Merge branch 'main' into install/force-torch-reinstall

.github/workflows/pypi-release.yml

@@ -0,0 +1,41 @@
+name: PyPI Release
+
+on:
+  push:
+    paths:
+      - 'ldm/invoke/_version.py'
+  workflow_dispatch:
+
+jobs:
+  release:
+    if: github.repository == 'invoke-ai/InvokeAI'
+    runs-on: ubuntu-22.04
+    env:
+      TWINE_USERNAME: __token__
+      TWINE_PASSWORD: ${{ secrets.PYPI_API_TOKEN }}
+      TWINE_NON_INTERACTIVE: 1
+    steps:
+      - name: checkout sources
+        uses: actions/checkout@v3
+
+      - name: install deps
+        run: pip install --upgrade build twine
+
+      - name: build package
+        run: python3 -m build
+
+      - name: check distribution
+        run: twine check dist/*
+
+      - name: check PyPI versions
+        if: github.ref == 'refs/heads/main'
+        run: |
+          pip install --upgrade requests
+          python -c "\
+          import scripts.pypi_helper; \
+          EXISTS=scripts.pypi_helper.local_on_pypi(); \
+          print(f'PACKAGE_EXISTS={EXISTS}')" >> $GITHUB_ENV
+
+      - name: upload package
+        if: env.PACKAGE_EXISTS == 'False' && env.TWINE_PASSWORD != ''
+        run: twine upload dist/*

installer/create_installer.sh

@@ -14,12 +14,13 @@ fi
 VERSION=$(cd ..; python -c "from ldm.invoke import __version__ as version; print(version)")
 PATCH=""
 VERSION="v${VERSION}${PATCH}"
+LATEST_TAG="v2.3-latest"
 
 echo Building installer for version $VERSION
 echo "Be certain that you're in the 'installer' directory before continuing."
 read -p "Press any key to continue, or CTRL-C to exit..."
 
-read -e -p "Commit and tag this repo with ${VERSION} and 'v2.3-latest'? [n]: " input
+read -e -p "Commit and tag this repo with '${VERSION}' and '${LATEST_TAG}'? [n]: " input
 RESPONSE=${input:='n'}
 if [ "$RESPONSE" == 'y' ]; then
     git commit -a
@@ -28,8 +29,9 @@ if [ "$RESPONSE" == 'y' ]; then
 	    echo "Existing/invalid tag"
 	    exit -1
     fi
-    git push origin :refs/tags/v2.3-latest
-    git tag -fa latest
+
+    git push origin :refs/tags/$LATEST_TAG
+    git tag -fa $LATEST_TAG
 fi
 
 # ----------------------

installer/installer.py

@@ -346,7 +346,6 @@ class InvokeAiInstance:
         # NOTE: currently the config script does its own arg parsing! this means the command-line switches
         # from the installer will also automatically propagate down to the config script.
         # this may change in the future with config refactoring!
-
         invokeai_configure.main()
 
     def install_user_scripts(self):

invokeai/backend/invoke_ai_web_server.py

@@ -626,9 +626,10 @@ class InvokeAIWebServer:
                         printable_parameters["init_mask"][:64] + "..."
                     )
 
-                print(
-                    f">> Image generation requested: {printable_parameters}\nESRGAN parameters: {esrgan_parameters}\nFacetool parameters: {facetool_parameters}"
-                )
+                print(f'\n>> Image Generation Parameters:\n\n{printable_parameters}\n')
+                print(f'>> ESRGAN Parameters: {esrgan_parameters}')
+                print(f'>> Facetool Parameters: {facetool_parameters}')
+
                 self.generate_images(
                     generation_parameters,
                     esrgan_parameters,
@@ -1154,7 +1155,7 @@ class InvokeAIWebServer:
                     image, os.path.basename(path), self.thumbnail_image_path
                 )
 
-                print(f'>> Image generated: "{path}"')
+                print(f'\n\n>> Image generated: "{path}"\n')
                 self.write_log_message(f'[Generated] "{path}": {command}')
 
                 if progress.total_iterations > progress.current_iteration:
@@ -1193,8 +1194,6 @@ class InvokeAIWebServer:
 
                 progress.set_current_iteration(progress.current_iteration + 1)
 
