_pre_compute_loss pass diff --git a/unsloth/models/llama.py b/unsloth/models/llama.py index 596042288d..535537a3a1 100644 --- a/unsloth/models/llama.py +++ b/unsloth/models/llama.py @@ -27,7 +27,14 @@ from transformers import __version__ as transformers_version from unsloth_zoo.utils import Version, _get_dtype from unsloth_zoo.hf_utils import dtype_from_config, add_dtype_kwargs, fix_lora_auto_mapping from unsloth_zoo.peft_utils import SKIP_QUANTIZATION_MODULES -from unsloth import DEVICE_TYPE, DEVICE_COUNT, DEVICE_TYPE_TORCH +from ..device_type import ( + is_hip, + get_device_type, + DEVICE_TYPE, + DEVICE_TYPE_TORCH, + DEVICE_COUNT, + ALLOW_PREQUANTIZED_MODELS, +) transformers_version = Version(transformers_version) # Transformers moved rotary embeddings out of all attention layers diff --git a/unsloth/models/loader.py b/unsloth/models/loader.py index 356bca8a29..5d1896fd5b 100644 --- a/unsloth/models/loader.py +++ b/unsloth/models/loader.py @@ -45,6 +45,14 @@ except: pass from huggingface_hub import HfFileSystem import importlib.util +from ..device_type import ( + is_hip, + get_device_type, + DEVICE_TYPE, + DEVICE_TYPE_TORCH, + DEVICE_COUNT, + ALLOW_PREQUANTIZED_MODELS, +) # https://github.com/huggingface/transformers/pull/26037 allows 4 bit loading! from unsloth_zoo.utils import Version, _get_dtype @@ -195,6 +203,12 @@ class FastLanguageModel(FastLlamaModel): old_model_name = model_name if not use_exact_model_name: model_name = get_model_name(model_name, load_in_4bit) + # Check if pre-quantized models are allowed + # For eg AMD GPUs need blocksize = 128, but our pre-quants are blocksize = 64 + if not ALLOW_PREQUANTIZED_MODELS and model_name.endswith(("-unsloth-bnb-4bit", "-bnb-4bit")): + model_name = model_name.removesuffix("-unsloth-bnb-4bit") + model_name = model_name.removesuffix("-bnb-4bit") + pass if USE_MODELSCOPE and not os.path.exists(model_name): from modelscope import snapshot_download @@ -306,6 +320,12 @@ class FastLanguageModel(FastLlamaModel): model_name = peft_config.base_model_name_or_path if not use_exact_model_name: model_name = get_model_name(model_name, load_in_4bit) + # Check if pre-quantized models are allowed + # For eg AMD GPUs need blocksize = 128, but our pre-quants are blocksize = 64 + if not ALLOW_PREQUANTIZED_MODELS and model_name.endswith(("-unsloth-bnb-4bit", "-bnb-4bit")): + model_name = model_name.removesuffix("-unsloth-bnb-4bit") + model_name = model_name.removesuffix("-bnb-4bit") + pass model_config = AutoConfig.from_pretrained( model_name, token = token, @@ -618,6 +638,12 @@ class FastModel(FastBaseModel): old_model_name = model_name if not use_exact_model_name: model_name = get_model_name(model_name, load_in_4bit) + # Check if pre-quantized models are allowed + # For eg AMD GPUs need blocksize = 128, but our pre-quants are blocksize = 64 + if not ALLOW_PREQUANTIZED_MODELS and model_name.endswith(("-unsloth-bnb-4bit", "-bnb-4bit")): + model_name = model_name.removesuffix("-unsloth-bnb-4bit") + model_name = model_name.removesuffix("-bnb-4bit") + pass # Check modelscope if USE_MODELSCOPE and not os.path.exists(model_name): @@ -833,7 +859,12 @@ class FastModel(FastBaseModel): model_name = peft_config.base_model_name_or_path if not use_exact_model_name: model_name = get_model_name(model_name, load_in_4bit) - + # Check if pre-quantized models are allowed + # For eg AMD GPUs need blocksize = 128, but our pre-quants are blocksize = 64 + if not ALLOW_PREQUANTIZED_MODELS and model_name.endswith(("-unsloth-bnb-4bit", "-bnb-4bit")): + model_name = model_name.removesuffix("-unsloth-bnb-4bit") + model_name = model_name.removesuffix("-bnb-4bit") + pass model_config = AutoConfig.from_pretrained( model_name, token = token, diff --git a/unsloth/models/rl_replacements.py b/unsloth/models/rl_replacements.py index a207514e72..7ecffdf63d 100644 --- a/unsloth/models/rl_replacements.py +++ b/unsloth/models/rl_replacements.py @@ -26,7 +26,14 @@ import torch import inspect from collections import defaultdict from unsloth_zoo.rl_replacements import RL_REPLACEMENTS, left_pack_padding -from unsloth import DEVICE_TYPE +from ..device_type import ( + is_hip, + get_device_type, + DEVICE_TYPE, + DEVICE_TYPE_TORCH, + DEVICE_COUNT, + ALLOW_PREQUANTIZED_MODELS, +) import textwrap RL_EXTRA_ARGS = defaultdict(list) diff --git a/unsloth/models/vision.py b/unsloth/models/vision.py index b2704876e3..f2bd7c306b 100644 --- a/unsloth/models/vision.py +++ b/unsloth/models/vision.py @@ -71,7 +71,14 @@ except: # Old HF Hub versions <= 0.0.25 from huggingface_hub.utils._token import get_token pass -from unsloth import DEVICE_TYPE, DEVICE_COUNT, DEVICE_TYPE_TORCH +from ..device_type import ( + is_hip, + get_device_type, + DEVICE_TYPE, + DEVICE_TYPE_TORCH, + DEVICE_COUNT, + ALLOW_PREQUANTIZED_MODELS, +) __all__ = [ "FastBaseModel",