ADOPT optimizer integration (#2032) [skip ci]

* adopt integration

* stuff

* doc and test for ADOPT

* rearrangement

* fixed formatting

* hacking pre-commit

* chore: lint

* update module doc for adopt optimizer

* remove un-necessary example yaml for adopt optimizer

* skip test adopt if torch<2.5.1

* formatting

* use version.parse

* specifies required torch version for adopt_adamw

---------

Co-authored-by: sunny <sunnyliu19981005@gmail.com>
Co-authored-by: Wing Lian <wing@axolotl.ai>

docs/config.qmd

@@ -405,6 +405,7 @@ lr_div_factor: # Learning rate div factor
 # - adamw_torch_fused
 # - adamw_torch_xla
 # - adamw_apex_fused
+# - adopt_adamw (only for torch version >= 2.5.1)
 # - adafactor
 # - adamw_anyprecision
 # - sgd

src/axolotl/core/trainer_builder.py

@@ -436,7 +436,13 @@ class AxolotlTrainer(SchedulerMixin, Trainer):
         if (
             self.args.loraplus_lr_ratio is None
             and self.args.alternate_optimizer
-            not in ["optimi_adamw", "ao_adamw_8bit", "ao_adamw_4bit", "ao_adamw_fp8"]
+            not in [
+                "optimi_adamw",
+                "ao_adamw_8bit",
+                "ao_adamw_4bit",
+                "ao_adamw_fp8",
+                "adopt_adamw",
+            ]
         ):
             return super().create_optimizer()
 
@@ -505,6 +511,14 @@ class AxolotlTrainer(SchedulerMixin, Trainer):
                 self.optimizer = (  # pylint: disable=attribute-defined-outside-init
                     AdamWFp8(optimizer_grouped_parameters, **optimizer_kwargs)
                 )
+            elif self.args.alternate_optimizer == "adopt_adamw":
+                from axolotl.utils.optimizers.adopt import ADOPT
+
+                self.optimizer = (  # pylint: disable=attribute-defined-outside-init
+                    ADOPT(
+                        optimizer_grouped_parameters, decoupled=True, **optimizer_kwargs
+                    )
+                )
 
         if is_sagemaker_mp_enabled():
             self.optimizer = smp.DistributedOptimizer(  # pylint: disable=attribute-defined-outside-init
@@ -1625,11 +1639,13 @@ class HFCausalTrainerBuilder(TrainerBuilderBase):
         if self.cfg.reward_model:
             trainer_kwargs["max_length"] = self.cfg.sequence_len
 
+        # pylint: disable=duplicate-code
         if self.cfg.optimizer in [
             "optimi_adamw",
             "ao_adamw_4bit",
             "ao_adamw_8bit",
             "ao_adamw_fp8",
+            "adopt_adamw",
         ]:
             # Set default so transformers doesn't throw
             training_arguments_kwargs["optim"] = "adamw_hf"

src/axolotl/utils/config/models/input/v0_4_1/__init__.py

@@ -428,6 +428,7 @@ class HyperparametersConfig(BaseModel):
                 "ao_adamw_4bit",
                 "ao_adamw_8bit",
                 "ao_adamw_fp8",
+                "adopt_adamw",
             ],
         ]
     ] = OptimizerNames.ADAMW_HF.value

