Public/automatic
1
0
Fork 0
mirror of https://github.com/vladmandic/automatic synced 2026-09-19 09:14:35 +02:00
Code Issues Packages Projects Releases Wiki Activity

Add --use-openvino

Browse Source
This commit is contained in:
Disty0
2023-08-22 03:02:19 +03:00
parent ebf7b90e3e
commit e70520efa9
7 changed files with 44 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/cmd_args.py
+1
View File
@@ -41,6 +41,7 @@ group.add_argument("--profile", action='store_true', help="Run profiler, default
group.add_argument("--disable-queue", action='store_true', help="Disable queues, default: %(default)s")
group.add_argument('--debug', default = False, action='store_true', help = "Run installer with debug logging, default: %(default)s")
group.add_argument('--use-directml', default = False, action='store_true', help = "Use DirectML if no compatible GPU is detected, default: %(default)s")
group.add_argument("--use-openvino", default = False, action='store_true', help="Use Intel OpenVINO backend, default: %(default)s")
group.add_argument("--use-ipex", default = False, action='store_true', help="Force use Intel OneAPI XPU backend, default: %(default)s")
group.add_argument("--use-cuda", default=False, action='store_true', help="Force use nVidia CUDA backend, default: %(default)s")
group.add_argument("--use-rocm", default=False, action='store_true', help="Force use AMD ROCm backend, default: %(default)s")
modules/intel/openvino/__init__.py
+12 -1
View File
@@ -8,7 +8,6 @@ from torch._dynamo.backends.registry import register_backend
from torch.fx.experimental.proxy_tensor import make_fx
from torch._inductor.compile_fx import compile_fx
from hashlib import sha256
import modules.shared
class ModelState:
def __init__(self):
@@ -113,3 +112,15 @@ def get_cached_file_name(*args, model_hash_str, device, cache_root):
file_name = None
model_hash_str = None
return file_name
def openvino_override_opts():
from modules import shared
if not shared.opts.cuda_compile:
shared.log.warn("OpenVINO: Enabling Torch Compile")
shared.opts.cuda_compile = True
if shared.opts.cuda_compile_backend != "openvino_fx":
shared.log.warn("OpenVINO: Setting Torch Compiler backend to OpenVINO FX")
shared.opts.cuda_compile_backend = "openvino_fx"
if shared.opts.sd_backend != "diffusers":
shared.log.warn("OpenVINO: Setting backend to Diffusers")
shared.opts.sd_backend = "diffusers"
modules/shared.py
+12 -9
View File
@@ -184,8 +184,11 @@ class State:
state = State()
state.server_start = time.time()
backend = Backend.DIFFUSERS if (cmd_opts.backend is not None) and (cmd_opts.backend.lower() == 'diffusers') else Backend.ORIGINAL # initial since we don't have opts loaded yet
if cmd_opts.use_openvino:
backend = Backend.DIFFUSERS
cmd_opts.backend = 'diffusers'
else:
backend = Backend.DIFFUSERS if (cmd_opts.backend is not None) and (cmd_opts.backend.lower() == 'diffusers') else Backend.ORIGINAL # initial since we don't have opts loaded yet
class OptionInfo:
@@ -342,7 +345,7 @@ else: # cuda
cross_attention_optimization_default ="Scaled-Dot-Product"
options_templates.update(options_section(('sd', "Stable Diffusion"), {
"sd_backend": OptionInfo("original", "Stable Diffusion backend", gr.Radio, lambda: {"choices": ["original", "diffusers"] }),
"sd_backend": OptionInfo("diffusers" if cmd_opts.use_openvino else "original", "Stable Diffusion backend", gr.Radio, lambda: {"choices": ["original", "diffusers"] }),
"sd_checkpoint_autoload": OptionInfo(True, "Stable Diffusion checkpoint autoload on server start"),
"sd_model_checkpoint": OptionInfo(default_checkpoint, "Stable Diffusion checkpoint", gr.Dropdown, lambda: {"choices": list_checkpoint_tiles()}, refresh=refresh_checkpoints),
"sd_model_refiner": OptionInfo('None', "Stable Diffusion refiner", gr.Dropdown, lambda: {"choices": ['None'] + list_checkpoint_tiles()}, refresh=refresh_checkpoints),
@@ -373,7 +376,7 @@ options_templates.update(options_section(('optimizations', "Optimizations"), {
options_templates.update(options_section(('cuda', "Compute Settings"), {
"memmon_poll_rate": OptionInfo(2, "VRAM usage polls per second during generation", gr.Slider, {"minimum": 0, "maximum": 40, "step": 1}),
