Mirrored_Repos/InvokeAI
1
0
Fork 0
mirror of https://github.com/invoke-ai/InvokeAI synced 2024-08-30 20:32:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add slider for VRAM cache in configure script (#4133)

Browse Source 
## What type of PR is this? (check all applicable)

- [X ] Feature

## Have you discussed this change with the InvokeAI team?
- [X] Yes
- [ ] No, because:

      
## Have you updated all relevant documentation?
- [ ] Yes
- [X] No - will be in release notes

## Description

On CUDA systems, this PR adds a new slider to the install-time configure
script for adjusting the VRAM cache and suggests a good starting value
based on the user's max VRAM (this is subject to verification).

On non-CUDA systems this slider is suppressed.

Please test on both CUDA and non-CUDA systems using:
```
invokeai-configure --root ~/invokeai-main/ --skip-sd --skip-support
```

To see and test the default values, move `invokeai.yaml` out of the way
before running.

**Note added 8 August 2023**

This PR also fixes the configure and model install scripts so that if
the window is too small to fit the user interface, the user will be
prompted to interactively resize the window and/or change font size
(with the option to give up). This will prevent `npyscreen` from
generating its horrible tracebacks.

## Related Tickets & Documents

<!--
For pull requests that relate or close an issue, please include them
below. 

For example having the text: "closes #1234" would connect the current
pull
request to issue 1234.  And when we merge the pull request, Github will
automatically close the issue.
-->

- Related Issue #
- Closes #

## QA Instructions, Screenshots, Recordings

<!-- 
Please provide steps on how to test changes, any hardware or 
software specifications as well as any other pertinent information. 
-->

## Added/updated tests?

- [ ] Yes
- [ ] No : _please replace this line with details on why tests
      have not been included_

## [optional] Are there any post deployment tasks we need to perform?
This commit is contained in:
Millun Atluri 2023-08-09 12:27:54 +10:00 committed by GitHub
commit 37c9b85549
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 96 additions and 38 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
invokeai/app/services/config.py
View File

@ -24,7 +24,7 @@ InvokeAI:
sequential_guidance: false
precision: float16
max_cache_size: 6
max_vram_cache_size: 2.7
max_vram_cache_size: 0.5
always_use_cpu: false
free_gpu_mem: false
Features:
@ -172,6 +172,7 @@ from typing import ClassVar, Dict, List, Set, Literal, Union, get_origin, get_ty
INIT_FILE = Path("invokeai.yaml")
DB_FILE = Path("invokeai.db")
LEGACY_INIT_FILE = Path("invokeai.init")
DEFAULT_MAX_VRAM = 0.5
class InvokeAISettings(BaseSettings):

