FEAT / Optim: Add GaLore optimizer (#29588)

* add galore v1

* add import

* add tests and doc

* fix doctest

* forward contrib credits from discussions

* forward contrib credits from discussions

* Apply suggestions from code review
Co-authored-by: Zach Mueller <muellerzr@gmail.com>

* fix failing tests'

* switch to `optim_target_modules` and clarify docs

* more clarification

* enhance lookup logic

* update a test to add peak memory

* add regex, all-linear and single string support

* add layer-wise optimization through DummyOptimizers and LRSchedulers

* forward contrib credits from discussions and original idea

* add a section about DDP not supported in layerwise

* Update src/transformers/trainer.py
Co-authored-by: Zach Mueller <muellerzr@gmail.com>

* fix self

* check only if layer_wise

* Update src/transformers/training_args.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* oops

* make use of intervals

* clarify comment

* add matching tests

* GaLoRe -> GaLore

* move to `get_scheduler`

* add note on docs

* add a warning

* adapt a bit the docs

* update docstring

* support original API

* Update docs/source/en/trainer.md

* slightly refactor

* Update docs/source/en/trainer.md
Co-authored-by: Matthew Douglas <38992547+matthewdouglas@users.noreply.github.com>

* Update src/transformers/training_args.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* fix args parsing and add tests

* remove warning for regex

* fix type hint

* add note about extra args

* make `is_regex` return optional

---------

Co-authored-by: Maxime <maximegmd @users.noreply.github.com>
Co-authored-by: Wing Lian <winglian @users.noreply.github.com>
Co-authored-by: Zach Mueller <muellerzr@gmail.com>
Co-authored-by: hiyouga <hiyouga@users.noreply.github.com>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>
Co-authored-by: Matthew Douglas <38992547+matthewdouglas@users.noreply.github.com>

docs/source/en/trainer.md

@@ -252,6 +252,136 @@ trainer = Trainer(..., args=training_args)
 
 NEFTune is disabled after training to restore the original embedding layer to avoid any unexpected behavior.
 
+## GaLore
+
+Gradient Low-Rank Projection (GaLore) is a memory-efficient low-rank training strategy that allows full-parameter learning but is more memory-efficient than common low-rank adaptation methods, such as LoRA.
+
+First make sure to install GaLore official repository:
+
+```bash
+pip install galore-torch
+```
+
+Then simply add one of `["galore_adamw", "galore_adafactor", "galore_adamw_8bit"]` in `optim` together with `optim_target_modules`, which can be a list of strings, regex or full path corresponding to the target module names you want to adapt. Below is an end-to-end example script (make sure to `pip install trl datasets`):
+
+```python
+import torch
+import datasets
+import trl
+
+from transformers import TrainingArguments, AutoConfig, AutoTokenizer, AutoModelForCausalLM
+
+train_dataset = datasets.load_dataset('imdb', split='train')
+
+args = TrainingArguments(
+    output_dir="./test-galore",
+    max_steps=100,
+    per_device_train_batch_size=2,
+    optim="galore_adamw",
+    optim_target_modules=["attn", "mlp"]
+)
+
+model_id = "google/gemma-2b"
+
+config = AutoConfig.from_pretrained(model_id)
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_config(config).to(0)
+
+trainer = trl.SFTTrainer(
+    model=model, 
+    args=args,
+    train_dataset=train_dataset,
+    dataset_text_field='text',
+    max_seq_length=512,
+)
+
+trainer.train()
+```
+
+To pass extra arguments supports by GaLore, you should pass correctly `optim_args`, for example:
+
+```python
+import torch
+import datasets
+import trl
+
+from transformers import TrainingArguments, AutoConfig, AutoTokenizer, AutoModelForCausalLM
+
+train_dataset = datasets.load_dataset('imdb', split='train')
+
+args = TrainingArguments(
+    output_dir="./test-galore",
+    max_steps=100,
+    per_device_train_batch_size=2,
+    optim="galore_adamw",
+    optim_target_modules=["attn", "mlp"],
+    optim_args="rank=64, update_proj_gap=100, scale=0.10",
+)
+
+model_id = "google/gemma-2b"
+
+config = AutoConfig.from_pretrained(model_id)
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_config(config).to(0)
+
+trainer = trl.SFTTrainer(
+    model=model, 
+    args=args,
+    train_dataset=train_dataset,
+    dataset_text_field='text',
+    max_seq_length=512,
+)
+
+trainer.train()
+```
+
+You can read more about the method in the [original repository](https://github.com/jiaweizzhao/GaLore) or the [paper](https://arxiv.org/abs/2403.03507).
+
+Currently you can only train Linear layers that are considered as GaLore layers and will use low-rank decomposition to be trained while remaining layers will be optimized in the conventional manner.
+
+Note it will take a bit of time before starting the training (~3 minutes for a 2B model on a NVIDIA A100), but training should go smoothly afterwards.
+
+You can also perform layer-wise optimization by post-pending the optimizer name with `layerwise` like below:
+
+```python
+import torch
+import datasets
+import trl
+
+from transformers import TrainingArguments, AutoConfig, AutoTokenizer, AutoModelForCausalLM
+
+train_dataset = datasets.load_dataset('imdb', split='train')
+
+args = TrainingArguments(
+    output_dir="./test-galore",
+    max_steps=100,
+    per_device_train_batch_size=2,
+    optim="galore_adamw_layerwise",
+    optim_target_modules=["attn", "mlp"]
+)
+
+model_id = "google/gemma-2b"
+
+config = AutoConfig.from_pretrained(model_id)
+
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_config(config).to(0)
+
+trainer = trl.SFTTrainer(
+    model=model, 
+    args=args,
+    train_dataset=train_dataset,
+    dataset_text_field='text',
+    max_seq_length=512,
+)
+
+trainer.train()
+```
+
+Note layerwise optimization is a bit experimental and does not support DDP (Distributed Data Parallel), thus you can run the training script only on a single GPU. Please see [this appropriate section](https://github.com/jiaweizzhao/GaLore?tab=readme-ov-file#train-7b-model-with-a-single-gpu-with-24gb-memory) for more details. Other features such as gradient clipping, DeepSpeed, etc might not be supported out of the box. Please [raise an issue on GitHub](https://github.com/huggingface/transformers/issues) if you encounter such issue.
+
 ## Accelerate and Trainer
 