"precision": OptionInfo("Autocast", "Precision type", gr.Radio, lambda: {"choices": ["Autocast", "Full"]}),
"cuda_dtype": OptionInfo("FP32" if sys.platform == "darwin" else "BF16" if devices.backend == "ipex" else "FP16", "Device precision type", gr.Radio, lambda: {"choices": ["FP32", "FP16", "BF16"]}),
"cuda_dtype": OptionInfo("FP32" if sys.platform == "darwin" or cmd_opts.use_openvino else "BF16" if devices.backend == "ipex" else "FP16", "Device precision type", gr.Radio, lambda: {"choices": ["FP32", "FP16", "BF16"]}),
"no_half": OptionInfo(False, "Use full precision for model (--no-half)", None, None, None),
"no_half_vae": OptionInfo(False, "Use full precision for VAE (--no-half-vae)"),
"upcast_sampling": OptionInfo(True if sys.platform == "darwin" else False, "Enable upcast sampling"),
@@ -385,8 +388,8 @@ options_templates.update(options_section(('cuda', "Compute Settings"), {
"cudnn_benchmark": OptionInfo(False, "Enable full-depth cuDNN benchmark feature"),
# "cuda_allow_tf32": OptionInfo(True, "Allow TF32 math ops"),
# "cuda_allow_tf16_reduced": OptionInfo(True, "Allow TF16 reduced precision math ops"),
"cuda_compile": OptionInfo(False, "Enable model compile (experimental)"),
"cuda_compile_backend": OptionInfo("none", "Model compile backend (experimental)", gr.Radio, lambda: {"choices": ['none', 'inductor', 'cudagraphs', 'aot_ts_nvfuser', 'hidet', 'ipex', 'openvino_fx']}),
"cuda_compile": OptionInfo(True if cmd_opts.use_openvino else False, "Enable model compile (experimental)"),
"cuda_compile_backend": OptionInfo("openvino_fx" if cmd_opts.use_openvino else "none", "Model compile backend (experimental)", gr.Radio, lambda: {"choices": ['none', 'inductor', 'cudagraphs', 'aot_ts_nvfuser', 'hidet', 'ipex', 'openvino_fx']}),
"cuda_compile_mode": OptionInfo("default", "Model compile mode (experimental)", gr.Radio, lambda: {"choices": ['default', 'reduce-overhead', 'max-autotune']}),
"cuda_compile_fullgraph": OptionInfo(False, "Model compile fullgraph"),
"cuda_compile_verbose": OptionInfo(False, "Model compile verbose mode"),
@@ -801,9 +804,9 @@ config_filename = cmd_opts.config
opts.load(config_filename)
cmd_opts = cmd_args.compatibility_args(opts, cmd_opts)
if cmd_opts.backend is None:
backend = Backend.DIFFUSERS if opts.data.get('sd_backend', 'original') == 'diffusers' else Backend.ORIGINAL
backend = Backend.DIFFUSERS if cmd_opts.use_openvino or opts.data.get('sd_backend', 'original') == 'diffusers' else Backend.ORIGINAL
else:
backend = Backend.DIFFUSERS if cmd_opts.backend.lower() == 'diffusers' else Backend.ORIGINAL
backend = Backend.DIFFUSERS if cmd_opts.use_openvino or cmd_opts.backend.lower() == 'diffusers' else Backend.ORIGINAL
opts.data['sd_backend'] = 'diffusers' if backend == Backend.DIFFUSERS else 'original'
opts.data['uni_pc_lower_order_final'] = opts.schedulers_use_loworder
opts.data['uni_pc_order'] = opts.schedulers_solver_order
@@ -997,7 +1000,7 @@ class Shared(sys.modules[__name__].__class__): # this class is here to provide s
@property
def backend(self):
return Backend.ORIGINAL if opts.data['sd_backend'] == 'original' else Backend.DIFFUSERS
return Backend.ORIGINAL if not cmd_opts.use_openvino and opts.data['sd_backend'] == 'original' else Backend.DIFFUSERS
@property
def sd_model_type(self):
modules/ui.py
+3
View File
@@ -15,6 +15,7 @@ from modules.ui_components import FormRow, FormColumn, FormGroup, ToolButton, Fo
from modules.paths import script_path, data_path
from modules.shared import opts, cmd_opts, readfile
from modules.dml import directml_override_opts
from modules.intel.openvino import openvino_override_opts
from modules import prompt_parser
import modules.codeformer_model
import modules.generation_parameters_copypaste as parameters_copypaste
@@ -966,6 +967,8 @@ def create_ui(startup_timer = None):
changed.append(key)
if cmd_opts.use_directml:
directml_override_opts()
if cmd_opts.use_openvino:
openvino_override_opts()
try:
opts.save(modules.shared.config_filename)
log.info(f'Settings changed: {len(changed)} {changed}')