src/axolotl/utils/optimizers/adopt.py

@@ -0,0 +1,508 @@
+"""
+Copied from https://github.com/iShohei220/adopt
+
+ADOPT: Modified Adam Can Converge with Any β2 with the Optimal Rate (2024)
+Taniguchi, Shohei and Harada, Keno and Minegishi, Gouki and Oshima, Yuta and Jeong, Seong Cheol and Nagahara, Go and Iiyama, Tomoshi and Suzuki, Masahiro and Iwasawa, Yusuke and Matsuo, Yutaka
+"""
+# mypy: ignore-errors
+# pylint: skip-file
+# mypy: allow-untyped-decorators
+# mypy: allow-untyped-defs
+from typing import List, Optional, Tuple, Union, cast
+
+import torch
+from torch import Tensor
+from torch.optim.optimizer import (
+    Optimizer,
+    ParamsT,
+    _default_to_fused_or_foreach,
+    _device_dtype_check_for_fused,
+    _disable_dynamo_if_unsupported,
+    _get_capturable_supported_devices,
+    _get_scalar_dtype,
+    _get_value,
+    _use_grad_for_differentiable,
+    _view_as_real,
+)
+
+__all__ = ["ADOPT", "adopt"]
+
+
+class ADOPT(Optimizer):
+    def __init__(
+        self,
+        params: ParamsT,
+        lr: Union[float, Tensor] = 1e-3,
+        betas: Tuple[float, float] = (0.9, 0.9999),
+        eps: float = 1e-6,
+        weight_decay: float = 0.0,
+        decoupled: bool = False,
+        *,
+        foreach: Optional[bool] = None,
+        maximize: bool = False,
+        capturable: bool = False,
+        differentiable: bool = False,
+        fused: Optional[bool] = None,
+    ):
+        if isinstance(lr, Tensor):
+            if foreach and not capturable:
+                raise ValueError(
+                    "lr as a Tensor is not supported for capturable=False and foreach=True"
+                )
+            if lr.numel() != 1:
+                raise ValueError("Tensor lr must be 1-element")
+        if not 0.0 <= lr:
+            raise ValueError(f"Invalid learning rate: {lr}")
+        if not 0.0 <= eps:
+            raise ValueError(f"Invalid epsilon value: {eps}")
+        if not 0.0 <= betas[0] < 1.0:
+            raise ValueError(f"Invalid beta parameter at index 0: {betas[0]}")
+        if not 0.0 <= betas[1] < 1.0:
+            raise ValueError(f"Invalid beta parameter at index 1: {betas[1]}")
+        if not 0.0 <= weight_decay:
+            raise ValueError(f"Invalid weight_decay value: {weight_decay}")
+
+        defaults = dict(
+            lr=lr,
+            betas=betas,
+            eps=eps,
+            weight_decay=weight_decay,
+            decoupled=decoupled,
+            maximize=maximize,
+            foreach=foreach,
+            capturable=capturable,
+            differentiable=differentiable,
+            fused=fused,
+        )
+        super().__init__(params, defaults)
+
+        if fused:
+            # TODO: support fused
+            raise RuntimeError("`fused` is not currently supported")
+
+            if differentiable:
+                raise RuntimeError("`fused` does not support `differentiable`")
+            self._step_supports_amp_scaling = True
+            # TODO(crcrpar): [low prec params & their higher prec copy]
+            # Support AMP with FP16/BF16 model params which would need
+            # higher prec copy of params to do update math in higher prec to
+            # alleviate the loss of information.
+            if foreach:
+                raise RuntimeError("`fused` and `foreach` cannot be `True` together.")
+
+    def __setstate__(self, state):
+        super().__setstate__(state)
+        for group in self.param_groups:
+            group.setdefault("maximize", False)
+            group.setdefault("foreach", None)
+            group.setdefault("capturable", False)
+            group.setdefault("differentiable", False)
+            fused = group.setdefault("fused", None)
+            for p in group["params"]:
+                p_state = self.state.get(p, [])
+                if len(p_state) != 0 and not torch.is_tensor(p_state["step"]):
+                    step_val = float(p_state["step"])
+                    p_state["step"] = (
+                        torch.tensor(
+                            step_val,
+                            dtype=_get_scalar_dtype(is_fused=fused),
+                            device=p.device,
+                        )
+                        if group["capturable"] or group["fused"]
+                        else torch.tensor(step_val, dtype=_get_scalar_dtype())
+                    )
+
+    def _init_group(
+        self,
+        group,
+        params_with_grad,
+        grads,
+        exp_avgs,
+        exp_avg_sqs,
+        state_steps,
+    ):
+        has_complex = False
+        for p in group["params"]:
+            if p.grad is not None:
+                has_complex |= torch.is_complex(p)
+                params_with_grad.append(p)
+                if p.grad.is_sparse:
+                    raise RuntimeError("ADOPT does not support sparse gradients")
+                grads.append(p.grad)
+
+                state = self.state[p]
+                # Lazy state initialization
+                if len(state) == 0:
+                    if group["fused"]:
+                        _device_dtype_check_for_fused(p)
+                    # note(crcrpar): [special device hosting for step]
+                    # Deliberately host `step` on CPU if both capturable and fused are off.
+                    # This is because kernel launches are costly on CUDA and XLA.
+                    state["step"] = (
+                        torch.zeros(
+                            (),
+                            dtype=_get_scalar_dtype(is_fused=group["fused"]),
+                            device=p.device,
+                        )
+                        if group["capturable"] or group["fused"]
+                        else torch.tensor(0.0, dtype=_get_scalar_dtype())
+                    )
+                    # Exponential moving average of gradient values
+                    state["exp_avg"] = torch.zeros_like(
+                        p, memory_format=torch.preserve_format
+                    )
+                    # Exponential moving average of squared gradient values
+                    state["exp_avg_sq"] = torch.zeros_like(
+                        p, memory_format=torch.preserve_format
+                    )
+
+                exp_avgs.append(state["exp_avg"])
+                exp_avg_sqs.append(state["exp_avg_sq"])
+
+                if group["differentiable"] and state["step"].requires_grad:
+                    raise RuntimeError(
+                        "`requires_grad` is not supported for `step` in differentiable mode"
+                    )
+
+                # Foreach without capturable does not support a tensor lr
+                if (
+                    group["foreach"]
+                    and torch.is_tensor(group["lr"])
+                    and not group["capturable"]
+                ):
+                    raise RuntimeError(
+                        "lr as a Tensor is not supported for capturable=False and foreach=True"
+                    )
+
+                state_steps.append(state["step"])
+        return has_complex
+
+    @_use_grad_for_differentiable
+    def step(self, closure=None):
+        """Perform a single optimization step.
+
+        Args:
+            closure (Callable, optional): A closure that reevaluates the model
+                and returns the loss.
+        """
+        self._cuda_graph_capture_health_check()
+
+        loss = None
+        if closure is not None:
+            with torch.enable_grad():
+                loss = closure()
+
+        for group in self.param_groups:
+            params_with_grad: List[Tensor] = []
+            grads: List[Tensor] = []
+            exp_avgs: List[Tensor] = []
+            exp_avg_sqs: List[Tensor] = []
+            state_steps: List[Tensor] = []
+            beta1, beta2 = group["betas"]
+
+            has_complex = self._init_group(
+                group,
+                params_with_grad,
+                grads,
+                exp_avgs,
+                exp_avg_sqs,
+                state_steps,
+            )
+
+            adopt(
+                params_with_grad,
+                grads,
+                exp_avgs,
+                exp_avg_sqs,
+                state_steps,
+                has_complex=has_complex,
+                beta1=beta1,
+                beta2=beta2,
+                lr=group["lr"],
+                weight_decay=group["weight_decay"],
+                decoupled=group["decoupled"],
+                eps=group["eps"],
+                maximize=group["maximize"],
+                foreach=group["foreach"],
+                capturable=group["capturable"],
+                differentiable=group["differentiable"],
+                fused=group["fused"],
+                grad_scale=getattr(self, "grad_scale", None),
+                found_inf=getattr(self, "found_inf", None),
+            )
+
+        return loss
+
+
+def _single_tensor_adopt(
+    params: List[Tensor],
+    grads: List[Tensor],
+    exp_avgs: List[Tensor],
+    exp_avg_sqs: List[Tensor],
+    state_steps: List[Tensor],
+    grad_scale: Optional[Tensor],
+    found_inf: Optional[Tensor],
+    *,
+    has_complex: bool,
+    beta1: float,
+    beta2: float,
+    lr: Union[float, Tensor],
+    weight_decay: float,
+    decoupled: bool,
+    eps: float,
+    maximize: bool,
+    capturable: bool,
+    differentiable: bool,
+):
+    assert grad_scale is None and found_inf is None
+
+    if torch.jit.is_scripting():
+        # this assert is due to JIT being dumb and not realizing that the ops below
