adv-optm 1.2.0__py3-none-any.whl

adv_optm/optim/Lion_Prodigy_adv.py

@@ -0,0 +1,341 @@
+import torch
+import torch.distributed as dist
+
+import math
+
+from typing import Tuple, Optional
+
+from ..util.BF16_Stochastic_Rounding import add_stochastic_
+from ..util.Effective_Shape import _get_effective_shape
+from ..util.NNMF import _nnmf,_unnmf
+from ..util.OrthoGrad import _orthogonalize_gradient
+from ..util.One_Bit_Boolean import _pack_bools, _unpack_bools
+
+class Lion_Prodigy_adv(torch.optim.Optimizer):
+    """
+    Implements the SMMF technique and Prodigy D-Adaptation method for Lion algorithm.
+
+    Args:
+        params (iterable): iterable of parameters to optimize or dicts defining
+            parameter groups.
+        lr (float, optional): learning rate (default: 1e-4).
+        betas (Tuple[float, float], optional): coefficients for computing
+            running averages of the update (default: (0.9, 0.99)).
+        weight_decay (float, optional): weight decay (L2 penalty) (default: 0.0).
+        vector_reshape (bool, optional): whether to reshape 1D vectors into 2D
+            matrices to apply low-rank compression (default: True).
+        stochastic_rounding (bool, optional): whether to use stochastic
+            rounding for BF16 parameter updates (default: True).
+        cautious_mask (bool): whether to use the cautious masking technique. (default: False).
+        clip_threshold (float, optional): whether to clip the gradients norm
+            per-parameter as proposed in the paper `Lions and Muons: Optimization via
+            Stochastic Frank-Wolfe` (https://arxiv.org/abs/2506.04192) to make Lion more stable
+            (default: 0.0).
+        nnmf_factor (bool): whether to use the factorization or use the
+            uncompressed optimizer. (default: True)
+        d0 (float):
+            Initial D estimate for D-adaptation (default 1e-6). Rarely needs changing.
+        d_coef (float):
+            Coefficient in the expression for the estimate of d (default 1.0).
+            Values such as 0.5 and 2.0 typically work as well. 
+            Changing this parameter is the preferred way to tune the method.
+        growth_rate (float):
+            prevent the D estimate from growing faster than this multiplicative rate.
+            Default is inf, for unrestricted. Values like 1.02 give a kind of learning
+            rate warmup effect.
+        fsdp_in_use (bool):
+            If you're using sharded parameters, this should be set to True. The optimizer
+            will attempt to auto-detect this, but if you're using an implementation other
+            than PyTorch's builtin version, the auto-detection won't work.
+        slice_p (int): Reduce memory usage by calculating LR adaptation statistics on only every 
+            pth entry of each tensor. For values greater than 1 this an an approximation to standard 
+            Prodigy. Values ~11 are reasonable (default 11).
+        prodigy_steps (int): If greater than zero, disable Prodigy's stepsize adjustments
+            after the specified optimiser step and release all state memory required by Prodigy
+            (default: 0).
+        d_limiter (bool): whether to clamp the new step size estimate (`d_hat`)
+            to prevent sudden, volatile increases in the adaptive step size (`d`).