 The [`Trainer`] class is powered by [Accelerate](https://hf.co/docs/accelerate), a library for easily training PyTorch models in distributed environments with support for integrations such as [FullyShardedDataParallel (FSDP)](https://pytorch.org/blog/introducing-pytorch-fully-sharded-data-parallel-api/) and [DeepSpeed](https://www.deepspeed.ai/).

src/transformers/optimization.py

@@ -24,6 +24,7 @@ from torch import nn
 from torch.optim import Optimizer
 from torch.optim.lr_scheduler import LambdaLR, ReduceLROnPlateau
 
+from .trainer_pt_utils import LayerWiseDummyOptimizer, LayerWiseDummyScheduler
 from .trainer_utils import SchedulerType
 from .utils import logging
 from .utils.versions import require_version
@@ -362,6 +363,32 @@ def get_scheduler(
     """
     name = SchedulerType(name)
     schedule_func = TYPE_TO_SCHEDULER_FUNCTION[name]
+
+    # If a `LayerWiseDummyOptimizer` is passed we extract the optimizer dict and
+    # recursively call `get_scheduler` to get the proper schedulers on each parameter
+    if optimizer is not None and isinstance(optimizer, LayerWiseDummyOptimizer):
+        optimizer_dict = optimizer.optimizer_dict
+        scheduler_dict = {}
+
+        for param in optimizer_dict.keys():
+            scheduler_dict[param] = get_scheduler(
+                name,
+                optimizer=optimizer_dict[param],
+                num_warmup_steps=num_warmup_steps,
+                num_training_steps=num_training_steps,
+            )
+
+        def scheduler_hook(param):
+            # Since the optimizer hook has been already attached we only need to
+            # attach the scheduler hook
+            if param.grad is not None:
+                scheduler_dict[param].step()
+
+        for param in optimizer_dict.keys():
+            param.register_post_accumulate_grad_hook(scheduler_hook)
+
+        return LayerWiseDummyScheduler()
+
     if name == SchedulerType.CONSTANT:
         return schedule_func(optimizer)
 

src/transformers/testing_utils.py

@@ -70,6 +70,7 @@ from .utils import (
     is_fsdp_available,
     is_ftfy_available,
     is_g2p_en_available,
+    is_galore_torch_available,
     is_ipex_available,
     is_jieba_available,
     is_jinja_available,
@@ -325,6 +326,14 @@ def require_bs4(test_case):
     return unittest.skipUnless(is_bs4_available(), "test requires BeautifulSoup4")(test_case)
 
