Refactor config class and reorganize image generation options (#4309)

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

- [X Refactor
- [X] Feature

## Have you discussed this change with the InvokeAI team?
- [X] Yes
      
## Have you updated all relevant documentation?
- [X] Yes

## Description

### Refactoring

This PR refactors `invokeai.app.services.config` to be easier to
maintain by splitting off the argument, environment and init file
parsing code from the InvokeAIAppConfig object. This will hopefully make
it easier for people to find the place where the various settings are
defined.

### New Features

In collaboration with @StAlKeR7779 , I have renamed and reorganized the
settings controlling image generation and model management to be more
intuitive. The relevant portion of the init file now looks like this:

```
  Model Cache:
    ram: 14.5
    vram: 0.5
    lazy_offload: true
  Device:
    precision: auto
    device: auto
  Generation:
    sequential_guidance: false
    attention_type: auto
    attention_slice_size: auto
    force_tiled_decode: false
```
Key differences are:
1. Split `Performance/Memory` into `Device`, `Generation` and `Model
Cache`
2. Added the ability to force the `device`. The value of this option is
one of {`auto`, `cpu`, `cuda`, `cuda:1`, `mps`}
3. Added the ability to force the `attention_type`. Possible values are
{`auto`, `normal`, `xformers`, `sliced`, `torch-sdp`}
4. Added the ability to force the `attention_slice_size` when `sliced`
attention is in use. The value of this option is one of {`auto`, `max`}
or an integer between 1 and 8.
 
@StAlKeR7779 Please confirm that I wired the `attention_type` and
`attention_slice_size` configuration options to the diffusers backend
correctly.

In addition, I have exposed the generation-related configuration options
to the TUI:


![image](https://github.com/invoke-ai/InvokeAI/assets/111189/8c0235d4-c3b0-494e-a1ab-ff45cdbfd9af)

### Backward Compatibility

This refactor should be backward compatible with earlier versions of
`invokeai.yaml`. If the user re-runs the `invokeai-configure` script,
`invokeai.yaml` will be upgraded to the current format. Several
configuration attributes had to be changed in order to preserve backward
compatibility. These attributes been changed in the code where
appropriate. For the record:

| Old Name | Preferred New Name | Comment |
| ------------| ---------------|------------|
| `max_cache_size` | `ram_cache_size` |
| `max_vram_cache` | `vram_cache_size` |
| `always_use_cpu` | `use_cpu` | Better to check conf.device == "cpu" |

Author: lstein

docs/features/CONFIGURATION.md

@@ -175,22 +175,27 @@ These configuration settings allow you to enable and disable various InvokeAI fe
 | `internet_available` | `true`     | When a resource is not available locally, try to fetch it via the internet |
 | `log_tokenization` | `false`      | Before each text2image generation, print a color-coded representation of the prompt to the console; this can help understand why a prompt is not working as expected |
 | `patchmatch` | `true`     | Activate the "patchmatch" algorithm for improved inpainting |
-| `restore`    | `true`     | Activate the facial restoration features (DEPRECATED; restoration features will be removed in 3.0.0) |
 
-### Memory/Performance
+### Generation
 
 These options tune InvokeAI's memory and performance characteristics.
 
-| Setting  | Default Value  |  Description |
-|----------|----------------|--------------|
-| `always_use_cpu`     | `false`      | Use the CPU to generate images, even if a GPU is available |
-| `free_gpu_mem`       | `false`      | Aggressively free up GPU memory after each operation; this will allow you to run in low-VRAM environments with some performance penalties |
-| `max_cache_size`       | `6`      | Amount of CPU RAM (in GB) to reserve for caching models in memory; more cache allows you to keep models in memory and switch among them quickly |
-| `max_vram_cache_size`  | `2.75`   | Amount of GPU VRAM (in GB) to reserve for caching models in VRAM; more cache speeds up generation but reduces the size of the images that can be generated. This can be set to zero to maximize the amount of memory available for generation. |
-| `precision`       | `auto`      | Floating point precision. One of `auto`, `float16` or `float32`. `float16` will consume half the memory of `float32` but produce slightly lower-quality images. The `auto` setting will guess the proper precision based on your video card and operating system |
-| `sequential_guidance`     | `false`      | Calculate guidance in serial rather than in parallel, lowering memory requirements at the cost of some performance loss |
-| `xformers_enabled`        | `true`      | If the x-formers memory-efficient attention module is installed, activate it for better memory usage and generation speed|
-| `tiled_decode`            | `false`     | If true, then during the VAE decoding phase the image will be decoded a section at a time, reducing memory consumption at the cost of a performance hit |
+| Setting               | Default Value | Description                                                                                                                                                                                                                                                      |
+|-----------------------|---------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| `sequential_guidance` | `false`       | Calculate guidance in serial rather than in parallel, lowering memory requirements at the cost of some performance loss                                                                                                                                          |
+| `attention_type`      | `auto`        | Select the type of attention to use. One of `auto`,`normal`,`xformers`,`sliced`, or `torch-sdp`                                                                                                                                                                  |
+| `attention_slice_size` | `auto`       | When "sliced" attention is selected, set the slice size. One of `auto`, `balanced`, `max` or the integers 1-8|
+| `force_tiled_decode`  | `false`       | Force the VAE step to decode in tiles, reducing memory consumption at the cost of performance |
+
+### Device
+
+These options configure the generation execution device.
+
+| Setting               | Default Value | Description                                                                                                                                                                                                                                                      |
+|-----------------------|---------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| `device`           | `auto`        | Preferred execution device. One of `auto`, `cpu`, `cuda`, `cuda:1`, `mps`. `auto` will choose the device depending on the hardware platform and the installed torch capabilities. |
+| `precision`           | `auto`        | Floating point precision. One of `auto`, `float16` or `float32`. `float16` will consume half the memory of `float32` but produce slightly lower-quality images. The `auto` setting will guess the proper precision based on your video card and operating system |
+
 
 ### Paths
 

invokeai/app/services/config/__init__.py

@@ -0,0 +1,8 @@
+"""
+Init file for InvokeAI configure package
+"""
+
+from .invokeai_config import (  # noqa F401
+    InvokeAIAppConfig,
+    get_invokeai_config,
+)

invokeai/app/services/config/base.py

@@ -0,0 +1,239 @@
+# Copyright (c) 2023 Lincoln Stein (https://github.com/lstein) and the InvokeAI Development Team
+
+"""
+Base class for the InvokeAI configuration system.
+It defines a type of pydantic BaseSettings object that
+is able to read and write from an omegaconf-based config file,
+with overriding of settings from environment variables and/or
+the command line.
+"""
+
+from __future__ import annotations
+import argparse
+import os
+import pydoc
+import sys
+from argparse import ArgumentParser
+from omegaconf import OmegaConf, DictConfig, ListConfig
+from pathlib import Path
+from pydantic import BaseSettings
+from typing import ClassVar, Dict, List, Literal, Union, get_origin, get_type_hints, get_args
+
+
+class PagingArgumentParser(argparse.ArgumentParser):
+    """
+    A custom ArgumentParser that uses pydoc to page its output.
+    It also supports reading defaults from an init file.
+    """
+
+    def print_help(self, file=None):
+        text = self.format_help()
+        pydoc.pager(text)
+
+
+class InvokeAISettings(BaseSettings):
+    """
+    Runtime configuration settings in which default values are
+    read from an omegaconf .yaml file.
+    """
+
+    initconf: ClassVar[DictConfig] = None
+    argparse_groups: ClassVar[Dict] = {}
+
+    def parse_args(self, argv: list = sys.argv[1:]):
+        parser = self.get_parser()
+        opt = parser.parse_args(argv)
+        for name in self.__fields__:
+            if name not in self._excluded():
+                value = getattr(opt, name)
+                if isinstance(value, ListConfig):
+                    value = list(value)
+                elif isinstance(value, DictConfig):
+                    value = dict(value)
+                setattr(self, name, value)
+
+    def to_yaml(self) -> str:
+        """
+        Return a YAML string representing our settings. This can be used
+        as the contents of `invokeai.yaml` to restore settings later.
+        """
+        cls = self.__class__
+        type = get_args(get_type_hints(cls)["type"])[0]
+        field_dict = dict({type: dict()})
+        for name, field in self.__fields__.items():
+            if name in cls._excluded_from_yaml():
+                continue
+            category = field.field_info.extra.get("category") or "Uncategorized"
+            value = getattr(self, name)
+            if category not in field_dict[type]:
+                field_dict[type][category] = dict()
+            # keep paths as strings to make it easier to read
+            field_dict[type][category][name] = str(value) if isinstance(value, Path) else value
+        conf = OmegaConf.create(field_dict)
+        return OmegaConf.to_yaml(conf)
+
+    @classmethod
+    def add_parser_arguments(cls, parser):
+        if "type" in get_type_hints(cls):
+            settings_stanza = get_args(get_type_hints(cls)["type"])[0]
+        else:
+            settings_stanza = "Uncategorized"
+
+        env_prefix = cls.Config.env_prefix if hasattr(cls.Config, "env_prefix") else settings_stanza.upper()
+
+        initconf = (
+            cls.initconf.get(settings_stanza)
+            if cls.initconf and settings_stanza in cls.initconf
+            else OmegaConf.create()
+        )
+
+        # create an upcase version of the environment in
+        # order to achieve case-insensitive environment
+        # variables (the way Windows does)
+        upcase_environ = dict()
+        for key, value in os.environ.items():
+            upcase_environ[key.upper()] = value
+
+        fields = cls.__fields__
+        cls.argparse_groups = {}
+
+        for name, field in fields.items():
+            if name not in cls._excluded():
+                current_default = field.default
+
+                category = field.field_info.extra.get("category", "Uncategorized")
+                env_name = env_prefix + "_" + name
+                if category in initconf and name in initconf.get(category):
+                    field.default = initconf.get(category).get(name)
+                if env_name.upper() in upcase_environ:
+                    field.default = upcase_environ[env_name.upper()]
+                cls.add_field_argument(parser, name, field)
+
+                field.default = current_default
+
+    @classmethod
+    def cmd_name(self, command_field: str = "type") -> str:
+        hints = get_type_hints(self)
+        if command_field in hints:
+            return get_args(hints[command_field])[0]
+        else:
+            return "Uncategorized"
+
+    @classmethod
+    def get_parser(cls) -> ArgumentParser:
+        parser = PagingArgumentParser(
+            prog=cls.cmd_name(),
+            description=cls.__doc__,
+        )
+        cls.add_parser_arguments(parser)
+        return parser
+
+    @classmethod
+    def add_subparser(cls, parser: argparse.ArgumentParser):
+        parser.add_parser(cls.cmd_name(), help=cls.__doc__)
+
+    @classmethod
+    def _excluded(self) -> List[str]:
+        # internal fields that shouldn't be exposed as command line options
+        return ["type", "initconf"]
+
+    @classmethod
+    def _excluded_from_yaml(self) -> List[str]:
+        # combination of deprecated parameters and internal ones that shouldn't be exposed as invokeai.yaml options
+        return [
+            "type",
+            "initconf",
+            "version",
+            "from_file",
+            "model",
+            "root",
+            "max_cache_size",
+            "max_vram_cache_size",
+            "always_use_cpu",
+            "free_gpu_mem",
+            "xformers_enabled",
+            "tiled_decode",
+        ]
+
+    class Config:
+        env_file_encoding = "utf-8"
+        arbitrary_types_allowed = True
+        case_sensitive = True
+
+    @classmethod
+    def add_field_argument(cls, command_parser, name: str, field, default_override=None):
+        field_type = get_type_hints(cls).get(name)
+        default = (
+            default_override
+            if default_override is not None
+            else field.default
+            if field.default_factory is None
+            else field.default_factory()
+        )
+        if category := field.field_info.extra.get("category"):
+            if category not in cls.argparse_groups:
+                cls.argparse_groups[category] = command_parser.add_argument_group(category)
+            argparse_group = cls.argparse_groups[category]
+        else:
+            argparse_group = command_parser
+
+        if get_origin(field_type) == Literal:
+            allowed_values = get_args(field.type_)
+            allowed_types = set()
+            for val in allowed_values:
+                allowed_types.add(type(val))
+            allowed_types_list = list(allowed_types)
+            field_type = allowed_types_list[0] if len(allowed_types) == 1 else int_or_float_or_str
+
+            argparse_group.add_argument(
+                f"--{name}",
+                dest=name,
+                type=field_type,
+                default=default,
+                choices=allowed_values,
+                help=field.field_info.description,
+            )
+
+        elif get_origin(field_type) == Union:
+            argparse_group.add_argument(
+                f"--{name}",
+                dest=name,
+                type=int_or_float_or_str,
+                default=default,
+                help=field.field_info.description,
+            )
+
+        elif get_origin(field_type) == list:
+            argparse_group.add_argument(
+                f"--{name}",
+                dest=name,
+                nargs="*",
+                type=field.type_,
+                default=default,
+                action=argparse.BooleanOptionalAction if field.type_ == bool else "store",
+                help=field.field_info.description,
+            )
+        else:
+            argparse_group.add_argument(
+                f"--{name}",
+                dest=name,
+                type=field.type_,
+                default=default,
+                action=argparse.BooleanOptionalAction if field.type_ == bool else "store",
+                help=field.field_info.description,
+            )
+
+
+def int_or_float_or_str(value: str) -> Union[int, float, str]:
+    """
+    Workaround for argparse type checking.
+    """
+    try:
+        return int(value)
+    except Exception as e:  # noqa F841
+        pass
+    try:
+        return float(value)
+    except Exception as e:  # noqa F841
+        pass
+    return str(value)