57
invokeai/backend/install/invokeai_configure.py
View File

@ -10,12 +10,15 @@ import sys
import argparse
import io
import os
import psutil
import shutil
import textwrap
import torch
import traceback
import yaml
import warnings
from argparse import Namespace
from enum import Enum
from pathlib import Path
from shutil import get_terminal_size
from typing import get_type_hints
@ -55,6 +58,7 @@ from invokeai.frontend.install.widgets import (
CyclingForm,
MIN_COLS,
MIN_LINES,
WindowTooSmallException,
)
from invokeai.backend.install.legacy_arg_parsing import legacy_parser
from invokeai.backend.install.model_install_backend import (
@ -79,6 +83,13 @@ Default_config_file = config.model_conf_path
SD_Configs = config.legacy_conf_path
PRECISION_CHOICES = ["auto", "float16", "float32"]
GB = 1073741824 # GB in bytes
HAS_CUDA = torch.cuda.is_available()
_, MAX_VRAM = torch.cuda.mem_get_info() if HAS_CUDA else (0, 0)
MAX_VRAM /= GB
MAX_RAM = psutil.virtual_memory().total / GB
INIT_FILE_PREAMBLE = """# InvokeAI initialization file
# This is the InvokeAI initialization file, which contains command-line default values.
@ -89,6 +100,12 @@ INIT_FILE_PREAMBLE = """# InvokeAI initialization file
logger = InvokeAILogger.getLogger()
class DummyWidgetValue(Enum):
zero = 0
true = True
false = False
# --------------------------------------------
def postscript(errors: None):
if not any(errors):
@ -381,13 +398,35 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
)
self.max_cache_size = self.add_widget_intelligent(
IntTitleSlider,
name="Size of the RAM cache used for fast model switching (GB)",
name="RAM cache size (GB). Make this at least large enough to hold a single full model.",
value=old_opts.max_cache_size,
out_of=20,
out_of=MAX_RAM,
lowest=3,
begin_entry_at=6,
scroll_exit=True,
)
if HAS_CUDA:
self.nextrely += 1
self.add_widget_intelligent(
npyscreen.TitleFixedText,
name="VRAM cache size (GB). Reserving a small amount of VRAM will modestly speed up the start of image generation.",
begin_entry_at=0,
editable=False,
color="CONTROL",
scroll_exit=True,
)
self.nextrely -= 1
self.max_vram_cache_size = self.add_widget_intelligent(
npyscreen.Slider,
value=old_opts.max_vram_cache_size,
out_of=round(MAX_VRAM * 2) / 2,
lowest=0.0,
relx=8,
step=0.25,
scroll_exit=True,
)
else:
self.max_vram_cache_size = DummyWidgetValue.zero
self.nextrely += 1
self.outdir = self.add_widget_intelligent(
FileBox,
@ -404,7 +443,7 @@ Use cursor arrows to make a checkbox selection, and space to toggle.
self.autoimport_dirs = {}
self.autoimport_dirs["autoimport_dir"] = self.add_widget_intelligent(
FileBox,
name=f"Folder to recursively scan for new checkpoints, ControlNets, LoRAs and TI models",
name="Folder to recursively scan for new checkpoints, ControlNets, LoRAs and TI models",
value=str(config.root_path / config.autoimport_dir),
select_dir=True,
must_exist=False,
@ -479,6 +518,7 @@ https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/LICENS
"outdir",
"free_gpu_mem",
"max_cache_size",
"max_vram_cache_size",
"xformers_enabled",
"always_use_cpu",
]:
@ -595,13 +635,13 @@ def maybe_create_models_yaml(root: Path):
# -------------------------------------
def run_console_ui(program_opts: Namespace, initfile: Path = None) -> (Namespace, Namespace):
# parse_args() will read from init file if present
invokeai_opts = default_startup_options(initfile)
invokeai_opts.root = program_opts.root
# The third argument is needed in the Windows 11 environment to
# launch a console window running this program.
set_min_terminal_size(MIN_COLS, MIN_LINES)
if not set_min_terminal_size(MIN_COLS, MIN_LINES):
raise WindowTooSmallException(
"Could not increase terminal size. Try running again with a larger window or smaller font size."
)
# the install-models application spawns a subprocess to install
# models, and will crash unless this is set before running.
@ -783,6 +823,7 @@ def main():
models_to_download = default_user_selections(opt)
new_init_file = config.root_path / "invokeai.yaml"
if opt.yes_to_all:
write_default_options(opt, new_init_file)
init_options = Namespace(precision="float32" if opt.full_precision else "float16")
@ -808,6 +849,8 @@ def main():
postscript(errors=errors)
if not opt.yes_to_all:
input("Press any key to continue...")
except WindowTooSmallException as e:
logger.error(str(e))
except KeyboardInterrupt:
print("\nGoodbye! Come back soon.")