 
+def require_galore_torch(test_case):
+    """
+    Decorator marking a test that requires GaLore. These tests are skipped when GaLore isn't installed.
+    https://github.com/jiaweizzhao/GaLore
+    """
+    return unittest.skipUnless(is_galore_torch_available(), "test requires GaLore")(test_case)
+
+
 def require_cv2(test_case):
     """
     Decorator marking a test that requires OpenCV.

src/transformers/trainer.py

@@ -83,6 +83,7 @@ from .trainer_pt_utils import (
     DistributedTensorGatherer,
     IterableDatasetShard,
     LabelSmoother,
+    LayerWiseDummyOptimizer,
     LengthGroupedSampler,
     SequentialDistributedSampler,
     distributed_broadcast_scalars,
@@ -111,6 +112,7 @@ from .trainer_utils import (
     RemoveColumnsCollator,
     TrainerMemoryTracker,
     TrainOutput,
+    check_target_module_exists,
     default_compute_objective,
     denumpify_detensorize,
     enable_full_determinism,
@@ -141,6 +143,7 @@ from .utils import (
     is_apex_available,
     is_bitsandbytes_available,
     is_datasets_available,
+    is_galore_torch_available,
     is_in_notebook,
     is_ipex_available,
     is_peft_available,
@@ -1010,7 +1013,17 @@ class Trainer:
                 },
             ]
 
-            optimizer_cls, optimizer_kwargs = Trainer.get_optimizer_cls_and_kwargs(self.args)
+            optimizer_cls, optimizer_kwargs = Trainer.get_optimizer_cls_and_kwargs(self.args, opt_model)
+
+            # Overwrite `params` in case it's created by `get_optimizer_cls_and_kwargs`
+            # e.g. for GaLore optimizer.
+            if "params" in optimizer_kwargs:
+                optimizer_grouped_parameters = optimizer_kwargs.pop("params")
+
+            # For layer-wise dummy optimizers we overwrite optimizer_grouped_parameters with `optimizer_dict`
+            # to avoid arguments conflicts.
+            if "optimizer_dict" in optimizer_kwargs:
+                optimizer_grouped_parameters = optimizer_kwargs.pop("optimizer_dict")
 
             self.optimizer = optimizer_cls(optimizer_grouped_parameters, **optimizer_kwargs)
             if optimizer_cls.__name__ == "Adam8bit":
@@ -1033,7 +1046,9 @@ class Trainer:
         return self.optimizer
 
     @staticmethod
-    def get_optimizer_cls_and_kwargs(args: TrainingArguments) -> Tuple[Any, Any]:
+    def get_optimizer_cls_and_kwargs(
+        args: TrainingArguments, model: Optional[PreTrainedModel] = None
+    ) -> Tuple[Any, Any]:
         """
         Returns the optimizer class and optimizer parameters based on the training arguments.
 
