change workflow to deploy from v2.3 branch

.github/workflows/mkdocs-material.yml

@@ -41,7 +41,7 @@ jobs:
             --verbose
 
       - name: deploy to gh-pages
-        if: ${{ github.ref == 'refs/heads/main' }}
+        if: ${{ github.ref == 'refs/heads/v2.3' }}
         run: |
           python -m \
             mkdocs gh-deploy \

docs/features/LORAS.md

@@ -41,16 +41,6 @@ Windows systems). If the `loras` folder does not already exist, just
 create it. The vast majority of LoRA models use the Kohya file format,
 which is a type of `.safetensors` file.
 
-!!! warning "LoRA Naming Restrictions"
-
-    InvokeAI will only recognize LoRA files that contain the
-    characters a-z, A-Z, 0-9 and the underscore character
-    _. Other characters, including the hyphen, will cause the
-    LoRA file not to load. These naming restrictions may be
-    relaxed in the future, but for now you will need to rename
-    files that contain hyphens, commas, brackets, and other
-    non-word characters.
-
 You may change where InvokeAI looks for the `loras` folder by passing the
 `--lora_directory` option to the `invoke.sh`/`invoke.bat` launcher, or
 by placing the option in `invokeai.init`. For example:

docs/features/index.md

@@ -33,11 +33,6 @@ title: Overview
         Restore mangled faces and make images larger with upscaling. Also see
         the [Embiggen Upscaling Guide](EMBIGGEN.md).
 
-    -   The [Using LoRA Models](LORAS.md)
-
-        Add custom subjects and styles using HuggingFace's repository of
-        embeddings.
-	
     -   The [Concepts Library](CONCEPTS.md)
 
         Add custom subjects and styles using HuggingFace's repository of

docs/installation/deprecated_documentation/INSTALL_LINUX.md

@@ -79,7 +79,7 @@ title: Manual Installation, Linux
         and obtaining an access token for downloading. It will then download and
         install the weights files for you.
 
-        Please look [here](../020_INSTALL_MANUAL.md) for a manual process for doing
+        Please look [here](../INSTALL_MANUAL.md) for a manual process for doing
         the same thing.
 
 7.  Start generating images!

docs/installation/deprecated_documentation/INSTALL_WINDOWS.md

@@ -75,7 +75,7 @@ Note that you will need NVIDIA drivers, Python 3.10, and Git installed beforehan
           obtaining an access token for downloading. It will then download and install the
           weights files for you.
 
-          Please look [here](../020_INSTALL_MANUAL.md) for a manual process for doing the
+          Please look [here](../INSTALL_MANUAL.md) for a manual process for doing the
           same thing.
 
 8. Start generating images!

docs/requirements-mkdocs.txt

@@ -1,5 +0,0 @@
-mkdocs
-mkdocs-material>=8, <9
-mkdocs-git-revision-date-localized-plugin
-mkdocs-redirects==1.2.0
-

installer/lib/installer.py

@@ -243,15 +243,16 @@ class InvokeAiInstance:
 
         # Note that we're installing pinned versions of torch and
         # torchvision here, which *should* correspond to what is
-        # in pyproject.toml.
+        # in pyproject.toml. This is to prevent torch 2.0 from
+        # being installed and immediately uninstalled and replaced with 1.13
         pip = local[self.pip]
 
         (
             pip[
                 "install",
                 "--require-virtualenv",
-                "torch~=2.0.0",
-                "torchvision>=0.14.1",
+                "torch~=1.13.1",
+                "torchvision~=0.14.1",
                 "--force-reinstall",
                 "--find-links" if find_links is not None else None,
                 find_links,

invokeai/backend/invoke_ai_web_server.py

@@ -25,7 +25,7 @@ from invokeai.backend.modules.parameters import parameters_to_command
 import invokeai.frontend.dist as frontend
 from ldm.generate import Generate
 from ldm.invoke.args import Args, APP_ID, APP_VERSION, calculate_init_img_hash
-from ldm.invoke.concepts_lib import get_hf_concepts_lib
+from ldm.invoke.concepts_lib import HuggingFaceConceptsLibrary
 from ldm.invoke.conditioning import (
     get_tokens_for_prompt_object,
     get_prompt_structure,
@@ -538,7 +538,7 @@ class InvokeAIWebServer:
             try:
                 local_triggers = self.generate.model.textual_inversion_manager.get_all_trigger_strings()
                 locals = [{'name': x} for x in sorted(local_triggers, key=str.casefold)]
-                concepts = get_hf_concepts_lib().list_concepts(minimum_likes=5)
+                concepts = HuggingFaceConceptsLibrary().list_concepts(minimum_likes=5)
                 concepts = [{'name': f'<{x}>'} for x in sorted(concepts, key=str.casefold) if f'<{x}>' not in local_triggers]
                 socketio.emit("foundTextualInversionTriggers", {'local_triggers': locals, 'huggingface_concepts': concepts})
             except Exception as e:

ldm/generate.py

@@ -13,16 +13,11 @@ import time
 import traceback
 from typing import List
 
-import warnings
-with warnings.catch_warnings():
-    warnings.filterwarnings("ignore", category=UserWarning)
-    import torch
-    
 import cv2
 import diffusers
 import numpy as np
 import skimage
-
+import torch
 import transformers
 from diffusers.pipeline_utils import DiffusionPipeline
 from diffusers.utils.import_utils import is_xformers_available
@@ -984,15 +979,13 @@ class Generate:
         seed_everything(random.randrange(0, np.iinfo(np.uint32).max))
         if self.embedding_path and not model_data.get("ti_embeddings_loaded"):
             print(f'>> Loading embeddings from {self.embedding_path}')
-            with warnings.catch_warnings():
-                warnings.filterwarnings("ignore", category=UserWarning)
-                for root, _, files in os.walk(self.embedding_path):
-                    for name in files:
-                        ti_path = os.path.join(root, name)
-                        self.model.textual_inversion_manager.load_textual_inversion(
-                            ti_path, defer_injecting_tokens=True
-                        )
-                model_data["ti_embeddings_loaded"] = True
+            for root, _, files in os.walk(self.embedding_path):
+                for name in files:
+                    ti_path = os.path.join(root, name)
+                    self.model.textual_inversion_manager.load_textual_inversion(
+                        ti_path, defer_injecting_tokens=True
+                    )
+            model_data["ti_embeddings_loaded"] = True
         print(
             f'>> Textual inversion triggers: {", ".join(sorted(self.model.textual_inversion_manager.get_all_trigger_strings()))}'
         )

ldm/invoke/CLI.py

@@ -9,6 +9,7 @@ from pathlib import Path
 from typing import Union
 
 import click
+
 from compel import PromptParser
 
 if sys.platform == "darwin":

ldm/invoke/_version.py

@@ -1,3 +1 @@
-__version__='2.3.5.post2'
-
-
+__version__='2.3.5'

ldm/invoke/ckpt_to_diffuser.py

@@ -620,10 +620,7 @@ def convert_ldm_vae_checkpoint(checkpoint, config):
     for key in keys:
         if key.startswith(vae_key):
             vae_state_dict[key.replace(vae_key, "")] = checkpoint.get(key)
-    new_checkpoint = convert_ldm_vae_state_dict(vae_state_dict,config)
-    return new_checkpoint
 
-def convert_ldm_vae_state_dict(vae_state_dict, config):
     new_checkpoint = {}
 
     new_checkpoint["encoder.conv_in.weight"] = vae_state_dict["encoder.conv_in.weight"]

ldm/invoke/concepts_lib.py

@@ -12,14 +12,6 @@ from urllib import request, error as ul_error
 from huggingface_hub import HfFolder, hf_hub_url, ModelSearchArguments, ModelFilter, HfApi
 from ldm.invoke.globals import Globals
 
-singleton = None
-
-def get_hf_concepts_lib():
-    global singleton
-    if singleton is None:
-        singleton = HuggingFaceConceptsLibrary()
-    return singleton
-
 class HuggingFaceConceptsLibrary(object):
     def __init__(self, root=None):
         '''

ldm/invoke/config/invokeai_update.py

@@ -6,7 +6,6 @@ import os
 import platform
 import psutil
 import requests
-import pkg_resources
 from rich import box, print
 from rich.console import Console, group
 from rich.panel import Panel
@@ -73,20 +72,10 @@ def welcome(versions: dict):
     )
     console.line()
 
-def get_extras():
-    extras = ''
-    try:
-        dist = pkg_resources.get_distribution('xformers')
-        extras = '[xformers]'
-    except pkg_resources.DistributionNotFound:
-        pass
-    return extras
-
 def main():
     versions = get_versions()
     if invokeai_is_running():
         print(f':exclamation: [bold red]Please terminate all running instances of InvokeAI before updating.[/red bold]')
-        input('Press any key to continue...')
         return
 
     welcome(versions)
@@ -105,15 +94,13 @@ def main():
     elif choice=='4':
         branch = Prompt.ask('Enter an InvokeAI branch name')
 
-    extras = get_extras()
-
     print(f':crossed_fingers: Upgrading to [yellow]{tag if tag else release}[/yellow]')
     if release:
-        cmd = f"pip install 'invokeai{extras} @ {INVOKE_AI_SRC}/{release}.zip' --use-pep517 --upgrade"
+        cmd = f'pip install {INVOKE_AI_SRC}/{release}.zip --use-pep517 --upgrade'
     elif tag:
-        cmd = f"pip install 'invokeai{extras} @ {INVOKE_AI_TAG}/{tag}.zip' --use-pep517 --upgrade"
+        cmd = f'pip install {INVOKE_AI_TAG}/{tag}.zip --use-pep517 --upgrade'
     else:
-        cmd = f"pip install 'invokeai{extras} @ {INVOKE_AI_BRANCH}/{branch}.zip' --use-pep517 --upgrade"
+        cmd = f'pip install {INVOKE_AI_BRANCH}/{branch}.zip --use-pep517 --upgrade'
     print('')
     print('')
     if os.system(cmd)==0:

ldm/invoke/model_manager.py

@@ -9,6 +9,7 @@ from __future__ import annotations
 import contextlib
 import gc
 import hashlib
+import io
 import os
 import re
 import sys
@@ -30,10 +31,11 @@ from huggingface_hub import scan_cache_dir
 from omegaconf import OmegaConf
 from omegaconf.dictconfig import DictConfig
 from picklescan.scanner import scan_file_path
+
 from ldm.invoke.devices import CPU_DEVICE
 from ldm.invoke.generator.diffusers_pipeline import StableDiffusionGeneratorPipeline
 from ldm.invoke.globals import Globals, global_cache_dir
-from ldm.util import ask_user, download_with_resume, url_attachment_name
+from ldm.util import ask_user, download_with_resume, instantiate_from_config, url_attachment_name
 
 
 class SDLegacyType(Enum):
@@ -368,9 +370,8 @@ class ModelManager(object):
         print(
             f">> Converting legacy checkpoint {model_name} into a diffusers model..."
         )
-        from .ckpt_to_diffuser import (
-            load_pipeline_from_original_stable_diffusion_ckpt,
-        )
+        from ldm.invoke.ckpt_to_diffuser import load_pipeline_from_original_stable_diffusion_ckpt
+
         if self._has_cuda():
             torch.cuda.empty_cache()
         pipeline = load_pipeline_from_original_stable_diffusion_ckpt(
@@ -432,7 +433,7 @@ class ModelManager(object):
                     **fp_args,
                 )
             except OSError as e:
-                if 'Revision Not Found' in str(e):
+                if str(e).startswith("fp16 is not a valid"):
                     pass
                 else:
                     print(
@@ -1229,17 +1230,6 @@ class ModelManager(object):
         return vae_path
                                
     def _load_vae(self, vae_config) -> AutoencoderKL:
-        using_fp16 = self.precision == "float16"
-        dtype = torch.float16 if using_fp16 else torch.float32
-        
-        # Handle the common case of a user shoving a VAE .ckpt into
-        # the vae field for a diffusers. We convert it into diffusers
-        # format and use it.
-        if isinstance(vae_config,(str,Path)):
-            return self.convert_vae(vae_config).to(dtype=dtype)
-        elif isinstance(vae_config,DictConfig) and (vae_path := vae_config.get('path')):
-            return self.convert_vae(vae_path).to(dtype=dtype)
-            
         vae_args = {}
         try:
             name_or_path = self.model_name_or_path(vae_config)
@@ -1247,6 +1237,7 @@ class ModelManager(object):
             return None
         if name_or_path is None:
             return None
+        using_fp16 = self.precision == "float16"
 
         vae_args.update(
             cache_dir=global_cache_dir("hub"),
@@ -1286,32 +1277,6 @@ class ModelManager(object):
 
         return vae
 
-    @staticmethod
-    def convert_vae(vae_path: Union[Path,str])->AutoencoderKL:
-        print("   | A checkpoint VAE was detected. Converting to diffusers format.")
-        vae_path = Path(Globals.root,vae_path).resolve()
-        
-        from .ckpt_to_diffuser import (
-            create_vae_diffusers_config,
-            convert_ldm_vae_state_dict,
-        )
-
-        vae_path = Path(vae_path)
-        if vae_path.suffix in ['.pt','.ckpt']:
-            vae_state_dict = torch.load(vae_path, map_location="cpu")
-        else:
-            vae_state_dict = safetensors.torch.load_file(vae_path)
-        if 'state_dict' in vae_state_dict:
-            vae_state_dict = vae_state_dict['state_dict']
-        # TODO: see if this works with 1.x inpaint models and 2.x models
-        config_file_path = Path(Globals.root,"configs/stable-diffusion/v1-inference.yaml")
-        original_conf = OmegaConf.load(config_file_path)
-        vae_config = create_vae_diffusers_config(original_conf, image_size=512) # TODO: fix
-        diffusers_vae = convert_ldm_vae_state_dict(vae_state_dict,vae_config)
-        vae = AutoencoderKL(**vae_config)
-        vae.load_state_dict(diffusers_vae)
-        return vae
-
     @staticmethod
     def _delete_model_from_cache(repo_id):
         cache_info = scan_cache_dir(global_cache_dir("diffusers"))

