File size: 6,453 Bytes
6142a25 |
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 |
import os
from omegaconf import OmegaConf
import numpy as np
import torch
import torchvision
from PIL import Image
from pytorch_lightning.callbacks import Callback
from pytorch_lightning.utilities.distributed import rank_zero_only
from pytorch_lightning.utilities import rank_zero_info
import time
class CUDACallback(Callback):
# see https://github.com/SeanNaren/minGPT/blob/master/mingpt/callback.py
def on_train_epoch_start(self, trainer, pl_module):
# Reset the memory use counter
torch.cuda.reset_peak_memory_stats(trainer.root_gpu)
torch.cuda.synchronize(trainer.root_gpu)
self.start_time = time.time()
def on_train_epoch_end(self, trainer, pl_module, outputs):
torch.cuda.synchronize(trainer.root_gpu)
max_memory = torch.cuda.max_memory_allocated(trainer.root_gpu) / 2 ** 20
epoch_time = (time.time() - self.start_time)/3600
try:
max_memory = trainer.training_type_plugin.reduce(max_memory)
epoch_time = trainer.training_type_plugin.reduce(epoch_time)
rank_zero_info(f"Average Epoch time: {epoch_time:.2f} hours")
rank_zero_info(f"Average Peak memory {max_memory:.2f}MiB")
except AttributeError:
pass
class SetupCallback(Callback):
def __init__(self, resume, now, logdir, ckptdir, cfgdir, config, lightning_config):
super().__init__()
self.resume = resume
self.now = now
self.logdir = logdir
self.ckptdir = ckptdir
self.cfgdir = cfgdir
self.config = config
self.lightning_config = lightning_config
def on_keyboard_interrupt(self, trainer, pl_module):
if trainer.global_rank == 0:
print("Summoning checkpoint.")
ckpt_path = os.path.join(self.ckptdir, "last.ckpt")
trainer.save_checkpoint(ckpt_path)
def on_pretrain_routine_start(self, trainer, pl_module):
if trainer.global_rank == 0:
# Create logdirs and save configs
os.makedirs(self.logdir, exist_ok=True)
os.makedirs(self.ckptdir, exist_ok=True)
os.makedirs(self.cfgdir, exist_ok=True)
if "callbacks" in self.lightning_config:
if 'metrics_over_trainsteps_checkpoint' in self.lightning_config['callbacks']:
os.makedirs(os.path.join(self.ckptdir, 'trainstep_checkpoints'), exist_ok=True)
print("Project config")
print(OmegaConf.to_yaml(self.config))
OmegaConf.save(self.config,
os.path.join(self.cfgdir, "{}-project.yaml".format(self.now)))
print("Lightning config")
print(OmegaConf.to_yaml(self.lightning_config))
OmegaConf.save(OmegaConf.create({"lightning": self.lightning_config}),
os.path.join(self.cfgdir, "{}-lightning.yaml".format(self.now)))
else:
# ModelCheckpoint callback created log directory --- remove it
if not self.resume and os.path.exists(self.logdir):
dst, name = os.path.split(self.logdir)
dst = os.path.join(dst, "child_runs", name)
os.makedirs(os.path.split(dst)[0], exist_ok=True)
try:
os.rename(self.logdir, dst)
except FileNotFoundError:
pass
class ImageLogger(Callback):
def __init__(self, batch_frequency=2000, max_images=4, clamp=True, increase_log_steps=True,
rescale=True, disabled=False, log_on_batch_idx=False, log_first_step=False,
log_images_kwargs=None, fixed_input=False):
super().__init__()
self.rescale = rescale
self.batch_freq = batch_frequency
self.max_images = max_images
if not increase_log_steps:
self.log_steps = [self.batch_freq]
self.clamp = clamp
self.disabled = disabled
self.log_on_batch_idx = log_on_batch_idx
self.log_images_kwargs = log_images_kwargs if log_images_kwargs else {}
self.log_first_step = log_first_step
self.fixed_input = fixed_input
@rank_zero_only
def log_local(self, save_dir, split, images, global_step, current_epoch, batch_idx):
root = os.path.join(save_dir, "image_log", split)
for k in images:
grid = torchvision.utils.make_grid(images[k], nrow=4)
if self.rescale:
grid = (grid + 1.0) / 2.0 # -1,1 -> 0,1; c,h,w
grid = grid.transpose(0, 1).transpose(1, 2).squeeze(-1)
grid = grid.numpy()
grid = (grid * 255).astype(np.uint8)
filename = "{}_gs-{:06}_e-{:06}_b-{:06}.png".format(k, global_step, current_epoch, batch_idx)
path = os.path.join(root, filename)
os.makedirs(os.path.split(path)[0], exist_ok=True)
Image.fromarray(grid).save(path)
def log_img(self, pl_module, batch, batch_idx, split="train"):
check_idx = batch_idx # if self.log_on_batch_idx else pl_module.global_step
if (self.check_frequency(check_idx) and # batch_idx % self.batch_freq == 0
hasattr(pl_module, "log_images") and
callable(pl_module.log_images) and
self.max_images > 0):
logger = type(pl_module.logger)
is_train = pl_module.training
if is_train:
pl_module.eval()
with torch.no_grad():
images = pl_module.log_images(batch, fixed_input=self.fixed_input, split=split, **self.log_images_kwargs)
for k in images:
N = min(images[k].shape[0], self.max_images)
images[k] = images[k][:N]
if isinstance(images[k], torch.Tensor):
images[k] = images[k].detach().cpu()
if self.clamp:
images[k] = torch.clamp(images[k], -1., 1.)
self.log_local(pl_module.logger.save_dir, split, images,
pl_module.global_step, pl_module.current_epoch, batch_idx)
if is_train:
pl_module.train()
def check_frequency(self, check_idx):
return check_idx % self.batch_freq == 0
def on_train_batch_end(self, trainer, pl_module, outputs, batch, batch_idx, dataloader_idx):
if not self.disabled:
self.log_img(pl_module, batch, batch_idx, split="train")
|