@@ -1171,6 +1186,132 @@ class Trainer:
             optimizer_cls = torch.optim.Adagrad
         elif args.optim == OptimizerNames.RMSPROP:
             optimizer_cls = torch.optim.RMSprop
+        elif args.optim in [
+            OptimizerNames.GALORE_ADAMW,
+            OptimizerNames.GALORE_ADAMW_8BIT,
+            OptimizerNames.GALORE_ADAFACTOR,
+            OptimizerNames.GALORE_ADAMW_LAYERWISE,
+            OptimizerNames.GALORE_ADAMW_8BIT_LAYERWISE,
+            OptimizerNames.GALORE_ADAFACTOR_LAYERWISE,
+        ]:
+            if not is_galore_torch_available():
+                raise ImportError(
+                    "You need to install `galore_torch` in order to use GaLore optimizers"
+                    " install it with `pip install git+https://github.com/jiaweizzhao/GaLore`"
+                )
+            from galore_torch import GaLoreAdafactor, GaLoreAdamW, GaLoreAdamW8bit
+
+            is_layerwise = args.optim.lower().endswith("layerwise")
+            if is_layerwise and args.parallel_mode == ParallelMode.DISTRIBUTED:
+                raise NotImplementedError("Layer-wise GaLore does not support DDP at this time")
+
+            optimizer_mapping = {
+                OptimizerNames.GALORE_ADAMW: GaLoreAdamW,
+                OptimizerNames.GALORE_ADAMW_8BIT: GaLoreAdamW8bit,
+                OptimizerNames.GALORE_ADAFACTOR: GaLoreAdafactor,
+                OptimizerNames.GALORE_ADAMW_LAYERWISE: GaLoreAdamW,
+                OptimizerNames.GALORE_ADAMW_8BIT_LAYERWISE: GaLoreAdamW8bit,
+                OptimizerNames.GALORE_ADAFACTOR_LAYERWISE: GaLoreAdafactor,
+            }
+
+            optimizer_cls = optimizer_mapping[args.optim]
+
+            if args.optim_target_modules is None:
+                raise ValueError(
+                    "You need to define a `optim_target_modules` in order to properly use GaLore optimizers"
+                )
+
+            if not isinstance(args.optim_target_modules, (list, str)):
+                raise ValueError(
+                    f"`optim_target_modules` has to be a list of strings, a string corresponding to a regex, or a specific module or 'all-linear', you passed {args.optim_target_modules}"
+                )
+
+            if model is None:
+                raise ValueError("You need to pass a model in order to correctly initialize a GaLore optimizer.")
+
+            logger.warning(
+                "Activated GaLoRE fine-tuning, depending on your model size and hardware, the training might take a while before starting. Please be patient !"
+            )
+
+            all_linear = (
+                isinstance(args.optim_target_modules, str)
+                and args.optim_target_modules.replace("_", "-") == "all-linear"
+            )
+
+            galore_params = []
+            galore_params_names = []
+            for module_name, module in model.named_modules():
+                target_module_exists, is_regex = check_target_module_exists(
+                    args.optim_target_modules, module_name, return_is_regex=True
+                )
+
+                if not isinstance(module, nn.Linear):
+                    # Warn in case we match but it's not a linear layer
+                    if target_module_exists and not is_regex:
+                        logger.warning(
+                            f"{module_name} has been matched but ignored as GaLore only supports linear layers. Please double check your `optim_target_modules`!"
+                        )
+
+                    continue
+
+                if not target_module_exists and not all_linear:
+                    continue
+
+                galore_params.append(module.weight)
+                galore_params_names.append(module_name + ".weight")
+
+            if len(galore_params) == 0:
+                raise ValueError(
+                    f"None of the target modules were found! ({args.optim_target_modules}). Please make sure to pass a valid `target_modules`."
+                )
+
+            non_galore_params = [p for n, p in model.named_parameters() if n not in galore_params_names]
+
+            galore_optim_kwargs = {
+                "rank": int(optim_args.pop("rank", 128)),
+                "update_proj_gap": int(optim_args.pop("update_proj_gap", 200)),
+                "scale": float(optim_args.pop("scale", 0.25)),
+                "proj_type": optim_args.pop("proj_type", "std"),
+            }
+
+            # The default args are from the official repository: https://github.com/jiaweizzhao/GaLore
+            param_groups = [
+                {"params": non_galore_params},
+                {"params": galore_params, **galore_optim_kwargs},
+            ]
+
+            if is_layerwise:
+                # For layer-wise optimizers, the optimization step is done through post accumulation
+                # gradient hooks. The trick is to first attach these hooks to the model parameters then
+                # create a dummy optimizer that will perform no-ops in the Trainer.
+                # See the original implementation or the nice implementation from @hiyouga
+                # here: https://github.com/hiyouga/LLaMA-Factory/commit/8664262cde3919e10eaecbd66e8c5d356856362e#diff-ebe08ab14496dfb9e06075f0fdd36799ef6d1535cc4dd4715b74c4e3e06fe3ba
+                if args.gradient_accumulation_steps != 1:
+                    raise ValueError("Layerwise GaLoRE optimizer do not support gradient accumulation !")
+
+                optimizer_dict = {}
+                for param in non_galore_params:
+                    param_groups = [{"params": [param]}]
+                    optimizer_dict[param] = optimizer_cls(param_groups, **optimizer_kwargs)
+                for param in galore_params:
+                    param_groups = [{"params": [param], **galore_optim_kwargs}]
+                    optimizer_dict[param] = optimizer_cls(param_groups, **optimizer_kwargs)
+
+                def optimizer_hook(param):
+                    if param.grad is not None:
+                        optimizer_dict[param].step()
+                        optimizer_dict[param].zero_grad()
+
+                for param in model.parameters():
+                    param.register_post_accumulate_grad_hook(optimizer_hook)
+
+                optimizer_cls = LayerWiseDummyOptimizer
+                optimizer_kwargs.update({"optimizer_dict": optimizer_dict})
+
+            optimizer_kwargs.update({"params": param_groups})
+
+            if args.optim == OptimizerNames.GALORE_ADAFACTOR:
+                optimizer_kwargs.update({"scale_parameter": False, "relative_step": False})
         else:
             raise ValueError(f"Trainer cannot instantiate unsupported optimizer: {args.optim}")
         return optimizer_cls, optimizer_kwargs

src/transformers/trainer_pt_utils.py

@@ -34,6 +34,7 @@ import numpy as np
 import torch
 import torch.distributed as dist
 from torch import nn
+from torch.optim.lr_scheduler import LRScheduler
 from torch.utils.data import Dataset, IterableDataset, RandomSampler, Sampler
 from torch.utils.data.distributed import DistributedSampler
 
@@ -1226,3 +1227,47 @@ class AcceleratorConfig:
 
     def to_dict(self):
         return copy.deepcopy(self.__dict__)
+
+
+class LayerWiseDummyOptimizer(torch.optim.Optimizer):
+    """
+    For Layer-wise optimizers such as GaLoRE optimizer, the optimization
+    step is already done through the post gradient hooks. Therefore
+    the trick is to create a dummy optimizer that can take arbitrary
+    args and kwargs and return a no-op during training.
+
+    Initial idea from @hiyouga in LLaMA-Factory:
+    https://github.com/hiyouga/LLaMA-Factory/commit/8664262cde3919e10eaecbd66e8c5d356856362e#diff-ebe08ab14496dfb9e06075f0fdd36799ef6d1535cc4dd4715b74c4e3e06fe3ba
+    """
+
+    def __init__(self, optimizer_dict=None, *args, **kwargs):
+        dummy_tensor = torch.randn(1, 1)
+        self.optimizer_dict = optimizer_dict
+        super().__init__([dummy_tensor], {"lr": 1e-03})
+
+    def zero_grad(self, set_to_none: bool = True) -> None:
+        pass
+
+    def step(self, closure=None) -> Optional[float]:
+        pass
+
+
+class LayerWiseDummyScheduler(LRScheduler):
+    """
+    For Layer-wise optimizers such as GaLoRE optimizer, the optimization and scheduling step
+    are already done through the post gradient hooks. Therefore
+    the trick is to create a dummy scheduler that can take arbitrary
+    args and kwargs and return a no-op during training.
+    """
+
+    def __init__(self, *args, **kwargs):
+        optimizer = LayerWiseDummyOptimizer()
+        last_epoch = -1
+        verbose = False
+        super().__init__(optimizer, last_epoch, verbose)
+
+    def get_lr(self):
+        return [group["lr"] for group in self.optimizer.param_groups]
+
+    def _get_closed_form_lr(self):
+        return self.base_lrs

src/transformers/trainer_utils.py

@@ -785,3 +785,42 @@ class RemoveColumnsCollator:
     def __call__(self, features: List[dict]):
         features = [self._remove_columns(feature) for feature in features]
         return self.data_collator(features)
+
+
+def check_target_module_exists(optim_target_modules, key: str, return_is_regex: bool = False):
+    """A helper method to check if the passed module's key name matches any of the target modules in the optim_target_modules.
+
+    Args:
+        optim_target_modules (`Union[str, List[str]]`):
+            A list of strings to try to match. Can be also a full string.
+        key (`str`):
+            A key to search any matches in optim_target_modules
+        return_is_regex (`bool`):
+            If set to `True`, the method will return whether the passed `optim_target_modules`
+            is a regex or not.
+
+    Returns:
+        `bool` : True of match object if key matches any target modules from config, False or
+        None if no match found
+        `bool` : If the matched target module is a regex to silence out the warnings in Trainer
+        for extra modules being found (only if `target_module_found=True` for an array of regex).
+    """
+    target_module_found = False
+    is_regex = False
+
+    if isinstance(optim_target_modules, str):
+        target_module_found = bool(re.fullmatch(optim_target_modules, key))
+        is_regex = True if not optim_target_modules == key else False
+    elif key in optim_target_modules:  # from here, target_module_found must be a list of str
+        # this module is specified directly in target_modules
+        target_module_found = True
+    elif any(target_key in key for target_key in optim_target_modules):
+        target_module_found = True
+    elif any(bool(re.fullmatch(optim_target_module, key)) for optim_target_module in optim_target_modules):
+        target_module_found = True
+        is_regex = True
+
+    if return_is_regex:
+        return target_module_found, is_regex
+
+    return target_module_found

