imaginAIry/imaginairy/utils/log_utils.py

463 lines
14 KiB
Python

"""Utilities for image generation logging"""
import logging
import logging.config
import re
import time
import warnings
from typing import Callable
import torch.cuda
from imaginairy.utils.memory_tracker import TorchRAMTracker
_CURRENT_LOGGING_CONTEXT = None
logger = logging.getLogger(__name__)
def log_conditioning(conditioning, description):
if _CURRENT_LOGGING_CONTEXT is None:
return
_CURRENT_LOGGING_CONTEXT.log_conditioning(conditioning, description)
def log_latent(latents, description):
if _CURRENT_LOGGING_CONTEXT is None:
return
if latents is None:
return
_CURRENT_LOGGING_CONTEXT.log_latents(latents, description)
def log_img(img, description):
if _CURRENT_LOGGING_CONTEXT is None:
return
_CURRENT_LOGGING_CONTEXT.log_img(img, description)
def log_progress_latent(latent):
if _CURRENT_LOGGING_CONTEXT is None:
return
_CURRENT_LOGGING_CONTEXT.log_progress_latent(latent)
def log_tensor(t, description=""):
if _CURRENT_LOGGING_CONTEXT is None:
return
_CURRENT_LOGGING_CONTEXT.log_img(t, description)
def increment_step():
if _CURRENT_LOGGING_CONTEXT is None:
return
_CURRENT_LOGGING_CONTEXT.step_count += 1
class TimingContext:
"""Tracks time and memory usage of a block of code"""
def __init__(
self,
description: str,
device: str | None = None,
callback_fn: Callable | None = None,
):
from imaginairy.utils import get_device
self.description = description
self._device = device or get_device()
self.callback_fn = callback_fn
self.start_time = None
self.end_time = None
self.duration = 0
self.memory_context = None
self.memory_start = None
self.memory_end = 0
self.memory_peak = 0
self.memory_peak_delta = 0
def start(self):
# supports repeated calls to start/stop
if self._device == "cuda":
self.memory_context = TorchRAMTracker(self.description)
self.memory_context.start()
if self.memory_start is None:
self.memory_start = self.memory_context.start_memory
self.end_time = None
self.start_time = time.time()
def stop(self):
# supports repeated calls to start/stop
self.end_time = time.time()
self.duration += self.end_time - self.start_time
if self._device == "cuda":
self.memory_context.stop()
self.memory_end = self.memory_context.end_memory
self.memory_peak = max(self.memory_context.peak_memory, self.memory_peak)
self.memory_peak_delta = max(
self.memory_context.peak_memory_delta, self.memory_peak_delta
)
if self.callback_fn is not None:
self.callback_fn(self)
def __enter__(self):
self.start()
return self
def __exit__(self, exc_type, exc_value, traceback):
self.stop()
class ImageLoggingContext:
def __init__(
self,
prompt,
model=None,
debug_img_callback=None,
img_outdir=None,
progress_img_callback=None,
progress_img_interval_steps=3,
progress_img_interval_min_s=0.1,
progress_latent_callback=None,
):
self.prompt = prompt
self.model = model
self.step_count = 0
self.image_count = 0
self.debug_img_callback = debug_img_callback
self.img_outdir = img_outdir
self.progress_img_callback = progress_img_callback
self.progress_img_interval_steps = progress_img_interval_steps
self.progress_img_interval_min_s = progress_img_interval_min_s
self.progress_latent_callback = progress_latent_callback
self.summary_context = TimingContext("total")
self.summary_context.start()
self.timing_contexts = {}
self.last_progress_img_ts = 0
self.last_progress_img_step = -1000
self._prev_log_context = None
def __enter__(self):
self.start()
def __exit__(self, exc_type, exc_val, exc_tb):
self.stop()
def start(self):
global _CURRENT_LOGGING_CONTEXT
self._prev_log_context = _CURRENT_LOGGING_CONTEXT
_CURRENT_LOGGING_CONTEXT = self
return self
def stop(self):
global _CURRENT_LOGGING_CONTEXT
_CURRENT_LOGGING_CONTEXT = self._prev_log_context
def timing(self, description):
if description not in self.timing_contexts:
def cb(context):
self.timing_contexts[description] = context
