imaginAIry/imaginairy/utils.py

177 lines
4.8 KiB
Python
Raw Normal View History

2022-09-08 03:59:30 +00:00
import importlib
import logging
import os.path
import platform
from contextlib import contextmanager
2022-09-08 03:59:30 +00:00
from functools import lru_cache
from typing import List, Optional
2022-09-08 03:59:30 +00:00
import numpy as np
import requests
2022-09-08 03:59:30 +00:00
import torch
from PIL import Image, ImageFilter
from torch import Tensor
2022-09-16 16:24:24 +00:00
from torch.nn import functional
2022-09-13 07:46:37 +00:00
from torch.overrides import handle_torch_function, has_torch_function_variadic
from transformers import cached_path
2022-09-08 03:59:30 +00:00
from imaginairy.img_log import log_img
logger = logging.getLogger(__name__)
2022-09-08 03:59:30 +00:00
@lru_cache()
def get_device():
if torch.cuda.is_available():
return "cuda"
2022-09-16 16:24:24 +00:00
if torch.backends.mps.is_available():
2022-09-17 19:24:27 +00:00
return "mps:0"
2022-09-16 16:24:24 +00:00
return "cpu"
2022-09-08 03:59:30 +00:00
@lru_cache()
def get_device_name(device_type):
if device_type == "cuda":
return torch.cuda.get_device_name(0)
return platform.processor()
def log_params(model):
2022-09-08 03:59:30 +00:00
total_params = sum(p.numel() for p in model.parameters())
logger.debug(f"{model.__class__.__name__} has {total_params * 1.e-6:.2f} M params.")
2022-09-08 03:59:30 +00:00
def instantiate_from_config(config):
2022-09-16 16:24:24 +00:00
if "target" not in config:
2022-09-08 03:59:30 +00:00
if config == "__is_first_stage__":
return None
2022-09-16 16:24:24 +00:00
if config == "__is_unconditional__":
2022-09-08 03:59:30 +00:00
return None
raise KeyError("Expected key `target` to instantiate.")
2022-09-16 16:24:24 +00:00
return get_obj_from_str(config["target"])(**config.get("params", {}))
2022-09-08 03:59:30 +00:00
def get_obj_from_str(string, reload=False):
module, cls = string.rsplit(".", 1)
if reload:
module_imp = importlib.import_module(module)
importlib.reload(module_imp)
return getattr(importlib.import_module(module, package=None), cls)
def _fixed_layer_norm(
2022-09-16 16:24:24 +00:00
input: Tensor, # noqa
normalized_shape: List[int],
weight: Optional[Tensor] = None,
bias: Optional[Tensor] = None,
eps: float = 1e-5,
) -> Tensor:
2022-09-16 16:24:24 +00:00
"""
Applies Layer Normalization for last certain number of dimensions.
See :class:`~torch.nn.LayerNorm` for details.
"""
if has_torch_function_variadic(input, weight, bias):
return handle_torch_function(
_fixed_layer_norm,
(input, weight, bias),
input,
normalized_shape,
weight=weight,
bias=bias,
eps=eps,
)
return torch.layer_norm(
input.contiguous(),
normalized_shape,
weight,
bias,
eps,
torch.backends.cudnn.enabled,
)
@contextmanager
def fix_torch_nn_layer_norm():
"""https://github.com/CompVis/stable-diffusion/issues/25#issuecomment-1221416526"""
orig_function = functional.layer_norm
functional.layer_norm = _fixed_layer_norm
try:
yield
finally:
functional.layer_norm = orig_function
def expand_mask(mask_image, size):
if size < 0:
threshold = 0.95
else:
threshold = 0.05
mask_image = mask_image.convert("L")
mask_image = mask_image.filter(ImageFilter.GaussianBlur(size))
log_img(mask_image, "init mask blurred")
mask = np.array(mask_image)
mask = mask.astype(np.float32) / 255.0
mask = mask[None, None]
mask[mask < threshold] = 0
mask[mask >= threshold] = 1
return Image.fromarray(np.uint8(mask.squeeze() * 255))
def img_path_to_torch_image(path):
image = Image.open(path).convert("RGB")
logger.info(f"Loaded input 🖼 of size {image.size} from {path}")
return pillow_img_to_torch_image(image)
def pillow_fit_image_within(image, max_height=512, max_width=512):
image = image.convert("RGB")
w, h = image.size
resize_ratio = min(max_width / w, max_height / h)
w, h = int(w * resize_ratio), int(h * resize_ratio)
w, h = map(lambda x: x - x % 64, (w, h)) # resize to integer multiple of 64
image = image.resize((w, h), resample=Image.Resampling.NEAREST)
return image, w, h
def pillow_img_to_torch_image(image):
image = image.convert("RGB")
image = np.array(image).astype(np.float32) / 255.0
image = image[None].transpose(0, 3, 1, 2)
image = torch.from_numpy(image)
return 2.0 * image - 1.0
def get_cache_dir():
xdg_cache_home = os.getenv("XDG_CACHE_HOME", None)
if xdg_cache_home is None:
user_home = os.getenv("HOME", None)
if user_home:
xdg_cache_home = os.path.join(user_home, ".cache")
if xdg_cache_home is not None:
return os.path.join(xdg_cache_home, "imaginairy", "weights")
return os.path.join(os.path.dirname(__file__), ".cached-downloads")
def get_cached_url_path(url):
try:
return cached_path(url)
except OSError:
pass
filename = url.split("/")[-1]
dest = get_cache_dir()
os.makedirs(dest, exist_ok=True)
dest_path = os.path.join(dest, filename)
if os.path.exists(dest_path):
return dest_path
2022-09-16 16:24:24 +00:00
r = requests.get(url) # noqa
with open(dest_path, "wb") as f:
f.write(r.content)
return dest_path