src/transformers/training_args.py

@@ -164,6 +164,12 @@ class OptimizerNames(ExplicitEnum):
     RMSPROP_BNB = "rmsprop_bnb"
     RMSPROP_8BIT = "rmsprop_bnb_8bit"
     RMSPROP_32BIT = "rmsprop_bnb_32bit"
+    GALORE_ADAMW = "galore_adamw"
+    GALORE_ADAMW_8BIT = "galore_adamw_8bit"
+    GALORE_ADAFACTOR = "galore_adafactor"
+    GALORE_ADAMW_LAYERWISE = "galore_adamw_layerwise"
+    GALORE_ADAMW_8BIT_LAYERWISE = "galore_adamw_8bit_layerwise"
+    GALORE_ADAFACTOR_LAYERWISE = "galore_adafactor_layerwise"
 
 
 # TODO: `TrainingArguments` users rely on it being fully mutable. In the future see if we can narrow this to a few keys: https://github.com/huggingface/transformers/pull/25903
@@ -696,6 +702,12 @@ class TrainingArguments:
             for instruction fine-tuning. Check out the [original paper](https://arxiv.org/abs/2310.05914) and the
             [original code](https://github.com/neelsjain/NEFTune). Support transformers `PreTrainedModel` and also
             `PeftModel` from peft.
+        optim_target_modules (`Union[str, List[str]]`, *optional*):
+            The target modules to optimize, i.e. the module names that you would like to train, right now this is used only for GaLore algorithm
+            https://arxiv.org/abs/2403.03507
+            See: https://github.com/jiaweizzhao/GaLore for more details. You need to make sure to pass a valid GaloRe
+            optimizer, e.g. one of: "galore_adamw", "galore_adamw_8bit", "galore_adafactor" and make sure that the target modules are `nn.Linear` modules
+            only.
     """
 
     framework = "pt"
@@ -1354,6 +1366,13 @@ class TrainingArguments:
         },
     )
 
+    optim_target_modules: Union[None, str, List[str]] = field(
+        default=None,
+        metadata={
+            "help": "Target modules for the optimizer defined in the `optim` argument. Only used for the GaLore optimizer at the moment."
+        },
+    )
+
     def __post_init__(self):
         # expand paths, if not os.makedirs("~/bar") will make directory
         # in the current directory instead of the actual home

src/transformers/utils/__init__.py

@@ -125,6 +125,7 @@ from .import_utils import (
     is_fsdp_available,
     is_ftfy_available,
     is_g2p_en_available,
+    is_galore_torch_available,
     is_in_notebook,
     is_ipex_available,
     is_jieba_available,

src/transformers/utils/import_utils.py

@@ -95,6 +95,7 @@ _accelerate_available, _accelerate_version = _is_package_available("accelerate",
 _apex_available = _is_package_available("apex")
 _aqlm_available = _is_package_available("aqlm")
 _bitsandbytes_available = _is_package_available("bitsandbytes")
+_galore_torch_available = _is_package_available("galore_torch")
 # `importlib.metadata.version` doesn't work with `bs4` but `beautifulsoup4`. For `importlib.util.find_spec`, reversed.
 _bs4_available = importlib.util.find_spec("bs4") is not None
 _coloredlogs_available = _is_package_available("coloredlogs")
@@ -309,6 +310,10 @@ def is_torchvision_available():
     return _torchvision_available
 
 
+def is_galore_torch_available():
+    return _galore_torch_available
+
+
 def is_pyctcdecode_available():
     return _pyctcdecode_available
 

tests/trainer/test_trainer.py

@@ -60,6 +60,7 @@ from transformers.testing_utils import (
     require_accelerate,
     require_bitsandbytes,
     require_deepspeed,
+    require_galore_torch,
     require_intel_extension_for_pytorch,
     require_optuna,
     require_peft,
@@ -84,7 +85,7 @@ from transformers.testing_utils import (
     slow,
     torch_device,
 )
-from transformers.trainer_utils import PREFIX_CHECKPOINT_DIR, HPSearchBackend
+from transformers.trainer_utils import PREFIX_CHECKPOINT_DIR, HPSearchBackend, check_target_module_exists
 from transformers.training_args import OptimizerNames
 from transformers.utils import (
     SAFE_WEIGHTS_INDEX_NAME,
@@ -114,6 +115,8 @@ if is_torch_available():
         GPT2Config,
         GPT2LMHeadModel,
         LineByLineTextDataset,
+        LlamaConfig,
+        LlamaForCausalLM,
         PreTrainedModel,
         Trainer,
         TrainerState,
@@ -146,6 +149,31 @@ class RegressionDataset:
         return result
 
 
+# Converting Bytes to Megabytes
+def bytes2megabytes(x):
+    return int(x / 2**20)
+
+
+# Copied from acclerate: https://github.com/huggingface/accelerate/blob/ee163b66fb7848892519e804688cb4ae981aacbe/src/accelerate/test_utils/scripts/external_deps/test_peak_memory_usage.py#L40C1-L73C68
+class TorchTracemalloc:
+    def __enter__(self):
+        gc.collect()
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+            torch.cuda.reset_max_memory_allocated()  # reset the peak gauge to zero
+            self.begin = torch.cuda.memory_allocated()
+        return self
+
+    def __exit__(self, *exc):
+        gc.collect()
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+            self.end = torch.cuda.memory_allocated()
+            self.peak = torch.cuda.max_memory_allocated()
+        self.used = bytes2megabytes(self.end - self.begin)
+        self.peaked = bytes2megabytes(self.peak - self.begin)
+
+
 @dataclasses.dataclass
 class RegressionTrainingArguments(TrainingArguments):
     a: float = 0.0
@@ -1069,6 +1097,293 @@ class TrainerIntegrationTest(TestCasePlus, TrainerIntegrationCommon):
         trainer.train()
         trainer.evaluate()
 
