File size: 6,413 Bytes
1c72248
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
import math
import torch
from torch.optim import Optimizer
from toolkit.optimizers.optimizer_utils import copy_stochastic, Auto8bitTensor, stochastic_grad_accummulation

class Adam8bit(Optimizer):
    """
    Implements Adam optimizer with 8-bit state storage and stochastic rounding.
    
    Arguments:
        params (iterable): Iterable of parameters to optimize or dicts defining parameter groups
        lr (float): Learning rate (default: 1e-3)
        betas (tuple): Coefficients for computing running averages of gradient and its square (default: (0.9, 0.999))
        eps (float): Term added to denominator to improve numerical stability (default: 1e-8)
        weight_decay (float): Weight decay coefficient (default: 0)
        decouple (bool): Use AdamW style decoupled weight decay (default: True)
    """
    def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, 
                 weight_decay=0, decouple=True):
        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]}")
        
        defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay,
                       decouple=decouple)
        super(Adam8bit, self).__init__(params, defaults)
        
        self.is_stochastic_rounding_accumulation = False
        
        # Setup stochastic grad accumulation hooks
        for group in self.param_groups:
            for param in group['params']:
                if param.requires_grad and param.dtype != torch.float32:
                    self.is_stochastic_rounding_accumulation = True
                    param.register_post_accumulate_grad_hook(
                        stochastic_grad_accummulation
                    )

    @property
    def supports_memory_efficient_fp16(self):
        return False

    @property
    def supports_flat_params(self):
        return True

    def step_hook(self):
        if not self.is_stochastic_rounding_accumulation:
            return
        # Copy over stochastically rounded grads
        for group in self.param_groups:
            for param in group['params']:
                if param.requires_grad and hasattr(param, "_accum_grad"):
                    param.grad = param._accum_grad
                    del param._accum_grad

    @torch.no_grad()
    def step(self, closure=None):
        """Performs a single optimization step.
        
        Arguments:
            closure (callable, optional): A closure that reevaluates the model and returns the loss.
        """
        # Call pre step
        self.step_hook()
        
        loss = None
        if closure is not None:
            loss = closure()

        for group in self.param_groups:
            beta1, beta2 = group['betas']
            eps = group['eps']
            lr = group['lr']
            decay = group['weight_decay']
            decouple = group['decouple']

            for p in group['params']:
                if p.grad is None:
                    continue

                grad = p.grad.data.to(torch.float32)
                p_fp32 = p.clone().to(torch.float32)

                # Apply weight decay (coupled variant)
                if decay != 0 and not decouple:
                    grad.add_(p_fp32.data, alpha=decay)

                state = self.state[p]

                # State initialization
                if len(state) == 0:
                    state['step'] = 0
                    # Exponential moving average of gradient values
                    state['exp_avg'] = Auto8bitTensor(
                        torch.zeros_like(p_fp32.data).detach())
                    # Exponential moving average of squared gradient values
                    state['exp_avg_sq'] = Auto8bitTensor(
                        torch.zeros_like(p_fp32.data).detach())

                exp_avg = state['exp_avg'].to(torch.float32)
                exp_avg_sq = state['exp_avg_sq'].to(torch.float32)

                state['step'] += 1
                bias_correction1 = 1 - beta1 ** state['step']
                bias_correction2 = 1 - beta2 ** state['step']

                # Adam EMA updates
                exp_avg.mul_(beta1).add_(grad, alpha=1-beta1)
                exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1-beta2)

                # Apply weight decay (decoupled variant)
                if decay != 0 and decouple:
                    p_fp32.data.mul_(1 - lr * decay)

                # Bias correction
                step_size = lr / bias_correction1
                denom = (exp_avg_sq.sqrt() / math.sqrt(bias_correction2)).add_(eps)

                # Take step
                p_fp32.data.addcdiv_(exp_avg, denom, value=-step_size)
                
                # Update state with stochastic rounding
                state['exp_avg'] = Auto8bitTensor(exp_avg)
                state['exp_avg_sq'] = Auto8bitTensor(exp_avg_sq)
                
                # Apply stochastic rounding to parameters
                copy_stochastic(p.data, p_fp32.data)

        return loss
    
    def state_dict(self):
        """Returns the state of the optimizer as a dict."""
        state_dict = super().state_dict()
        
        # Convert Auto8bitTensor objects to regular state dicts
        for param_id, param_state in state_dict['state'].items():
            for key, value in param_state.items():
                if isinstance(value, Auto8bitTensor):
                    param_state[key] = {
                        '_type': 'Auto8bitTensor',
                        'state': value.state_dict()
                    }
        
        return state_dict

    def load_state_dict(self, state_dict):
        """Loads the optimizer state."""
        # First, load the basic state
        super().load_state_dict(state_dict)
        
        # Then convert any Auto8bitTensor states back to objects
        for param_id, param_state in self.state.items():
            for key, value in param_state.items():
                if isinstance(value, dict) and value.get('_type') == 'Auto8bitTensor':
                    param_state[key] = Auto8bitTensor(value['state'])