DL-Art-School/codes/models/ExtensibleTrainer.py
James Betker 44a19cd37c ExtensibleTrainer mods to support advanced checkpointing for stylegan2
Basically: stylegan2 makes use of gradient-based normalizers. These
make it so that I cannot use gradient checkpointing. But I love gradient
checkpointing. It makes things really, really fast and memory conscious.

So - only don't checkpoint when we run the regularizer loss. This is a
bit messy, but speeds up training by at least 20%.

Also: pytorch: please make checkpointing a first class citizen.
2020-11-12 15:45:07 -07:00

365 lines
16 KiB
Python

import logging
import os
import torch
from torch.nn.parallel import DataParallel
import torch.nn as nn
from torch.nn.parallel.distributed import DistributedDataParallel
import models.lr_scheduler as lr_scheduler
import models.networks as networks
from models.base_model import BaseModel
from models.steps.injectors import create_injector
from models.steps.steps import ConfigurableStep
from models.experiments.experiments import get_experiment_for_name
import torchvision.utils as utils
logger = logging.getLogger('base')
class ExtensibleTrainer(BaseModel):
def __init__(self, opt, cached_networks={}):
super(ExtensibleTrainer, self).__init__(opt)
if opt['dist']:
self.rank = torch.distributed.get_rank()
else:
self.rank = -1 # non dist training
train_opt = opt['train']
# env is used as a global state to store things that subcomponents might need.
self.env = {'device': self.device,
'rank': self.rank,
'opt': opt,
'step': 0,
'dist': opt['dist']
}
if opt['path']['models'] is not None:
self.env['base_path'] = os.path.join(opt['path']['models'])
self.mega_batch_factor = 1
if self.is_train:
self.mega_batch_factor = train_opt['mega_batch_factor']
self.env['mega_batch_factor'] = self.mega_batch_factor
self.batch_factor = self.mega_batch_factor
self.checkpointing_cache = opt['checkpointing_enabled']
self.netsG = {}
self.netsD = {}
# Note that this is on the chopping block. It should be integrated into an injection point.
self.netF = networks.define_F().to(self.device) # Used to compute feature loss.
for name, net in opt['networks'].items():
# Trainable is a required parameter, but the default is simply true. Set it here.
if 'trainable' not in net.keys():
net['trainable'] = True
if name in cached_networks.keys():
new_net = cached_networks[name]
else:
new_net = None
if net['type'] == 'generator':
if new_net is None:
new_net = networks.define_G(net, None, opt['scale']).to(self.device)
self.netsG[name] = new_net
elif net['type'] == 'discriminator':
if new_net is None:
new_net = networks.define_D_net(net, opt['datasets']['train']['target_size']).to(self.device)
self.netsD[name] = new_net
else:
raise NotImplementedError("Can only handle generators and discriminators")
if not net['trainable']:
new_net.eval()
if net['wandb_debug']:
import wandb
wandb.watch(new_net, log='all', log_freq=3)
# Initialize the train/eval steps
self.step_names = []
self.steps = []
for step_name, step in opt['steps'].items():
step = ConfigurableStep(step, self.env)
self.step_names.append(step_name) # This could be an OrderedDict, but it's a PITA to integrate with AMP below.
self.steps.append(step)
# step.define_optimizers() relies on the networks being placed in the env, so put them there. Even though
# they aren't wrapped yet.
self.env['generators'] = self.netsG
self.env['discriminators'] = self.netsD
# Define the optimizers from the steps
for s in self.steps:
s.define_optimizers()
self.optimizers.extend(s.get_optimizers())
if self.is_train:
# Find the optimizers that are using the default scheduler, then build them.
def_opt = []
for s in self.steps:
def_opt.extend(s.get_optimizers_with_default_scheduler())
self.schedulers = lr_scheduler.get_scheduler_for_name(train_opt['default_lr_scheme'], def_opt, train_opt)
else:
self.schedulers = []
# Wrap networks in distributed shells.
dnets = []
all_networks = [g for g in self.netsG.values()] + [d for d in self.netsD.values()]
for anet in all_networks:
if opt['dist']:
dnet = DistributedDataParallel(anet,
device_ids=[torch.cuda.current_device()],
find_unused_parameters=False)
else:
dnet = DataParallel(anet, device_ids=opt['gpu_ids'])
if self.is_train:
dnet.train()
else:
dnet.eval()
dnets.append(dnet)
if not opt['dist']:
self.netF = DataParallel(self.netF, device_ids=opt['gpu_ids'])