+            (default: True)
+    """
+
+    def __init__(
+        self,
+        params,
+        lr: float = 1,
+        betas: Tuple[float, float] = (0.9, 0.99),
+        weight_decay: float = 0.0,
+        vector_reshape: bool = True,
+        stochastic_rounding: bool = True,
+        orthogonal_gradient: bool = False,
+        cautious_mask: bool = False,
+        clip_threshold: float = 0.0,
+        nnmf_factor: bool = False,
+        # prodigy parameters
+        beta3: float = None,
+        d0: float = 1e-6,
+        d_coef: float = 1,
+        growth_rate: float = float('inf'),
+        safeguard_warmup: bool = False,
+        fsdp_in_use: bool = False,
+        slice_p: int = 11,
+        prodigy_steps: int = 0,
+        d_limiter: bool = True,
+    ):
+        if not lr > 0.0:
+            raise ValueError(f"Learning rate must be > 0.0, but got {lr}")
+        if not all(0.0 <= beta <= 1.0 for beta in betas):
+            raise ValueError(f"Betas should be in [0.0, 1.0], but got {betas}")
+        if not weight_decay >= 0.0:
+            raise ValueError(f"Weight decay must be >= 0.0, but got {weight_decay}")
+
+        defaults = dict(
+            lr=lr,
+            betas=betas,
+            weight_decay=weight_decay,
+            vector_reshape=vector_reshape,
+            orthogonal_gradient=orthogonal_gradient,
+            clip_threshold=clip_threshold,
+            beta3=beta3, d=d0, d0=d0, d_max=d0, d_numerator=0.0, d_coef=d_coef,
+            growth_rate=growth_rate, safeguard_warmup=safeguard_warmup, k=0, slice_p=slice_p,
+            fsdp_in_use=fsdp_in_use,
+            prodigy_steps=prodigy_steps,
+            d_limiter=d_limiter,
+        )
+        self.stochastic_rounding = stochastic_rounding
+        self.cautious_mask = cautious_mask
+        self.factored = nnmf_factor
+        self.fsdp_in_use = fsdp_in_use
+        super().__init__(params, defaults)
+        # Global state for accumulating metrics across parameter updates within a single step.
+        self.init_step()
+
+    @property
+    def supports_fused_back_pass(self) -> bool:
+        return True
+
+    @property
+    def supports_memory_efficient_fp16(self) -> bool:
+        return True
+
+    @property
+    def supports_flat_params(self) -> bool:
+        return False
+
+    def init_step(self):
+        """Resets accumulators and calculates dlr for the upcoming step."""
+        self.d_denom = 0.0
+        
+        g_group = self.param_groups[0]
+        self.beta1, self.beta2 = g_group['betas']
+        self.beta3 = g_group['beta3']
+        if self.beta3 is None:
+            self.beta3 = math.sqrt(self.beta2)
+        
+        k = g_group['k']
+        self.d = g_group['d']
+        lr = g_group['lr']
+
+        self.dlr = self.d * lr
+
+        self.d_numerator = g_group.get('d_numerator', 0.0) * self.beta3
+
+    @torch.no_grad()
+    def step_parameter(self, p: torch.Tensor, group: dict, i: Optional[int] = None):
+        """Performs a single optimization step on a single parameter."""
+        if p.grad is None:
+            return
+
+        if hasattr(p, "_fsdp_flattened"):
+            self.fsdp_in_use = True
+
+        grad = p.grad
+        if grad.dtype != torch.float32 and self.factored:
+            grad = grad.float()
+        if group["clip_threshold"] > 0.0:
+            grad_norm = torch.norm(grad.detach())
+            if grad_norm > group["clip_threshold"]:
+                clip_coef = group["clip_threshold"] / grad_norm
+                grad.mul_(clip_coef)
+        if group["orthogonal_gradient"]:
+            grad = _orthogonalize_gradient(p, grad)
+        state = self.state[p]
+
+        # State Initialization
+        if 'step' not in state:
+            state['step'] = 0
+
+            should_factor = (
+                self.factored and
+                not (len(p.shape) == 1 and not group['vector_reshape'])
+            )
+
+            state['factored'] = should_factor
+
+            dtype = torch.float32 if self.factored else p.dtype
+
+            slice_p = group['slice_p']
+
+            # D-Adaptation states
+            state['s'] = torch.zeros_like(p.flatten()[::slice_p]).detach()
+            if p.any():
+                state['p0'] = p.flatten()[::slice_p].detach().clone()
+            else:
+                state['p0'] = torch.tensor(0, device=p.device, dtype=p.dtype)
+
+            if state['factored']:
+                state['effective_shape'] = _get_effective_shape(p.numel())
+                d1, d2 = state['effective_shape']