13
invokeai/frontend/install/model_install.py
View File

@ -28,7 +28,6 @@ from npyscreen import widget
from invokeai.backend.util.logging import InvokeAILogger
from invokeai.backend.install.model_install_backend import (
ModelInstallList,
InstallSelections,
ModelInstall,
SchedulerPredictionType,
@ -41,12 +40,12 @@ from invokeai.frontend.install.widgets import (
SingleSelectColumns,
TextBox,
BufferBox,
FileBox,
set_min_terminal_size,
select_stable_diffusion_config_file,
CyclingForm,
MIN_COLS,
MIN_LINES,
WindowTooSmallException,
)
from invokeai.app.services.config import InvokeAIAppConfig
@ -156,7 +155,7 @@ class addModelsForm(CyclingForm, npyscreen.FormMultiPage):
BufferBox,
name="Log Messages",
editable=False,
max_height=15,
max_height=6,
)
self.nextrely += 1
@ -693,7 +692,11 @@ def select_and_download_models(opt: Namespace):
# needed to support the probe() method running under a subprocess
torch.multiprocessing.set_start_method("spawn")
set_min_terminal_size(MIN_COLS, MIN_LINES)
if not set_min_terminal_size(MIN_COLS, MIN_LINES):
raise WindowTooSmallException(
"Could not increase terminal size. Try running again with a larger window or smaller font size."
)
installApp = AddModelApplication(opt)
try:
installApp.run()
@ -787,6 +790,8 @@ def main():
curses.echo()
curses.endwin()
logger.info("Goodbye! Come back soon.")
except WindowTooSmallException as e:
logger.error(str(e))
except widget.NotEnoughSpaceForWidget as e:
if str(e).startswith("Height of 1 allocated"):
logger.error("Insufficient vertical space for the interface. Please make your window taller and try again")

61
invokeai/frontend/install/widgets.py
View File

@ -21,31 +21,40 @@ MIN_COLS = 130
MIN_LINES = 38
class WindowTooSmallException(Exception):
pass
# -------------------------------------
def set_terminal_size(columns: int, lines: int):
ts = get_terminal_size()
width = max(columns, ts.columns)
height = max(lines, ts.lines)
def set_terminal_size(columns: int, lines: int) -> bool:
OS = platform.uname().system
if OS == "Windows":
pass
# not working reliably - ask user to adjust the window
# _set_terminal_size_powershell(width,height)
elif OS in ["Darwin", "Linux"]:
_set_terminal_size_unix(width, height)
screen_ok = False
while not screen_ok:
ts = get_terminal_size()
width = max(columns, ts.columns)
height = max(lines, ts.lines)
# check whether it worked....
ts = get_terminal_size()
pause = False
if ts.columns < columns:
print("\033[1mThis window is too narrow for the user interface.\033[0m")
pause = True
if ts.lines < lines:
print("\033[1mThis window is too short for the user interface.\033[0m")
pause = True
if pause:
input("Maximize the window then press any key to continue..")
if OS == "Windows":
pass
# not working reliably - ask user to adjust the window
# _set_terminal_size_powershell(width,height)
elif OS in ["Darwin", "Linux"]:
_set_terminal_size_unix(width, height)
# check whether it worked....
ts = get_terminal_size()
if ts.columns < columns or ts.lines < lines:
print(
f"\033[1mThis window is too small for the interface. InvokeAI requires {columns}x{lines} (w x h) characters, but window is {ts.columns}x{ts.lines}\033[0m"
)
resp = input(
"Maximize the window and/or decrease the font size then press any key to continue. Type [Q] to give up.."
)
if resp.upper().startswith("Q"):
break
else:
screen_ok = True
return screen_ok
def _set_terminal_size_powershell(width: int, height: int):
@ -80,14 +89,14 @@ def _set_terminal_size_unix(width: int, height: int):
sys.stdout.flush()
def set_min_terminal_size(min_cols: int, min_lines: int):
def set_min_terminal_size(min_cols: int, min_lines: int) -> bool:
# make sure there's enough room for the ui
term_cols, term_lines = get_terminal_size()
if term_cols >= min_cols and term_lines >= min_lines:
return
return True
cols = max(term_cols, min_cols)
lines = max(term_lines, min_lines)
set_terminal_size(cols, lines)
return set_terminal_size(cols, lines)
class IntSlider(npyscreen.Slider):
@ -164,7 +173,7 @@ class FloatSlider(npyscreen.Slider):
class FloatTitleSlider(npyscreen.TitleText):
_entry_type = FloatSlider
_entry_type = npyscreen.Slider
class SelectColumnBase: