Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 01:04:32 +02:00
Code Issues Packages Projects Releases Wiki Activity

VRAM efficient refiner loading for compiler

Browse Source
This commit is contained in:
Disty0
2023-07-11 11:00:45 +03:00
parent 6466d3cb02
commit a844a83d9d
1 changed files with 16 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/sd_models.py
+16 -12
View File
@@ -591,7 +591,7 @@ def load_diffuser(checkpoint_info=None, already_loaded_state_dict=None, timer=No
shared.log.debug(f'Diffusers load config: {diffusers_load_config}')
sd_model = None
base_sent_to_cpu=False
try:
devices.set_cuda_params()
if shared.cmd_opts.ckpt is not None and model_data.initial: # initial load
@@ -715,20 +715,20 @@ def load_diffuser(checkpoint_info=None, already_loaded_state_dict=None, timer=No
shared.log.debug('Diffusers: enable channels last')
sd_model.unet.to(memory_format=torch.channels_last)
if devices.backend == 'ipex':
sd_model.unet.training = False
if shared.opts.cuda_compile and shared.opts.cuda_compile_mode == 'ipex':
shared.log.info("Model compile enabled: IPEX Optimize Graph Mode")
base_sent_to_cpu=False
if (shared.opts.cuda_compile and torch.cuda.is_available()) or devices.backend == 'ipex':
if op == 'refiner':
gpu_vram = memory_stats().get('gpu', {})
if not shared.opts.diffusers_move_base and (gpu_vram.get('total', 0) - gpu_vram.get('used', 0)) >= 7 if "StableDiffusionXL" in sd_model.__class__.__name__ else 3.5:
free_vram = gpu_vram.get('total', 0) - gpu_vram.get('used', 0)
refiner_enough_vram = free_vram >= 7 if "StableDiffusionXL" in sd_model.__class__.__name__ else 3
if not shared.opts.diffusers_move_base and refiner_enough_vram:
sd_model.to(devices.device)
base_sent_to_cpu=False
else:
if shared.opts.diffusers_move_base:
pass
else:
shared.log.info(f"Not enough VRAM to optimize refiner, using RAM as fallback. Free VRAM: {gpu_vram.get('total', 0) - gpu_vram.get('used', 0)} GB")
if not refiner_enough_vram and not (shared.opts.diffusers_move_base and shared.opts.diffusers_move_refiner):
shared.log.warning(f"Not enough VRAM to use refiner, using RAM as fallback. Free VRAM: {free_vram} GB\n"
+ "Enabled 'Move base model to CPU when using refiner' and 'Move refiner model to CPU when not in use'\n"
+ "Enable the settings above and apply the settings to suppress this warning.")
shared.opts.diffusers_move_base=True
shared.opts.diffusers_move_refiner=True
shared.log.debug('Moving base model to CPU')
@@ -738,12 +738,16 @@ def load_diffuser(checkpoint_info=None, already_loaded_state_dict=None, timer=No
base_sent_to_cpu=True
else:
sd_model.to(devices.device)
if devices.backend == 'ipex':
sd_model.unet.training = False
if shared.opts.cuda_compile and shared.opts.cuda_compile_mode == 'ipex':
shared.log.info("Model compile enabled: IPEX Optimize Graph Mode")
sd_model.unet = torch.xpu.optimize(sd_model.unet, dtype=devices.dtype, auto_kernel_selection=True, optimize_lstm=True, # pylint: disable=attribute-defined-outside-init
graph_mode=True if shared.opts.cuda_compile and shared.opts.cuda_compile_mode == 'ipex' else False)
shared.log.info("Applied IPEX Optimize")
if shared.opts.cuda_compile and torch.cuda.is_available():
sd_model.to(devices.device)
elif shared.opts.cuda_compile and torch.cuda.is_available():
import torch._dynamo # pylint: disable=unused-import,redefined-outer-name
log_level = logging.WARNING if shared.opts.cuda_compile_verbose else logging.CRITICAL # pylint: disable=protected-access
torch._logging.set_logs(dynamo=log_level, aot=log_level, inductor=log_level) # pylint: disable=protected-access