tc = TimingContext(description, callback_fn=cb)
self.timing_contexts[description] = tc
return self.timing_contexts[description]
def get_performance_stats(self) -> dict[str, dict[str, float]]:
# calculate max peak seen in any timing context
self.summary_context.stop()
self.timing_contexts["total"] = self.summary_context
# move total to the end
self.timing_contexts["total"] = self.timing_contexts.pop("total")
if torch.cuda.is_available():
self.summary_context.memory_peak = max(
max(context.memory_peak, context.memory_start, context.memory_end)
for context in self.timing_contexts.values()
)
performance_stats = {}
for context in self.timing_contexts.values():
performance_stats[context.description] = {
"duration": context.duration,
"memory_start": context.memory_start,
"memory_end": context.memory_end,
"memory_peak": context.memory_peak,
"memory_peak_delta": context.memory_peak_delta,
}
return performance_stats
def log_conditioning(self, conditioning, description):
if not self.debug_img_callback:
return
img = conditioning_to_img(conditioning)
self.debug_img_callback(
img, description, self.image_count, self.step_count, self.prompt
)
def log_latents(self, latents, description):
if "predicted_latent" in description:
if self.progress_latent_callback is not None:
self.progress_latent_callback(latents)
if (
self.step_count - self.last_progress_img_step
) > self.progress_img_interval_steps and (
time.perf_counter() - self.last_progress_img_ts
> self.progress_img_interval_min_s
):
self.log_progress_latent(latents)
self.last_progress_img_step = self.step_count
self.last_progress_img_ts = time.perf_counter()
if not self.debug_img_callback:
return
if latents.shape[1] != 4:
# logger.info(f"Didn't save tensor of shape {samples.shape} for {description}")
return
try:
shape_str = ",".join(tuple(latents.shape))
except TypeError:
shape_str = str(latents.shape)
description = f"{description}-{shape_str}"
for latent in latents:
self.image_count += 1
latent = latent.unsqueeze(0)
img = latent_to_raw_image(latent)
self.debug_img_callback(
img, description, self.image_count, self.step_count, self.prompt
)
# for img in model_latents_to_pillow_imgs(latents):
# self.image_count += 1
# self.debug_img_callback(
# img, description, self.image_count, self.step_count, self.prompt
# )
def log_img(self, img, description):
if not self.debug_img_callback:
return
import torch
from torchvision.transforms import ToPILImage
self.image_count += 1
if isinstance(img, torch.Tensor):
img = ToPILImage()(img.squeeze().cpu().detach())
img = img.copy()
self.debug_img_callback(
img, description, self.image_count, self.step_count, self.prompt
)
def log_progress_latent(self, latent):
from imaginairy.utils.img_utils import model_latents_to_pillow_imgs
if not self.progress_img_callback:
return
for img in model_latents_to_pillow_imgs(latent):
self.progress_img_callback(img)
def log_tensor(self, t, description=""):
if not self.debug_img_callback:
return
if len(t.shape) == 2:
self.log_img(t, description)
def log_indexed_graph_of_tensor(self):
pass
# def img_callback(self, img, description, step_count, prompt):
# steps_path = os.path.join(self.img_outdir, "steps", f"{self.file_num:08}_S{prompt.seed}")
# os.makedirs(steps_path, exist_ok=True)
# filename = f"{self.file_num:08}_S{prompt.seed}_step{step_count:04}_{filesafe_text(description)[:40]}.jpg"
# destination = os.path.join(steps_path, filename)
# draw = ImageDraw.Draw(img)
# draw.text((10, 10), str(description))
# img.save(destination)
def filesafe_text(t):
return re.sub(r"[^a-zA-Z0-9.,\[\]() -]+", "_", t)[:130]
def conditioning_to_img(conditioning):
from torchvision.transforms import ToPILImage
return ToPILImage()(conditioning)
class ColorIndentingFormatter(logging.Formatter):
RED = "\033[31m"
GREEN = "\033[32m"
YELLOW = "\033[33m"
RESET = "\033[0m"