+        # have overloads to handle both float and Tensor lrs, so we just assert it's
+        # a float since most people using JIT are using floats
+        assert isinstance(lr, float)
+
+    for i, param in enumerate(params):
+        grad = grads[i] if not maximize else -grads[i]
+        exp_avg = exp_avgs[i]
+        exp_avg_sq = exp_avg_sqs[i]
+        step_t = state_steps[i]
+
+        # If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
+        if not torch._utils.is_compiling() and capturable:
+            capturable_supported_devices = _get_capturable_supported_devices()
+            assert (
+                param.device.type == step_t.device.type
+                and param.device.type in capturable_supported_devices
+            ), f"If capturable=True, params and state_steps must be on supported devices: {capturable_supported_devices}."
+
+        # update step
+        step_t += 1
+
+        if weight_decay != 0:
+            if decoupled:
+                param.add_(param, alpha=-lr * weight_decay)
+            else:
+                grad = grad.add(param, alpha=weight_decay)
+
+        if torch.is_complex(param):
+            grad = torch.view_as_real(grad)
+            if exp_avg is not None:
+                exp_avg = torch.view_as_real(exp_avg)
+            if exp_avg_sq is not None:
+                exp_avg_sq = torch.view_as_real(exp_avg_sq)
+            param = torch.view_as_real(param)
+
+        step = step_t if capturable or differentiable else _get_value(step_t)
+        if step == 1:
+            exp_avg_sq.addcmul_(grad, grad.conj())
+            continue
+
+        denom = torch.clamp(exp_avg_sq.sqrt(), eps)
+        if step == 2:
+            exp_avg.addcdiv_(grad, denom)
+        else:
+            exp_avg.mul_(beta1).addcdiv_(grad, denom, value=1 - beta1)
+
+        param.add_(exp_avg, alpha=-lr)
+        exp_avg_sq.mul_(beta2).addcmul_(grad, grad.conj(), value=1 - beta2)
+
+
+def _multi_tensor_adopt(
+    params: List[Tensor],
+    grads: List[Tensor],
+    exp_avgs: List[Tensor],
+    exp_avg_sqs: List[Tensor],
+    state_steps: List[Tensor],
+    grad_scale: Optional[Tensor],
+    found_inf: Optional[Tensor],
+    *,
+    has_complex: bool,
+    beta1: float,
+    beta2: float,
+    lr: Union[float, Tensor],
+    weight_decay: float,
+    decoupled: bool,
+    eps: float,
+    maximize: bool,
+    capturable: bool,
+    differentiable: bool,
+):
+    if len(params) == 0:
+        return
+
+    if isinstance(lr, Tensor) and not capturable:
+        raise RuntimeError(
+            "lr as a Tensor is not supported for capturable=False and foreach=True"
+        )
+
+    # If compiling, the compiler will handle cudagraph checks, see note [torch.compile x capturable]
+    if not torch._utils.is_compiling() and capturable:
+        capturable_supported_devices = _get_capturable_supported_devices(
+            supports_xla=False
+        )
+        assert all(
+            p.device.type == step.device.type
+            and p.device.type in capturable_supported_devices
+            for p, step in zip(params, state_steps)
+        ), f"If capturable=True, params and state_steps must be on supported devices: {capturable_supported_devices}."
+
+    assert grad_scale is None and found_inf is None
+
+    assert not differentiable, "_foreach ops don't support autograd"
+
+    grouped_tensors = Optimizer._group_tensors_by_device_and_dtype(
+        [params, grads, exp_avgs, exp_avg_sqs, state_steps]  # type: ignore[list-item]
+    )
+    for (
+        device_params_,
+        device_grads_,
+        device_exp_avgs_,
+        device_exp_avg_sqs_,
+        device_state_steps_,
+    ), _ in grouped_tensors.values():
+        device_params = cast(List[Tensor], device_params_)
+        device_grads = cast(List[Tensor], device_grads_)
+        device_exp_avgs = cast(List[Tensor], device_exp_avgs_)
+        device_exp_avg_sqs = cast(List[Tensor], device_exp_avg_sqs_)
+        device_state_steps = cast(List[Tensor], device_state_steps_)
+
+        # Handle complex parameters
+        if has_complex:
+            _view_as_real(
+                device_params, device_grads, device_exp_avgs, device_exp_avg_sqs
+            )
+
+        if maximize:
+            device_grads = torch._foreach_neg(device_grads)  # type: ignore[assignment]
+
+        # Update steps
+        # If steps are on CPU, foreach will fall back to the slow path, which is a for-loop calling t.add(1) over
+        # and over. 1 will then be wrapped into a Tensor over and over again, which is slower than if we just
+        # wrapped it once now. The alpha is required to assure we go to the right overload.
+        if not torch._utils.is_compiling() and device_state_steps[0].is_cpu:
+            torch._foreach_add_(
+                device_state_steps, torch.tensor(1.0, device="cpu"), alpha=1.0
+            )
+        else:
+            torch._foreach_add_(device_state_steps, 1)
+
+        if weight_decay != 0:
+            if decoupled:
+                torch._foreach_add_(
+                    device_params, device_params, alpha=-lr * weight_decay
+                )
+            else:
+                # Re-use the intermediate memory (device_grads) already allocated for maximize
+                if maximize:
+                    torch._foreach_add_(device_grads, device_params, alpha=weight_decay)
+                else:
+                    device_grads = torch._foreach_add(  # type: ignore[assignment]
+                        device_grads, device_params, alpha=weight_decay
+                    )
+
+        if device_state_steps[0] == 1:
+            torch._foreach_addcmul_(device_exp_avg_sqs, device_grads, device_grads)
+            continue
+
+        exp_avg_sq_sqrt = torch._foreach_sqrt(device_exp_avg_sqs)
+        exp_avg_sq_sqrt = torch._foreach_maximum(exp_avg_sq_sqrt, eps)
+
+        if device_state_steps[0] == 2:
+            torch._foreach_addcdiv_(device_exp_avgs, device_grads, exp_avg_sq_sqrt)
+        else:
+            torch._foreach_mul_(device_exp_avgs, beta1)
+            torch._foreach_addcdiv_(
+                device_exp_avgs, device_grads, exp_avg_sq_sqrt, value=1 - beta1
+            )
+
+        torch._foreach_add_(device_params, device_exp_avgs, alpha=-lr)
+        torch._foreach_mul_(device_exp_avg_sqs, beta2)
+        torch._foreach_addcmul_(
+            device_exp_avg_sqs, device_grads, device_grads, value=1 - beta2
+        )
+
+
+@_disable_dynamo_if_unsupported(single_tensor_fn=_single_tensor_adopt)
+def adopt(
+    params: List[Tensor],
+    grads: List[Tensor],
+    exp_avgs: List[Tensor],
+    exp_avg_sqs: List[Tensor],
+    state_steps: List[Tensor],
+    # kwonly args with defaults are not supported by functions compiled with torchscript issue #70627
+    # setting this as kwarg for now as functional API is compiled by torch/distributed/optim
+    foreach: Optional[bool] = None,
+    capturable: bool = False,
+    differentiable: bool = False,
+    fused: Optional[bool] = None,
+    grad_scale: Optional[Tensor] = None,
+    found_inf: Optional[Tensor] = None,
+    has_complex: bool = False,
+    *,
+    beta1: float,
+    beta2: float,
+    lr: Union[float, Tensor],
+    weight_decay: float,
+    decoupled: bool,
+    eps: float,
+    maximize: bool,
+):
+    r"""Functional API that performs ADOPT algorithm computation."""
+    # Respect when the user inputs False/True for foreach or fused. We only want to change
+    # the default when neither have been user-specified. Note that we default to foreach
+    # and pass False to use_fused. This is not a mistake--we want to give the fused impl
+    # bake-in time before making it the default, even if it is typically faster.
+    if fused is None and foreach is None:
+        _, foreach = _default_to_fused_or_foreach(
+            params, differentiable, use_fused=False
+        )
+        # Do not flip on foreach for the unsupported case where lr is a Tensor and capturable=False.
+        if foreach and isinstance(lr, Tensor) and not capturable:
+            foreach = False
+    if fused is None:
+        fused = False
+    if foreach is None:
+        foreach = False
+
+    # this check is slow during compilation, so we skip it
+    # if it's strictly needed we can add this check back in dynamo
+    if not torch._utils.is_compiling() and not all(
+        isinstance(t, torch.Tensor) for t in state_steps
+    ):
+        raise RuntimeError(
+            "API has changed, `state_steps` argument must contain a list of singleton tensors"
+        )
+
+    if foreach and torch.jit.is_scripting():
+        raise RuntimeError("torch.jit.script not supported with foreach optimizers")
+    if fused and torch.jit.is_scripting():
+        raise RuntimeError("torch.jit.script not supported with fused optimizers")
+
+    # if fused and not torch.jit.is_scripting():
+    #     func = _fused_adopt
+    # elif foreach and not torch.jit.is_scripting():
+    if foreach and not torch.jit.is_scripting():
+        func = _multi_tensor_adopt
+    else:
+        func = _single_tensor_adopt
+
+    func(
+        params,
+        grads,
+        exp_avgs,
+        exp_avg_sqs,
+        state_steps,
+        has_complex=has_complex,
+        beta1=beta1,
+        beta2=beta2,
+        lr=lr,
+        weight_decay=weight_decay,
+        decoupled=decoupled,
+        eps=eps,
+        maximize=maximize,
+        capturable=capturable,
+        differentiable=differentiable,
+        grad_scale=grad_scale,
+        found_inf=found_inf,
+    )

tests/e2e/test_optimizers.py

@@ -13,7 +13,7 @@ from axolotl.train import train
 from axolotl.utils.config import normalize_config
 from axolotl.utils.dict import DictDefault
 
-from .utils import with_temp_dir
+from .utils import require_torch_2_5_1, with_temp_dir
 
 LOG = logging.getLogger("axolotl.tests.e2e")
 os.environ["WANDB_DISABLED"] = "true"
@@ -65,3 +65,46 @@ class TestCustomOptimizers(unittest.TestCase):
 
         train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
         assert (Path(temp_dir) / "adapter_model.bin").exists()
+
+    @with_temp_dir
+    @require_torch_2_5_1
+    def test_adopt_adamw(self, temp_dir):
+        # pylint: disable=duplicate-code
+        cfg = DictDefault(
+            {
+                "base_model": "JackFram/llama-68m",
+                "tokenizer_type": "LlamaTokenizer",
+                "sequence_len": 1024,
+                "load_in_8bit": True,
+                "adapter": "lora",
+                "lora_r": 8,
+                "lora_alpha": 16,
+                "lora_dropout": 0.05,
+                "lora_target_linear": True,
+                "val_set_size": 0.1,
+                "special_tokens": {
+                    "unk_token": "<unk>",
+                    "bos_token": "<s>",
+                    "eos_token": "</s>",
+                },
+                "datasets": [
+                    {
+                        "path": "mhenrichsen/alpaca_2k_test",
+                        "type": "alpaca",
+                    },
+                ],
+                "num_epochs": 1,
+                "micro_batch_size": 8,
+                "gradient_accumulation_steps": 1,
+                "output_dir": temp_dir,
+                "learning_rate": 0.00001,
+                "optimizer": "adopt_adamw",
+                "lr_scheduler": "cosine",
+            }
+        )
+        normalize_config(cfg)
+        cli_args = TrainerCliArgs()
+        dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
+
+        train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
+        assert (Path(temp_dir) / "adapter_model.bin").exists()

tests/e2e/utils.py

@@ -6,11 +6,13 @@ import shutil
 import tempfile
 import unittest
 from functools import wraps
-from importlib.metadata import version
 from pathlib import Path
 
 import torch
 
+# from importlib.metadata import version
+from packaging import version
+
 
 def with_temp_dir(test_func):
     @wraps(test_func)
@@ -43,12 +45,24 @@ def require_torch_2_3_1(test_case):
     """
 
     def is_min_2_3_1():
-        torch_version = version("torch")
+        torch_version = version.parse(torch.__version__)
-        return torch_version >= "2.3.1"
+        return torch_version >= version.parse("2.3.1")
 
     return unittest.skipUnless(is_min_2_3_1(), "test torch 2.3.1")(test_case)
 
 
+def require_torch_2_5_1(test_case):
+    """
+    Decorator marking a test that requires torch >= 2.3.1
+    """
+
+    def is_min_2_5_1():
+        torch_version = version.parse(torch.__version__)
+        return torch_version >= version.parse("2.5.1")
+
+    return unittest.skipUnless(is_min_2_5_1(), "test torch 2.5.1")(test_case)
+
+
 def is_hopper():
     compute_capability = torch.cuda.get_device_capability()
     return compute_capability == (9, 0)