+                state['mu_m_nmf'] = torch.zeros(d1, device=p.device, dtype=dtype) 
+                state['mv_m_nmf'] = torch.zeros(d2, device=p.device, dtype=dtype)
+                packed_d2 = (d2 + 7) // 8
+                state['sign'] = torch.zeros((d1, packed_d2), dtype=torch.uint8, device=p.device)
+            else: # Fallback to standard Lion
+                state['exp_avg'] = torch.zeros_like(p, device=p.device, dtype=dtype)
+
+        if state['factored']:
+            # Factored Path
+            d1, d2 = state['effective_shape']
+            grad_reshaped = grad.view(d1, d2)
+            # Reconstruct momentum m_{t-1}
+            exp_avg = _unnmf((state['mu_m_nmf'], state['mv_m_nmf']))
+            unpacked_sign = _unpack_bools(state['sign'], original_m=d2)
+            torch.where(unpacked_sign, exp_avg, -exp_avg, out=exp_avg)
+            del unpacked_sign
+            if exp_avg.dtype != torch.float32:
+                exp_avg = exp_avg.float()
+
+            # Compute update term c_t = β1*m_{t-1} + (1-β1)*g_t
+            signed_update = exp_avg.clone().mul_(self.beta1).add_(grad_reshaped, alpha=self.d * (1-self.beta1)).sign_()
+
+            if self.cautious_mask:
+                mask = (signed_update * grad_reshaped > 0).to(grad_reshaped.dtype)
+                mask.div_(mask.mean().clamp_(min=1e-3))
+                signed_update.mul_(mask)
+                del mask
+
+            # Parameter update: p_t = p_{t-1} - lr * sign(c_t)
+            update_for_param = signed_update.view(p.shape).mul(self.dlr)
+
+            # Update momentum m_t = β2*m_{t-1} + (1-β2)*lr*g_t
+            exp_avg.mul_(self.beta2).add_(grad_reshaped, alpha=self.d * (1 - self.beta2))
+            del grad_reshaped
+
+            # Compress new momentum m_t and store factors
+            state['sign'] = _pack_bools(exp_avg > 0)
+            _nnmf(exp_avg.abs(), out=(state['mu_m_nmf'], state['mv_m_nmf']))
+            del exp_avg
+
+        else:
+            # Fallback to standard Lion logic
+            exp_avg = state["exp_avg"]
+
+            # Compute update term and sign for the update
+            if exp_avg.dtype != torch.float32 and self.factored:
+                exp_avg = exp_avg.float()
+            signed_update = exp_avg.clone().mul_(self.beta1).add_(grad, alpha=self.d * (1-self.beta1)).sign_()
+
+            if self.cautious_mask:
+                mask = (signed_update * grad > 0).to(grad.dtype)
+                mask.div_(mask.mean().clamp_(min=1e-3))
+                signed_update.mul_(mask)
+                del mask
+
+            update_for_param = signed_update.mul(self.dlr)
+
+            # Update momentum 
+            exp_avg.mul_(self.beta2).add_(grad, alpha=self.d * (1 - self.beta2))
+
+        prodigy_steps = group['prodigy_steps']
+        if prodigy_steps <= 0 or group['k'] < prodigy_steps:
+            # --- Accumulate Prodigy stats ---
+            d0, safeguard_warmup, slice_p = group['d0'], group['safeguard_warmup'], group['slice_p']
+            s, p0 = state['s'], state['p0']
+            grad_flat = grad.flatten().float()
+            p_flat = p.data.flatten().float()
+            p0 = p0.float()
+
+            self.d_numerator += (self.d / d0) * self.dlr * torch.dot(grad_flat[::slice_p], p0.data - p_flat[::slice_p]).item()
+
+            alpha = ((self.d / d0) * self.d) if safeguard_warmup else ((self.d / d0) * self.dlr)
+            s.mul_(self.beta3).add_(grad_flat[::slice_p], alpha=alpha)
+            self.d_denom += s.abs().sum().item()
+
+            del s, p0, grad_flat, p_flat, alpha
+        else:
+            # Free memory if prodigy_steps is reached
+            if 's' in state:
+                del state['s']
+            if 'p0' in state:
+                del state['p0']
+
+        if group["weight_decay"] != 0:
+            if p.dtype == torch.bfloat16 and self.stochastic_rounding:
+                add_stochastic_(p.data, p.data,
+                                alpha=-group["weight_decay"] * self.dlr)
+            else:
+                p.data.add_(
+                    p.data, alpha=-group["weight_decay"] * self.dlr
+                )
+
+        if p.dtype == torch.bfloat16 and self.stochastic_rounding:
+            add_stochastic_(p.data, -update_for_param)
+        else:
+            p.data.add_(-update_for_param)
+
+        del update_for_param
+
+    @torch.no_grad()
+    def step(self, closure: Optional[callable] = None):
+        """Performs a single optimization step."""