def format(self, record):
s = super().format(record)
color = ""
reset = ""
if record.levelno >= logging.ERROR:
color = self.RED
elif record.levelno >= logging.WARNING:
color = self.YELLOW
if _CURRENT_LOGGING_CONTEXT is not None:
s = f" {s}"
if color is None and not s.startswith(" "):
color = self.GREEN
if color:
reset = self.RESET
s = f"{color}{s}{reset}"
return s
def configure_logging(level="INFO"):
fmt = "%(message)s"
if level == "DEBUG":
fmt = "%(asctime)s [%(levelname)s] %(name)s:%(lineno)d: %(message)s"
LOGGING_CONFIG = {
"version": 1,
"disable_existing_loggers": True,
"formatters": {
"standard": {
"format": fmt,
"class": "imaginairy.utils.log_utils.ColorIndentingFormatter",
},
},
"handlers": {
"default": {
"level": level,
"formatter": "standard",
"class": "logging.StreamHandler",
"stream": "ext://sys.stdout", # Default is stderr
},
},
"loggers": {
"": { # root logger
"handlers": ["default"],
"level": "WARNING",
"propagate": False,
},
"imaginairy": {"handlers": ["default"], "level": level, "propagate": False},
"transformers.modeling_utils": {
"handlers": ["default"],
"level": "ERROR",
"propagate": False,
},
# disable https://github.com/huggingface/transformers/blob/17a55534f5e5df10ac4804d4270bf6b8cc24998d/src/transformers/models/clip/configuration_clip.py#L330
"transformers.models.clip.configuration_clip": {
"handlers": ["default"],
"level": "ERROR",
"propagate": False,
},
# disable the stupid triton is not available messages
# https://github.com/facebookresearch/xformers/blob/6425fd0cacb1a6579aa2f0c4a570b737cb10e9c3/xformers/__init__.py#L52
"xformers": {"handlers": ["default"], "level": "ERROR", "propagate": False},
},
}
suppress_annoying_logs_and_warnings()
logging.config.dictConfig(LOGGING_CONFIG)
def disable_transformers_custom_logging():
from transformers.modeling_utils import logger as modeling_logger
from transformers.utils.logging import _configure_library_root_logger
_configure_library_root_logger()
_logger = modeling_logger.parent
_logger.handlers = []
_logger.propagate = True
_logger.setLevel(logging.NOTSET)
modeling_logger.handlers = []
modeling_logger.propagate = True
modeling_logger.setLevel(logging.ERROR)
def disable_pytorch_lighting_custom_logging():
try:
from pytorch_lightning import _logger as pytorch_logger
except ImportError:
return
try:
from pytorch_lightning.utilities.seed import log
log.setLevel(logging.NOTSET)
log.handlers = []
log.propagate = False
except ImportError:
pass
pytorch_logger.setLevel(logging.NOTSET)
def disable_common_warnings():
warnings.filterwarnings(
"ignore",
category=UserWarning,
message=r"The operator .*?is not currently supported.*",
)
warnings.filterwarnings(
"ignore", category=UserWarning, message=r"The parameter 'pretrained' is.*"
)
warnings.filterwarnings(
"ignore", category=UserWarning, message=r"Arguments other than a weight.*"
)
warnings.filterwarnings("ignore", category=DeprecationWarning)
def suppress_annoying_logs_and_warnings():
disable_transformers_custom_logging()
# disable_pytorch_lighting_custom_logging()
disable_common_warnings()
def latent_to_raw_image(tensor):
"""
Converts a tensor of size (1, 4, x, y) into a PIL image of size (x*4, y*4).
Args:
tensor (numpy.ndarray): A tensor of size (1, 4, x, y).
Returns:
PIL.Image: An image representing the tensor.
"""
from PIL import Image
if tensor.ndim != 4 or tensor.shape[0] != 1:
msg = f"Tensor must be of shape (1, c, x, y). got shape: {tensor.shape}"
raise ValueError(msg)
_, c, x, y = tensor.shape
full_image = Image.new("L", (x, y * c))
# Process each channel
for i in range(c):
# Extract the channel
channel = tensor[0, i, :, :]
# Normalize and convert to an image
channel_image = Image.fromarray(
(channel / channel.max() * 255).cpu().numpy().astype("uint8")
)
# Paste the channel image into the full image
full_image.paste(channel_image, (0, i * y))
return full_image