ldm/invoke/readline.py

@@ -13,7 +13,7 @@ import re
 import atexit
 from typing import List
 from ldm.invoke.args import Args
-from ldm.invoke.concepts_lib import get_hf_concepts_lib
+from ldm.invoke.concepts_lib import HuggingFaceConceptsLibrary
 from ldm.invoke.globals import Globals
 from ldm.modules.lora_manager import LoraManager
 
@@ -287,7 +287,7 @@ class Completer(object):
     def _concept_completions(self, text, state):
         if self.concepts is None:
             # cache Concepts() instance so we can check for updates in concepts_list during runtime.
-            self.concepts = get_hf_concepts_lib()
+            self.concepts = HuggingFaceConceptsLibrary()
             self.embedding_terms.update(set(self.concepts.list_concepts()))
         else:
             self.embedding_terms.update(set(self.concepts.list_concepts()))

ldm/modules/embedding_manager.py

@@ -6,7 +6,7 @@ from torch import nn
 
 import sys
 
-from ldm.invoke.concepts_lib import get_hf_concepts_lib
+from ldm.invoke.concepts_lib import HuggingFaceConceptsLibrary
 from ldm.data.personalized import per_img_token_list
 from transformers import CLIPTokenizer
 from functools import partial
@@ -39,7 +39,7 @@ class EmbeddingManager(nn.Module):
         super().__init__()
 
         self.embedder = embedder
-        self.concepts_library=get_hf_concepts_lib()
+        self.concepts_library=HuggingFaceConceptsLibrary()
 
         self.string_to_token_dict = {}
         self.string_to_param_dict = nn.ParameterDict()

ldm/modules/textual_inversion_manager.py

@@ -3,16 +3,14 @@ from dataclasses import dataclass
 from pathlib import Path
 from typing import Optional, Union
 
-import warnings
-with warnings.catch_warnings():
-    warnings.filterwarnings("ignore", category=UserWarning)
-    import safetensors.torch
-    import torch
+import safetensors.torch
+import torch
 from picklescan.scanner import scan_file_path
 from transformers import CLIPTextModel, CLIPTokenizer
 
 from compel.embeddings_provider import BaseTextualInversionManager
-from ldm.invoke.concepts_lib import get_hf_concepts_lib
+from ldm.invoke.concepts_lib import HuggingFaceConceptsLibrary
+
 
 @dataclass
 class TextualInversion:
@@ -36,7 +34,7 @@ class TextualInversionManager(BaseTextualInversionManager):
         self.tokenizer = tokenizer
         self.text_encoder = text_encoder
         self.full_precision = full_precision
-        self.hf_concepts_library = get_hf_concepts_lib()
+        self.hf_concepts_library = HuggingFaceConceptsLibrary()
         self.trigger_to_sourcefile = dict()
         default_textual_inversions: list[TextualInversion] = []
         self.textual_inversions = default_textual_inversions

pyproject.toml

@@ -32,9 +32,9 @@ dependencies = [
   "albumentations",
   "click",
   "clip_anytorch",
-  "compel~=1.1.5",
+  "compel~=1.1.0",
   "datasets",
-  "diffusers[torch]~=0.16.1",
+  "diffusers[torch]~=0.15.0",
   "dnspython==2.2.1",
   "einops",
   "eventlet",
@@ -76,7 +76,7 @@ dependencies = [
   "taming-transformers-rom1504",
   "test-tube>=0.7.5",
   "torch-fidelity",
-  "torch~=2.0.0",
+  "torch~=1.13.1",
   "torchmetrics",
   "torchvision>=0.14.1",
   "transformers~=4.26",
@@ -108,7 +108,7 @@ requires-python = ">=3.9, <3.11"
 "test" = ["pytest-cov", "pytest>6.0.0"]
 "xformers" = [
   "triton; sys_platform=='linux'",
-  "xformers~=0.0.19; sys_platform!='darwin'",
+  "xformers~=0.0.16; sys_platform!='darwin'",
 ]
 
 [project.scripts]