-            print(generation_parameters)
-
             def diffusers_step_callback_adapter(*cb_args, **kwargs):
                 if isinstance(cb_args[0], PipelineIntermediateState):
                     progress_state: PipelineIntermediateState = cb_args[0]
@@ -1305,8 +1304,6 @@ class InvokeAIWebServer:
 
             rfc_dict["variations"] = variations
 
-            print(parameters)
-
             if rfc_dict["type"] == "img2img":
                 rfc_dict["strength"] = parameters["strength"]
                 rfc_dict["fit"] = parameters["fit"]  # TODO: Noncompliant

ldm/generate.py

@@ -574,7 +574,7 @@ class Generate:
             print('>> Could not generate image.')
 
         toc = time.time()
-        print('>> Usage stats:')
+        print('\n>> Usage stats:')
         print(
             f'>>   {len(results)} image(s) generated in', '%4.2fs' % (
                 toc - tic)

ldm/invoke/CLI.py

@@ -4,6 +4,10 @@ import sys
 import shlex
 import traceback
 
+from argparse import Namespace
+from pathlib import Path
+from typing import Optional, Union
+
 if sys.platform == "darwin":
     os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1"
 
@@ -16,10 +20,10 @@ from ldm.invoke.pngwriter import PngWriter, retrieve_metadata, write_metadata
 from ldm.invoke.image_util import make_grid
 from ldm.invoke.log import write_log
 from ldm.invoke.model_manager import ModelManager
-from pathlib import Path
-from argparse import Namespace
-import pyparsing
+
+import click  # type: ignore
 import ldm.invoke
+import pyparsing  # type: ignore
 
 # global used in multiple functions (fix)
 infile = None
@@ -69,8 +73,10 @@ def main():
 
     # these two lines prevent a horrible warning message from appearing
     # when the frozen CLIP tokenizer is imported
-    import transformers
+    import transformers  # type: ignore
     transformers.logging.set_verbosity_error()
+    import diffusers
+    diffusers.logging.set_verbosity_error()
 
     # Loading Face Restoration and ESRGAN Modules
     gfpgan,codeformer,esrgan = load_face_restoration(opt)
@@ -570,7 +576,7 @@ def set_default_output_dir(opt:Args, completer:Completer):
     completer.set_default_dir(opt.outdir)
 
 
-def import_model(model_path:str, gen, opt, completer):
+def import_model(model_path: str, gen, opt, completer):
     '''
     model_path can be (1) a URL to a .ckpt file; (2) a local .ckpt file path; or
     (3) a huggingface repository id
@@ -579,12 +585,28 @@ def import_model(model_path:str, gen, opt, completer):
 
     if model_path.startswith(('http:','https:','ftp:')):
         model_name = import_ckpt_model(model_path, gen, opt, completer)
+
     elif os.path.exists(model_path) and model_path.endswith(('.ckpt','.safetensors')) and os.path.isfile(model_path):
         model_name = import_ckpt_model(model_path, gen, opt, completer)
-    elif re.match('^[\w.+-]+/[\w.+-]+$',model_path):
-        model_name = import_diffuser_model(model_path, gen, opt, completer)
+
     elif os.path.isdir(model_path):
-        model_name = import_diffuser_model(Path(model_path), gen, opt, completer)
+
+        # Allow for a directory containing multiple models.
+        models = list(Path(model_path).rglob('*.ckpt')) + list(Path(model_path).rglob('*.safetensors'))
+
+        if models:
+            # Only the last model name will be used below.
+            for model in sorted(models):
+
+                if click.confirm(f'Import {model.stem} ?', default=True):
+                    model_name = import_ckpt_model(model, gen, opt, completer)
+                    print()
+        else:
+            model_name = import_diffuser_model(Path(model_path), gen, opt, completer)
+
+    elif re.match(r'^[\w.+-]+/[\w.+-]+$', model_path):
+        model_name = import_diffuser_model(model_path, gen, opt, completer)
+
     else:
         print(f'** {model_path} is neither the path to a .ckpt file nor a diffusers repository id. Can\'t import.')
 
