2022-03-25 10:03:54 +00:00
|
|
|
import torch
|
2022-03-30 10:14:59 +00:00
|
|
|
|
2022-03-25 10:03:54 +00:00
|
|
|
from colossalai.utils import multi_tensor_applier
|
2022-03-30 10:14:59 +00:00
|
|
|
from colossalai.registry import OPTIMIZERS
|
2022-04-02 09:04:05 +00:00
|
|
|
from colossalai.nn.optimizer import CPU_ADAM_CNT
|
2022-07-26 09:25:24 +00:00
|
|
|
from typing import Optional
|
|
|
|
from .nvme_optimizer import NVMeOptimizer
|
2022-03-30 10:14:59 +00:00
|
|
|
|
2022-03-25 10:03:54 +00:00
|
|
|
|
2022-03-30 10:14:59 +00:00
|
|
|
@OPTIMIZERS.register_module
|
2022-07-26 09:25:24 +00:00
|
|
|
class HybridAdam(NVMeOptimizer):
|
2022-03-30 10:14:59 +00:00
|
|
|
"""Implements Adam algorithm.
|
|
|
|
|
|
|
|
Supports parameters updating on both GPU and CPU, depanding on the device of paramters.
|
|
|
|
But the parameters and gradients should on the same device:
|
|
|
|
* Parameters on CPU and gradients on CPU is allowed.
|
|
|
|
* Parameters on GPU and gradients on GPU is allowed.
|
|
|
|
* Parameters on GPU and gradients on CPU is **not** allowed.
|
2022-04-01 08:27:03 +00:00
|
|
|
|
2022-03-30 10:14:59 +00:00
|
|
|
Requires ColossalAI to be installed via ``pip install .``
|
|
|
|
|
|
|
|
This version of Hybrid Adam is an hybrid of CPUAdam and FusedAdam.
|
2022-04-01 08:27:03 +00:00
|
|
|
|
|
|
|
* For parameters updating on CPU, it uses CPUAdam.
|
|
|
|
* For parameters updating on GPU, it uses FusedAdam.
|
|
|
|
* Hybird precision calculation of fp16 and fp32 is supported, eg fp32 parameters and fp16 gradients.
|
2022-03-30 10:14:59 +00:00
|
|
|
|
|
|
|
:class:`colossalai.nn.optimizer.HybridAdam` may be used as a drop-in replacement for ``torch.optim.AdamW``,
|
|
|
|
or ``torch.optim.Adam`` with ``adamw_mode=False``
|
|
|
|
|
|
|
|
Adam was been proposed in `Adam: A Method for Stochastic Optimization`_.
|
|
|
|
|
|
|
|
Arguments:
|
|
|
|
model_params (iterable): iterable of parameters of dicts defining
|
|
|
|
parameter groups.
|
|
|
|
lr (float, optional): learning rate. (default: 1e-3)
|
|
|
|
betas (Tuple[float, float], optional): coefficients used for computing
|
|
|
|
running averages of gradient and its square. (default: (0.9, 0.999))
|
|
|
|
eps (float, optional): term added to the denominator to improve
|
|
|
|
numerical stability. (default: 1e-8)
|
|
|
|
weight_decay (float, optional): weight decay (L2 penalty) (default: 0)
|
|
|
|
amsgrad (boolean, optional): whether to use the AMSGrad variant of this
|
|
|
|
algorithm from the paper `On the Convergence of Adam and Beyond`_
|
|
|
|
(default: False) NOT SUPPORTED yet in CPUAdam!
|
|
|
|
adamw_mode (boolean, optional): Apply L2 regularization or weight decay
|
|
|
|
True for decoupled weight decay(also known as AdamW) (default: True)
|
|
|
|
simd_log (boolean, optional): whether to show if you are using SIMD to
|
|
|
|
accelerate. (default: False)
|
2022-07-26 09:25:24 +00:00
|
|
|
nvme_offload_fraction (float, optional): Fraction of params to be offloaded to NVMe. Defaults to 0.0.
|
|
|
|
offload_dir (Optional[str], optional): Directory to save NVMe offload files.
|
|
|
|
If it's ``None``, a random temporary directory will be used. Defaults to None.
|
2022-04-01 08:27:03 +00:00
|
|
|
|
|
|
|
.. _Adam\: A Method for Stochastic Optimization:
|
2022-03-30 10:14:59 +00:00
|
|
|
https://arxiv.org/abs/1412.6980
|
|
|
|
.. _On the Convergence of Adam and Beyond:
|
|
|
|
https://openreview.net/forum?id=ryQu7f-RZ
|
|
|
|
"""
|
|
|
|
|
2022-03-25 10:03:54 +00:00
|
|
|
# Number of fp32 shards for per parameter
|
|
|
|
# Param weight, grad, momentum and variance
|
|
|
|
num_fp32_shards_per_param = 4
|
|
|
|
|
|
|
|
def __init__(self,
|
|
|
|
model_params,
|
|
|
|
lr=1e-3,
|
|
|
|
bias_correction=True,
|
|
|
|
betas=(0.9, 0.999),
|
|
|
|
eps=1e-8,
|
|
|
|
weight_decay=0,
|
|
|
|
adamw_mode=True,
|
2022-07-26 09:25:24 +00:00
|
|
|
simd_log=False,
|
|
|
|
nvme_offload_fraction: float = 0.0,
|
|
|
|
nvme_offload_dir: Optional[str] = None):
|
2022-03-25 10:03:54 +00:00
|
|
|
|
|
|
|
default_args = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay, bias_correction=bias_correction)
|
2022-07-26 09:25:24 +00:00
|
|
|
super(HybridAdam, self).__init__(model_params, default_args, nvme_offload_fraction, nvme_offload_dir)
|
2022-04-02 09:04:05 +00:00
|
|
|
self.opt_id = CPU_ADAM_CNT()
|
2022-03-25 10:03:54 +00:00
|
|
|
self.adamw_mode = adamw_mode
|
|
|
|
try:
|
|
|
|
import cpu_adam
|
|
|
|
import colossal_C
|
|
|
|
except ImportError:
|
|
|
|
raise ImportError('Please install colossalai from source code to use HybridAdam')
|
2022-04-01 08:27:03 +00:00
|
|
|
|
2022-03-25 10:03:54 +00:00
|
|
|
self.cpu_adam_op = cpu_adam
|
|
|
|
self.cpu_adam_op.create_adam(self.opt_id, lr, betas[0], betas[1], eps, weight_decay, adamw_mode, simd_log)
|
|
|
|
|
|
|
|
self.gpu_adam_op = colossal_C.multi_tensor_adam
|
|
|
|
self._dummy_overflow_buf = torch.cuda.IntTensor([0])
|
|
|
|
|
|
|
|
def __del__(self):
|
2022-07-26 09:25:24 +00:00
|
|
|
super().__del__()
|
|
|
|
if getattr(self, 'cpu_adam_op', None):
|
2022-03-25 10:03:54 +00:00
|
|
|
self.cpu_adam_op.destroy_adam(self.opt_id)
|
|
|
|
|
|
|
|
@torch.no_grad()
|
|
|
|
def step(self, closure=None):
|
|
|
|
loss = None
|
|
|
|
if closure is not None:
|
|
|
|
with torch.enable_grad():
|
|
|
|
loss = closure()
|
|
|
|
|
2022-07-26 09:25:24 +00:00
|
|
|
self._pre_step('exp_avg', 'exp_avg_sq')
|
2022-03-25 10:03:54 +00:00
|
|
|
for _, group in enumerate(self.param_groups):
|
|
|
|
g_l, p_l, m_l, v_l = [], [], [], []
|
|
|
|
group_step = 0
|
|
|
|
for _, p in enumerate(group['params']):
|
|
|
|
|
|
|
|
if p.grad is None:
|
|
|
|
continue
|
|
|
|
|
|
|
|
state = self.state[p]
|
|
|
|
|
|
|
|
target_device = p.device
|
|
|
|
if len(state) == 0:
|
|
|
|
state['step'] = 0
|
|
|
|
|
|
|
|
# gradient momentums
|
2022-07-29 11:33:24 +00:00
|
|
|
state['exp_avg'] = torch.zeros_like(p, dtype=torch.float, device=target_device)
|
2022-03-25 10:03:54 +00:00
|
|
|
# gradient variances
|
2022-07-29 11:33:24 +00:00
|
|
|
state['exp_avg_sq'] = torch.zeros_like(p, dtype=torch.float, device=target_device)
|
2022-07-26 09:25:24 +00:00
|
|
|
self._post_state_init(p)
|
2022-03-25 10:03:54 +00:00
|
|
|
|
|
|
|
state['step'] += 1
|
|
|
|
group_step = state['step']
|
|
|
|
beta1, beta2 = group['betas']
|
|
|
|
|
|
|
|
if target_device.type == 'cpu':
|
|
|
|
assert state['exp_avg'].device.type == 'cpu', "exp_avg should stay on cpu"
|
|
|
|
assert state['exp_avg_sq'].device.type == 'cpu', "exp_avg should stay on cpu"
|
2022-07-26 09:25:24 +00:00
|
|
|
self._pre_update(p, 'exp_avg', 'exp_avg_sq')
|
2022-03-25 10:03:54 +00:00
|
|
|
self.cpu_adam_op.adam_update(self.opt_id, state['step'], group['lr'], beta1, beta2, group['eps'],
|
|
|
|
group['weight_decay'], group['bias_correction'], p.data, p.grad.data,
|
|
|
|
state['exp_avg'], state['exp_avg_sq'], -1)
|
2022-07-26 09:25:24 +00:00
|
|
|
self._post_update(p, 'exp_avg', 'exp_avg_sq')
|
2022-03-25 10:03:54 +00:00
|
|
|
|
|
|
|
elif target_device.type == 'cuda':
|
|
|
|
assert state['exp_avg'].device.type == 'cuda', "exp_avg should stay on cuda"
|
|
|
|
assert state['exp_avg_sq'].device.type == 'cuda', "exp_avg should stay on cuda"
|
|
|
|
|
|
|
|
# record the state by gruop and update at once
|
|
|
|
g_l.append(p.grad.data)
|
|
|
|
p_l.append(p.data)
|
|
|
|
m_l.append(state['exp_avg'])
|
2022-04-01 08:27:03 +00:00
|
|
|
v_l.append(state['exp_avg_sq'])
|
2022-03-25 10:03:54 +00:00
|
|
|
|
|
|
|
else:
|
|
|
|
raise RuntimeError
|
|
|
|
if len(g_l) > 0:
|
|
|
|
adamw_mode = 1 if self.adamw_mode else 0
|
|
|
|
bias_correction = 1 if group['bias_correction'] else 0
|
2022-04-01 08:27:03 +00:00
|
|
|
multi_tensor_applier(self.gpu_adam_op, self._dummy_overflow_buf, [g_l, p_l, m_l, v_l], group['lr'],
|
|
|
|
group['betas'][0], group['betas'][1], group['eps'], group_step, adamw_mode,
|
|
|
|
bias_correction, group['weight_decay'])
|
2022-07-26 09:25:24 +00:00
|
|
|
self._post_step()
|
2022-03-25 10:03:54 +00:00
|
|
|
return loss
|