# Backpush the wrapped networks into the network dicts..
self.networks = {}
found = 0
for dnet in dnets:
for net_dict in [self.netsD, self.netsG]:
for k, v in net_dict.items():
if v == dnet.module:
net_dict[k] = dnet
self.networks[k] = dnet
found += 1
assert found == len(self.netsG) + len(self.netsD)
# Replace the env networks with the wrapped networks
self.env['generators'] = self.netsG
self.env['discriminators'] = self.netsD
self.print_network() # print network
self.load() # load G and D if needed
# Load experiments
self.experiments = []
if 'experiments' in opt.keys():
self.experiments = [get_experiment_for_name(e) for e in op['experiments']]
# Setting this to false triggers SRGAN to call the models update_model() function on the first iteration.
self.updated = True
def feed_data(self, data, step, need_GT=True):
self.env['step'] = step
self.batch_factor = self.mega_batch_factor
self.opt['checkpointing_enabled'] = self.checkpointing_cache
# The batch factor can be adjusted on a period to allow known high-memory steps to fit in GPU memory.
if 'mod_batch_factor' in self.opt['train'].keys() and \
self.env['step'] % self.opt['train']['mod_batch_factor_every'] == 0:
self.batch_factor = self.opt['train']['mod_batch_factor']
if self.opt['train']['mod_batch_factor_also_disable_checkpointing']:
self.opt['checkpointing_enabled'] = False
self.eval_state = {}
for o in self.optimizers:
o.zero_grad()
torch.cuda.empty_cache()
self.lq = [t.to(self.device) for t in torch.chunk(data['LQ'], chunks=self.batch_factor, dim=0)]
if need_GT:
self.hq = [t.to(self.device) for t in torch.chunk(data['GT'], chunks=self.batch_factor, dim=0)]
input_ref = data['ref'] if 'ref' in data.keys() else data['GT']
self.ref = [t.to(self.device) for t in torch.chunk(input_ref, chunks=self.batch_factor, dim=0)]
else:
self.hq = self.lq
self.ref = self.lq
self.dstate = {'lq': self.lq, 'hq': self.hq, 'ref': self.ref}
for k, v in data.items():
if k not in ['LQ', 'ref', 'GT'] and isinstance(v, torch.Tensor):
self.dstate[k] = [t.to(self.device) for t in torch.chunk(v, chunks=self.batch_factor, dim=0)]
def optimize_parameters(self, step):
# Some models need to make parametric adjustments per-step. Do that here.
for net in self.networks.values():
if hasattr(net.module, "update_for_step"):
net.module.update_for_step(step, os.path.join(self.opt['path']['models'], ".."))
# Iterate through the steps, performing them one at a time.
state = self.dstate
for step_num, s in enumerate(self.steps):
train_step = True
# 'every' is used to denote steps that should only occur at a certain integer factor rate. e.g. '2' occurs every 2 steps.
# Note that the injection points for the step might still be required, so address this by setting train_step=False
if 'every' in s.step_opt.keys() and step % s.step_opt['every'] != 0:
train_step = False
# Steps can opt out of early (or late) training, make sure that happens here.
if 'after' in s.step_opt.keys() and step < s.step_opt['after'] or 'before' in s.step_opt.keys() and step > s.step_opt['before']:
continue
# Steps can choose to not execute if a state key is missing.
if 'requires' in s.step_opt.keys():
requirements_met = True
for requirement in s.step_opt['requires']:
if requirement not in state.keys():
requirements_met = False
if not requirements_met:
continue
if train_step:
# Only set requires_grad=True for the network being trained.
nets_to_train = s.get_networks_trained()
enabled = 0
for name, net in self.networks.items():
net_enabled = name in nets_to_train
if net_enabled:
enabled += 1
# Networks can opt out of training before a certain iteration by declaring 'after' in their definition.
if 'after' in self.opt['networks'][name].keys() and step < self.opt['networks'][name]['after']:
net_enabled = False
for p in net.parameters():
if p.dtype != torch.int64 and p.dtype != torch.bool and not hasattr(p, "DO_NOT_TRAIN"):
p.requires_grad = net_enabled
else:
p.requires_grad = False
assert enabled == len(nets_to_train)
# Update experiments
[e.before_step(self.opt, self.step_names[step_num], self.env, nets_to_train, state) for e in self.experiments]
for o in s.get_optimizers():
o.zero_grad()
# Now do a forward and backward pass for each gradient accumulation step.
new_states = {}
for m in range(self.batch_factor):
ns = s.do_forward_backward(state, m, step_num, train=train_step)
for k, v in ns.items():
if k not in new_states.keys():
new_states[k] = [v]
else:
new_states[k].append(v)
# Push the detached new state tensors into the state map for use with the next step.
for k, v in new_states.items():
# State is immutable to reduce complexity. Overwriting existing state keys is not supported.
assert k not in state.keys()
state[k] = v
if train_step:
# And finally perform optimization.
[e.before_optimize(state) for e in self.experiments]
s.do_step(step)
[e.after_optimize(state) for e in self.experiments]
# Record visual outputs for usage in debugging and testing.
if 'visuals' in self.opt['logger'].keys() and self.rank <= 0 and step % self.opt['logger']['visual_debug_rate'] == 0:
sample_save_path = os.path.join(self.opt['path']['models'], "..", "visual_dbg")
for v in self.opt['logger']['visuals']:
if v not in state.keys():
continue # This can happen for several reasons (ex: 'after' defs), just ignore it.
for i, dbgv in enumerate(state[v]):
if 'recurrent_visual_indices' in self.opt['logger'].keys() and len(dbgv.shape)==5:
for rvi in self.opt['logger']['recurrent_visual_indices']:
rdbgv = dbgv[:, rvi]
if rdbgv.shape[1] > 3:
rdbgv = rdbgv[:, :3, :, :]
os.makedirs(os.path.join(sample_save_path, v), exist_ok=True)
utils.save_image(rdbgv.float(), os.path.join(sample_save_path, v, "%05i_%02i_%02i.png" % (step, rvi, i)))
else:
if dbgv.shape[1] > 3:
dbgv = dbgv[:,:3,:,:]
os.makedirs(os.path.join(sample_save_path, v), exist_ok=True)
utils.save_image(dbgv.float(), os.path.join(sample_save_path, v, "%05i_%02i.png" % (step, i)))
# Some models have their own specific visual debug routines.
for net_name, net in self.networks.items():
if hasattr(net.module, "visual_dbg"):
model_vdbg_dir = os.path.join(sample_save_path, net_name)
os.makedirs(model_vdbg_dir, exist_ok=True)
net.module.visual_dbg(step, model_vdbg_dir)
def compute_fea_loss(self, real, fake):
with torch.no_grad():
logits_real = self.netF(real.to(self.device))
logits_fake = self.netF(fake.to(self.device))
return nn.L1Loss().to(self.device)(logits_fake, logits_real)
def test(self):
for net in self.netsG.values():
net.eval()
with torch.no_grad():
# This can happen one of two ways: Either a 'validation injector' is provided, in which case we run that.
# Or, we run the entire chain of steps in "train" mode and use eval.output_state.
if 'injectors' in self.opt['eval'].keys():
state = {}
for inj in self.opt['eval']['injectors'].values():
# Need to move from mega_batch mode to batch mode (remove chunks)
for k, v in self.dstate.items():
state[k] = v[0]
inj = create_injector(inj, self.env)
state.update(inj(state))
else:
# Iterate through the steps, performing them one at a time.
state = self.dstate
for step_num, s in enumerate(self.steps):
ns = s.do_forward_backward(state, 0, step_num, train=False)
for k, v in ns.items():
state[k] = [v]
self.eval_state = {}
for k, v in state.items():
if isinstance(v, list):
self.eval_state[k] = [s.detach().cpu() if isinstance(s, torch.Tensor) else s for s in v]
else:
self.eval_state[k] = [v.detach().cpu() if isinstance(v, torch.Tensor) else v]
for net in self.netsG.values():
net.train()
# Fetches a summary of the log.
def get_current_log(self, step):
log = {}
for s in self.steps:
log.update(s.get_metrics())
for e in self.experiments:
log.update(e.get_log_data())
# Some generators can do their own metric logging.
for net_name, net in self.networks.items():
if hasattr(net.module, "get_debug_values"):
log.update(net.module.get_debug_values(step, net_name))
return log
def get_current_visuals(self, need_GT=True):
# Conforms to an archaic format from MMSR.
return {'LQ': self.eval_state['lq'][0].float().cpu(),
'GT': self.eval_state['hq'][0].float().cpu(),
'rlt': self.eval_state[self.opt['eval']['output_state']][0].float().cpu()}
def print_network(self):
for name, net in self.networks.items():
s, n = self.get_network_description(net)
net_struc_str = '{}'.format(net.__class__.__name__)
if self.rank <= 0:
logger.info('Network {} structure: {}, with parameters: {:,d}'.format(name, net_struc_str, n))
logger.info(s)
def load(self):
for netdict in [self.netsG, self.netsD]:
for name, net in netdict.items():
if not self.opt['networks'][name]['trainable']:
continue
load_path = self.opt['path']['pretrain_model_%s' % (name,)]
if load_path is not None:
if self.rank <= 0:
logger.info('Loading model for [%s]' % (load_path,))
self.load_network(load_path, net, self.opt['path']['strict_load'])
def save(self, iter_step):
for name, net in self.networks.items():
# Don't save non-trainable networks.
if self.opt['networks'][name]['trainable']:
self.save_network(net, name, iter_step)
def force_restore_swapout(self):
# Legacy method. Do nothing.
pass