@@ -602,7 +624,7 @@ def import_model(model_path:str, gen, opt, completer):
     completer.update_models(gen.model_manager.list_models())
     print(f'>> {model_name} successfully installed')
 
-def import_diffuser_model(path_or_repo:str, gen, opt, completer)->str:
+def import_diffuser_model(path_or_repo: Union[Path, str], gen, _, completer) -> Optional[str]:
     manager = gen.model_manager
     default_name = Path(path_or_repo).stem
     default_description = f'Imported model {default_name}'
@@ -625,7 +647,7 @@ def import_diffuser_model(path_or_repo:str, gen, opt, completer)->str:
         return None
     return model_name
 
-def import_ckpt_model(path_or_url:str, gen, opt, completer)->str:
+def import_ckpt_model(path_or_url: Union[Path, str], gen, opt, completer) -> Optional[str]:
     manager = gen.model_manager
     default_name = Path(path_or_url).stem
     default_description = f'Imported model {default_name}'
@@ -1133,8 +1155,8 @@ def report_model_error(opt:Namespace, e:Exception):
         for arg in yes_to_all.split():
             sys.argv.append(arg)
 
-    from ldm.invoke.config import configure_invokeai
-    configure_invokeai.main()
+    from ldm.invoke.config import invokeai_configure
+    invokeai_configure.main()
     print('** InvokeAI will now restart')
     sys.argv = previous_args
     main() # would rather do a os.exec(), but doesn't exist?

ldm/invoke/_version.py

@@ -1 +1 @@
-__version__='2.3.0-rc3'
+__version__='2.3.0-rc4'

ldm/invoke/args.py

@@ -196,6 +196,7 @@ class Args(object):
             elif os.path.exists(legacyinit):
                 print(f'>> WARNING: Old initialization file found at {legacyinit}. This location is deprecated. Please move it to {Globals.root}/invokeai.init.')
                 sysargs.insert(0,f'@{legacyinit}')
+            Globals.log_tokenization = self._arg_parser.parse_args(sysargs).log_tokenization
 
             self._arg_switches = self._arg_parser.parse_args(sysargs)
             return self._arg_switches
@@ -599,6 +600,12 @@ class Args(object):
             help=f'Set the default sampler. Supported samplers: {", ".join(SAMPLER_CHOICES)}',
             default='k_lms',
         )
+        render_group.add_argument(
+            '--log_tokenization',
+            '-t',
+            action='store_true',
+            help='shows how the prompt is split into tokens'
+        )
         render_group.add_argument(
             '-f',
             '--strength',
@@ -744,7 +751,7 @@ class Args(object):
                 invoke> !fetch 0000015.8929913.png
                 invoke> a fantastic alien landscape -W 576 -H 512 -s 60 -A plms -C 7.5
                 invoke> !fetch /path/to/images/*.png prompts.txt
- 
+
             !replay /path/to/prompts.txt
             Replays all the prompts contained in the file prompts.txt.
 
@@ -756,6 +763,7 @@ class Args(object):
             !models                                   -- list models in configs/models.yaml
             !switch <model_name>                      -- switch to model named <model_name>
             !import_model /path/to/weights/file.ckpt  -- adds a .ckpt model to your config
+            !import_model /path/to/weights/           -- interactively import models from a directory
             !import_model http://path_to_model.ckpt   -- downloads and adds a .ckpt model to your config
             !import_model hakurei/waifu-diffusion     -- downloads and adds a diffusers model to your config
             !optimize_model <model_name>              -- converts a .ckpt model to a diffusers model

ldm/invoke/conditioning.py

@@ -17,6 +17,7 @@ from ..models.diffusion import cross_attention_control
 from ..models.diffusion.shared_invokeai_diffusion import InvokeAIDiffuserComponent
 from ..modules.encoders.modules import WeightedFrozenCLIPEmbedder
 from ..modules.prompt_to_embeddings_converter import WeightedPromptFragmentsToEmbeddingsConverter
+from ldm.invoke.globals import Globals
 
 
 def get_uc_and_c_and_ec(prompt_string, model, log_tokens=False, skip_normalize_legacy_blend=False):
@@ -92,9 +93,9 @@ def _get_conditioning_for_prompt(parsed_prompt: Union[Blend, FlattenedPrompt], p
     Process prompt structure and tokens, and return (conditioning, unconditioning, extra_conditioning_info)
     """
 
-    if log_tokens:
-        print(f">> Parsed prompt to {parsed_prompt}")
-        print(f">> Parsed negative prompt to {parsed_negative_prompt}")
+    if log_tokens or Globals.log_tokenization:
+        print(f"\n>> [TOKENLOG] Parsed Prompt: {parsed_prompt}")
+        print(f"\n>> [TOKENLOG] Parsed Negative Prompt: {parsed_negative_prompt}")
 
     conditioning = None
     cac_args: cross_attention_control.Arguments = None
@@ -235,7 +236,7 @@ def _get_embeddings_and_tokens_for_prompt(model, flattened_prompt: FlattenedProm
     fragments = [x.text for x in flattened_prompt.children]
     weights = [x.weight for x in flattened_prompt.children]
     embeddings, tokens = model.get_learned_conditioning([fragments], return_tokens=True, fragment_weights=[weights])
-    if log_tokens:
+    if log_tokens or Globals.log_tokenization:
         text = " ".join(fragments)
         log_tokenization(text, model, display_label=log_display_label)
 
@@ -273,12 +274,12 @@ def log_tokenization(text, model, display_label=None):
     # usually tokens have '</w>' to indicate end-of-word,
     # but for readability it has been replaced with ' '
     """
-
     tokens = model.cond_stage_model.tokenizer.tokenize(text)
     tokenized = ""
     discarded = ""
     usedTokens = 0
     totalTokens = len(tokens)
+
     for i in range(0, totalTokens):
         token = tokens[i].replace('</w>', ' ')
         # alternate color
@@ -288,8 +289,11 @@ def log_tokenization(text, model, display_label=None):
             usedTokens += 1
         else:  # over max token length
             discarded = discarded + f"\x1b[0;3{s};40m{token}"
-    print(f"\n>> Tokens {display_label or ''} ({usedTokens}):\n{tokenized}\x1b[0m")
+
+    if usedTokens > 0:
+        print(f'\n>> [TOKENLOG] Tokens {display_label or ""} ({usedTokens}):')
+        print(f'{tokenized}\x1b[0m')
+    
     if discarded != "":
-        print(
-            f">> Tokens Discarded ({totalTokens - usedTokens}):\n{discarded}\x1b[0m"
-        )
+        print(f'\n>> [TOKENLOG] Tokens Discarded ({totalTokens - usedTokens}):')
+        print(f'{discarded}\x1b[0m')

ldm/invoke/config/invokeai_configure.py

@@ -127,8 +127,8 @@ script do it for you. Manual installation is described at:
 
 https://invoke-ai.github.io/InvokeAI/installation/020_INSTALL_MANUAL/
 
-You may download the recommended models (about 10GB total), select a customized set, or
-completely skip this step.
+You may download the recommended models (about 15GB total), install all models (40 GB!!) 
+select a customized set, or completely skip this step.
 """
     )
     completer.set_options(["recommended", "customized", "skip"])
@@ -435,9 +435,7 @@ def _download_diffusion_weights(
             )
         except OSError as e:
             if str(e).startswith("fp16 is not a valid"):
-                print(
-                    f"Could not fetch half-precision version of model {repo_id}; fetching full-precision instead"
-                )
+                pass
             else:
                 print(f"An unexpected error occurred while downloading the model: {e})")
         if path:
@@ -868,7 +866,7 @@ def initialize_rootdir(root: str, yes_to_all: bool = False):
     ):
         os.makedirs(os.path.join(root, name), exist_ok=True)
 
-    configs_src = Path(configs.__path__[-1])
+    configs_src = Path(configs.__path__[0])
     configs_dest = Path(root) / "configs"
     if not os.path.samefile(configs_src, configs_dest):
         shutil.copytree(configs_src, configs_dest, dirs_exist_ok=True)

ldm/invoke/generator/diffusers_pipeline.py

@@ -4,7 +4,6 @@ import dataclasses
 import inspect
 import secrets
 import sys
-import warnings
 from dataclasses import dataclass, field
 from typing import List, Optional, Union, Callable, Type, TypeVar, Generic, Any
 
@@ -641,7 +640,6 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
 
     @property
     def cond_stage_model(self):
-        warnings.warn("legacy compatibility layer", DeprecationWarning)
         return self.prompt_fragments_to_embeddings_converter
 
     @torch.inference_mode()

ldm/invoke/generator/inpaint.py

@@ -194,7 +194,8 @@ class Inpaint(Img2Img):
         """
 
         self.enable_image_debugging = enable_image_debugging
-        self.infill_method = infill_method or infill_methods()[0], # The infill method to use
+        infill_method = infill_method or infill_methods()[0]
+        self.infill_method = infill_method
 
         self.inpaint_width = inpaint_width
         self.inpaint_height = inpaint_height

ldm/invoke/model_manager.py

@@ -18,7 +18,7 @@ import warnings
 import safetensors.torch
 from pathlib import Path
 from shutil import move, rmtree
-from typing import Union, Any
+from typing import Any, Optional, Union
 from huggingface_hub import scan_cache_dir
 from ldm.util import download_with_progress_bar
 