+        loss = None
+        if closure is not None:
+            with torch.enable_grad():
+                loss = closure()
+
+        for group in self.param_groups:
+            for i, p in enumerate(group["params"]):
+                if p.grad is not None:
+                    self.step_parameter(p, group, i)
+
+
+        self.calculate_d()
+        self.init_step()
+        return loss
+
+    def calculate_d(self):
+        """Calculates the new `d` based on the accumulated stats."""
+        g_group = self.param_groups[0]
+        # Only perform d-adaptation if prodigy_steps has not been reached
+        prodigy_active = not (g_group.get('prodigy_steps', 0) > 0 and g_group['k'] >= g_group['prodigy_steps'])
+
+        if prodigy_active:
+            d_max, d_coef, growth_rate = g_group['d_max'], g_group['d_coef'], g_group['growth_rate']
+            
+            if self.fsdp_in_use and dist.is_available() and dist.is_initialized():
+                # Use the device of the first parameter to avoid hardcoding '.cuda()'
+                device = self.param_groups[0]['params'][0].device
+                dist_tensor = torch.tensor([self.d_numerator, self.d_denom], device=device)
+                dist.all_reduce(dist_tensor, op=dist.ReduceOp.SUM)
+                global_d_numerator = dist_tensor[0].item()
+                global_d_denom = dist_tensor[1].item()
+            else:
+                global_d_numerator = self.d_numerator
+                global_d_denom = self.d_denom
+
+            d_hat = self.d
+            if global_d_denom > 0:
+                d_hat = d_coef * global_d_numerator / global_d_denom
+                if g_group.get('d_limiter', False):
+                    d_hat = min(self.d * (2 ** 0.25), d_hat)
+                if self.d == g_group['d0']:
+                    self.d = max(self.d, d_hat)
+                d_max = max(d_max, d_hat)
+                self.d = min(d_max, self.d * growth_rate)
+
+            for group in self.param_groups:
+                group['d_numerator'] = global_d_numerator
+                group['d'] = self.d
+                group['d_max'] = d_max
+        # Increment step counter for all groups, regardless of whether d was updated
+        for group in self.param_groups:
+            group['k'] += 1

adv_optm/optim/Lion_adv.py

@@ -0,0 +1,210 @@
+import torch
+
+from typing import Tuple, Optional
+
+from ..util.BF16_Stochastic_Rounding import add_stochastic_
+from ..util.Effective_Shape import _get_effective_shape
+from ..util.NNMF import _nnmf,_unnmf
+from ..util.OrthoGrad import _orthogonalize_gradient
+from ..util.One_Bit_Boolean import _pack_bools, _unpack_bools
+
+class Lion_adv(torch.optim.Optimizer):
+    """
+    Implements the SMMF technique for Lion algorithm.