+    def test_galore_matched_modules(self):
+        regex_patterns = [r".*.attn.*", r".*.mlp.*"]
+
+        module_names = [
+            "model.transformer.h.0.ln_1",
+            "model.transformer.h.0.attn.q_proj",
+            "model.lm_head",
+            "model.transformer.h.0.mlp.up_proj",
+        ]
+        expected_values = [False, True, False, True]
+
+        for expected_value, module_name in zip(expected_values, module_names):
+            is_module_matched, is_regex = check_target_module_exists(regex_patterns, module_name, return_is_regex=True)
+            self.assertTrue(is_module_matched == expected_value)
+            if is_module_matched:
+                self.assertTrue(is_regex)
+
+        exact_patterns = ["q_proj", "up_proj"]
+
+        module_names = [
+            "model.transformer.h.0.ln_1",
+            "model.transformer.h.0.attn.q_proj",
+            "model.lm_head",
+            "model.transformer.h.0.mlp.up_proj",
+        ]
+        expected_values = [False, True, False, True]
+
+        for expected_value, module_name in zip(expected_values, module_names):
+            is_module_matched, is_regex = check_target_module_exists(exact_patterns, module_name, return_is_regex=True)
+            self.assertTrue(is_module_matched == expected_value)
+            if is_module_matched:
+                self.assertFalse(is_regex)
+
+        simple_regex = r".*.attn.*"
+
+        module_names = [
+            "model.transformer.h.0.ln_1",
+            "model.transformer.h.0.attn.q_proj",
+            "model.lm_head",
+            "model.transformer.h.0.mlp.up_proj",
+        ]
+        expected_values = [False, True, False, False]
+
+        for expected_value, module_name in zip(expected_values, module_names):
+            is_module_matched, is_regex = check_target_module_exists(simple_regex, module_name, return_is_regex=True)
+            self.assertTrue(is_module_matched == expected_value)
+            if is_module_matched:
+                self.assertTrue(is_regex)
+
+        simple_regex = "model.transformer.h.0.attn.q_proj"
+
+        module_names = [
+            "model.transformer.h.0.ln_1",
+            "model.transformer.h.0.attn.q_proj",
+            "model.lm_head",
+            "model.transformer.h.0.mlp.up_proj",
+        ]
+        expected_values = [False, True, False, False]
+
+        for expected_value, module_name in zip(expected_values, module_names):
+            is_module_matched, is_regex = check_target_module_exists(simple_regex, module_name, return_is_regex=True)
+            self.assertTrue(is_module_matched == expected_value)
+            if is_module_matched:
+                self.assertFalse(is_regex)
+
+        target_modules = ["attn", "mlp"]
+
+        module_names = [
+            "model.transformer.h.0.ln_1",
+            "model.transformer.h.0.attn.q_proj",
+            "model.lm_head",
+            "model.transformer.h.0.mlp.up_proj",
+        ]
+        expected_values = [False, True, False, True]
+
+        for expected_value, module_name in zip(expected_values, module_names):
+            is_module_matched, is_regex = check_target_module_exists(target_modules, module_name, return_is_regex=True)
+            self.assertTrue(is_module_matched == expected_value)
+            if is_module_matched:
+                self.assertFalse(is_regex)
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore(self):
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adamw",
+                optim_target_modules=[r".*attn.*", r".*mlp.*"],
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore_extra_args(self):
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adamw",
+                optim_args="rank=64, update_proj_gap=100, scale=0.10",
+                optim_target_modules=[r".*attn.*", r".*mlp.*"],
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore_layerwise(self):
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adamw_layerwise",
+                optim_target_modules=[r".*attn.*", r".*mlp.*"],
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore_layerwise_with_scheduler(self):
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adamw_layerwise",
+                lr_scheduler_type="cosine",
+                optim_target_modules=[r".*attn.*", r".*mlp.*"],
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore_adamw_8bit(self):
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adamw_8bit",
+                optim_target_modules=[r".*attn.*", r".*mlp.*"],
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore_adafactor(self):
+        # These are the intervals of the peak memory usage of training such a tiny model
+        # if the peak memory goes outside that range, then we know there might be a bug somewhere
+        upper_bound_pm = 700
+        lower_bound_pm = 650
+
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir, TorchTracemalloc() as tracemalloc:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adafactor",
+                optim_target_modules=[r".*attn.*", r".*mlp.*"],
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+        galore_peak_memory = tracemalloc.peaked + bytes2megabytes(tracemalloc.begin)
+
+        self.assertTrue(galore_peak_memory < upper_bound_pm)
+        self.assertTrue(lower_bound_pm < galore_peak_memory)
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore_adafactor_attention_only(self):
+        # These are the intervals of the peak memory usage of training such a tiny model
+        # if the peak memory goes outside that range, then we know there might be a bug somewhere
+        upper_bound_pm = 700
+        lower_bound_pm = 650
+
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir, TorchTracemalloc() as tracemalloc:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adafactor",
+                optim_target_modules=["q_proj", "k_proj", "v_proj"],
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+        galore_peak_memory = tracemalloc.peaked + bytes2megabytes(tracemalloc.begin)
+        self.assertTrue(galore_peak_memory < upper_bound_pm)
+        self.assertTrue(lower_bound_pm < galore_peak_memory)
+
+    @require_galore_torch
+    @require_torch_gpu
+    def test_galore_adafactor_all_linear(self):
+        # These are the intervals of the peak memory usage of training such a tiny model
+        # if the peak memory goes outside that range, then we know there might be a bug somewhere
+        upper_bound_pm = 700
+        lower_bound_pm = 650
+
+        config = LlamaConfig(vocab_size=100, hidden_size=32, num_hidden_layers=3, num_attention_heads=4)
+        tiny_llama = LlamaForCausalLM(config)
+        x = torch.randint(0, 100, (128,))
+        train_dataset = RepeatDataset(x)
+
+        with tempfile.TemporaryDirectory() as tmpdir, TorchTracemalloc() as tracemalloc:
+            # Trainer without inf/nan filter
+            args = TrainingArguments(
+                tmpdir,
+                learning_rate=1e-9,
+                logging_steps=5,
+                optim="galore_adafactor",
+                optim_target_modules="all-linear",
+            )
+            trainer = Trainer(tiny_llama, args, train_dataset=train_dataset)
+
+            # Check this works
+            _ = trainer.train()
+
+        galore_peak_memory = tracemalloc.peaked + bytes2megabytes(tracemalloc.begin)
+        self.assertTrue(galore_peak_memory < upper_bound_pm)
+        self.assertTrue(lower_bound_pm < galore_peak_memory)
+
     @require_torch_multi_accelerator
     def test_data_is_not_parallelized_when_model_is_parallel(self):
         model = RegressionModel()