+
+    This optimizer combines the Lion update rule with the memory-saving low-rank
+    compression (SMMF) technique from https://arxiv.org/abs/2412.08894.
+
+    Args:
+        params (iterable): iterable of parameters to optimize or dicts defining
+            parameter groups.
+        lr (float, optional): learning rate (default: 1e-4).
+        betas (Tuple[float, float], optional): coefficients for computing
+            running averages of the update (default: (0.9, 0.99)).
+        weight_decay (float, optional): weight decay (L2 penalty) (default: 0.0).
+        vector_reshape (bool, optional): whether to reshape 1D vectors into 2D
+            matrices to apply low-rank compression (default: True).
+        stochastic_rounding (bool, optional): whether to use stochastic
+            rounding for BF16 parameter updates (default: True).
+        cautious_mask (bool): whether to use the cautious masking technique. (default: False).
+        clip_threshold (float, optional): whether to clip the gradients norm
+            per-parameter as proposed in the paper `Lions and Muons: Optimization via
+            Stochastic Frank-Wolfe` (https://arxiv.org/abs/2506.04192) to make Lion more stable
+            (default: 0.0).
+        nnmf_factor (bool): whether to use the factorization or use the
+            uncompressed optimizer. (default: True)
+    """
+
+    def __init__(
+        self,
+        params,
+        lr: float = 1e-4,
+        betas: Tuple[float, float] = (0.9, 0.99),
+        weight_decay: float = 0.0,
+        vector_reshape: bool = True,
+        stochastic_rounding: bool = True,
+        orthogonal_gradient: bool = False,
+        cautious_mask: bool = False,
+        clip_threshold: float = 0.0,
+        nnmf_factor: bool = True,
+    ):
+        if not lr > 0.0:
+            raise ValueError(f"Learning rate must be > 0.0, but got {lr}")
+        if not all(0.0 <= beta <= 1.0 for beta in betas):
+            raise ValueError(f"Betas should be in [0.0, 1.0], but got {betas}")
+        if not weight_decay >= 0.0:
+            raise ValueError(f"Weight decay must be >= 0.0, but got {weight_decay}")
+
+        defaults = dict(
+            lr=lr,
+            betas=betas,
+            weight_decay=weight_decay,
+            vector_reshape=vector_reshape,
+            orthogonal_gradient=orthogonal_gradient,
+            clip_threshold=clip_threshold,
+        )
+        self.stochastic_rounding = stochastic_rounding
+        self.cautious_mask = cautious_mask
+        self.factored = nnmf_factor
+        super().__init__(params, defaults)
+
+    @property
+    def supports_fused_back_pass(self) -> bool:
+        return True
+
+    @property
+    def supports_memory_efficient_fp16(self) -> bool:
+        return True
+
+    @property
+    def supports_flat_params(self) -> bool:
+        return False
+
+    @torch.no_grad()
+    def step_parameter(self, p: torch.Tensor, group: dict, i: Optional[int] = None):
+        """Performs a single optimization step on a single parameter."""
+        if p.grad is None:
+            return
+
+        grad = p.grad
+        if grad.dtype != torch.float32 and self.factored:
+            grad = grad.float()
+        if group["clip_threshold"] > 0.0:
+            grad_norm = torch.norm(grad.detach())
+            if grad_norm > group["clip_threshold"]:
+                clip_coef = group["clip_threshold"] / grad_norm
+                grad.mul_(clip_coef)
+        if group["orthogonal_gradient"]:
+            grad = _orthogonalize_gradient(p, grad)
+        state = self.state[p]
+
+        # State Initialization
+        if 'step' not in state:
+            state['step'] = 0
+
+            should_factor = (
+                self.factored and
+                not (len(p.shape) == 1 and not group['vector_reshape'])
+            )
+
+            state['factored'] = should_factor
+
+            dtype = torch.float32 if self.factored else p.dtype
+
+            if state['factored']:
+                state['effective_shape'] = _get_effective_shape(p.numel())
+                d1, d2 = state['effective_shape']
+                state['mu_m_nmf'] = torch.zeros(d1, device=p.device, dtype=dtype) 
+                state['mv_m_nmf'] = torch.zeros(d2, device=p.device, dtype=dtype)
+                packed_d2 = (d2 + 7) // 8
+                state['sign'] = torch.zeros((d1, packed_d2), dtype=torch.uint8, device=p.device)
+            else: # Fallback to standard Lion
+                state['exp_avg'] = torch.zeros_like(p, device=p.device, dtype=dtype)
+
+        state['step'] += 1
+        beta1, beta2 = group["betas"]
+        lr = group["lr"]
+
+        if state['factored']:
+            # Factored Path
+            d1, d2 = state['effective_shape']
+            grad_reshaped = grad.view(d1, d2)
+            # Reconstruct momentum m_{t-1}
+            exp_avg = _unnmf((state['mu_m_nmf'], state['mv_m_nmf']))
+            unpacked_sign = _unpack_bools(state['sign'], original_m=d2)
+            torch.where(unpacked_sign, exp_avg, -exp_avg, out=exp_avg)
+            del unpacked_sign
+            if exp_avg.dtype != torch.float32:
+                exp_avg = exp_avg.float()
+
+            # Compute update term c_t
+            signed_update = exp_avg.clone().mul_(beta1).add_(grad_reshaped, alpha=(1-beta1)).sign_()
+
+            if self.cautious_mask:
+                mask = (signed_update * grad_reshaped > 0).to(grad_reshaped.dtype)
+                mask.div_(mask.mean().clamp_(min=1e-3))
+                signed_update.mul_(mask)
+                del mask
+
+            # Parameter update
+            update_for_param = signed_update.view(p.shape).mul_(lr)
+
+            # Standard Lion momentum update
+            exp_avg.mul_(beta2).add_(grad_reshaped, alpha=1-beta2)
+            del grad_reshaped
+
+            # Compress new momentum m_t and store factors
+            state['sign'] = _pack_bools(exp_avg > 0)
+            _nnmf(exp_avg.abs(), out=(state['mu_m_nmf'], state['mv_m_nmf']))
+            del exp_avg
+
+        else:
+            # Fallback to standard Lion logic
+            exp_avg = state["exp_avg"]
+
+            # Compute update term and sign for the update
+            if exp_avg.dtype != torch.float32 and self.factored:
+                exp_avg = exp_avg.float()
+            signed_update = exp_avg.clone().mul_(beta1).add_(grad, alpha=(1-beta1)).sign_()
+
+            if self.cautious_mask:
+                mask = (signed_update * grad > 0).to(grad.dtype)
+                mask.div_(mask.mean().clamp_(min=1e-3))
+                signed_update.mul_(mask)
+                del mask
+
+            update_for_param = signed_update.mul_(lr)
+
+            # Standard Lion momentum update
+            exp_avg.mul_(beta2).add_(grad, alpha=1-beta2)
+
+        if group["weight_decay"] != 0:
+            if p.dtype == torch.bfloat16 and self.stochastic_rounding:
+                add_stochastic_(p.data, p.data,
+                                alpha=-group["weight_decay"] * lr)
+            else:
+                p.data.add_(
+                    p.data, alpha=-group["weight_decay"] * lr
+                )
+
+        if p.dtype == torch.bfloat16 and self.stochastic_rounding:
+            add_stochastic_(p.data, -update_for_param)
+        else:
+            p.data.add_(-update_for_param)
+
+        del update_for_param
+
+    @torch.no_grad()
+    def step(self, closure: Optional[callable] = None):
+        """Performs a single optimization step."""
+        loss = None
+        if closure is not None:
+            with torch.enable_grad():
+                loss = closure()
+
+        for group in self.param_groups:
+            for i, p in enumerate(group["params"]):
+                if p.grad is not None:
+                    self.step_parameter(p, group, i)
+
+        return loss