@@ -484,12 +484,11 @@ class ModelManager(object):
                     **pipeline_args,
                     **fp_args,
                 )
-
             except OSError as e:
                 if str(e).startswith('fp16 is not a valid'):
-                    print(f'Could not fetch half-precision version of model {name_or_path}; fetching full-precision instead')
+                    pass
                 else:
-                    print(f'An unexpected error occurred while downloading the model: {e})')
+                    print(f'** An unexpected error occurred while downloading the model: {e})')
             if pipeline:
                 break
 
@@ -881,14 +880,14 @@ class ModelManager(object):
         print('** Migration is done. Continuing...')
 
 
-    def _resolve_path(self, source:Union[str,Path], dest_directory:str)->Path:
+    def _resolve_path(self, source: Union[str, Path], dest_directory: str) -> Optional[Path]:
         resolved_path = None
-        if source.startswith(('http:','https:','ftp:')):
+        if str(source).startswith(('http:','https:','ftp:')):
             basename = os.path.basename(source)
             if not os.path.isabs(dest_directory):
                 dest_directory = os.path.join(Globals.root,dest_directory)
             dest = os.path.join(dest_directory,basename)
-            if download_with_progress_bar(source,dest):
+            if download_with_progress_bar(str(source), Path(dest)):
                 resolved_path = Path(dest)
         else:
             if not os.path.isabs(source):
@@ -1040,7 +1039,7 @@ class ModelManager(object):
                 vae = AutoencoderKL.from_pretrained(name_or_path, **vae_args, **fp_args)
             except OSError as e:
                 if str(e).startswith('fp16 is not a valid'):
-                    print('  | Half-precision version of model not available; fetching full-precision instead')
+                    pass
                 else:
                     deferred_error = e
             if vae:

ldm/invoke/training/textual_inversion.py

@@ -295,7 +295,8 @@ class textualInversionForm(npyscreen.FormMultiPageAction):
             for idx in range(len(model_names))
             if "default" in conf[model_names[idx]]
         ]
-        return (model_names, defaults[0])
+        default = defaults[0] if len(defaults)>0 else 0
+        return (model_names, default)
 
     def marshall_arguments(self) -> dict:
         args = dict()

ldm/util.py

@@ -284,9 +284,9 @@ class ProgressBar():
 
 def download_with_progress_bar(url:str, dest:Path)->bool:
     try:
-        if not os.path.exists(dest):
-            os.makedirs((os.path.dirname(dest) or '.'), exist_ok=True)
-            request.urlretrieve(url,dest,ProgressBar(os.path.basename(dest)))
+        if not dest.exists():
+            dest.parent.mkdir(parents=True, exist_ok=True)
+            request.urlretrieve(url,dest,ProgressBar(dest.stem))
             return True
         else:
             return True

pyproject.toml

@@ -36,6 +36,7 @@ classifiers = [
 dependencies = [
   "accelerate",
   "albumentations",
+  "click",
   "clip_anytorch",          # replacing "clip @ https://github.com/openai/CLIP/archive/eaa22acb90a5876642d0507623e859909230a52d.zip",
   "datasets",
   "diffusers[torch]~=0.11",
@@ -98,13 +99,13 @@ dependencies = [
 
 # legacy entrypoints; provided for backwards compatibility
 "invoke.py" = "ldm.invoke.CLI:main"
-"configure_invokeai.py" = "ldm.invoke.config.configure_invokeai:main"
+"configure_invokeai.py" = "ldm.invoke.config.invokeai_configure:main"
 "textual_inversion.py" = "ldm.invoke.training.textual_inversion:main"
 "merge_embeddings.py" = "ldm.invoke.merge_diffusers:main"
 
 # modern entrypoints
 "invokeai" = "ldm.invoke.CLI:main"
-"invokeai-configure" = "ldm.invoke.config.configure_invokeai:main"
+"invokeai-configure" = "ldm.invoke.config.invokeai_configure:main"
 "invokeai-merge" = "ldm.invoke.merge_diffusers:main"   # note name munging
 "invokeai-ti" = "ldm.invoke.training.textual_inversion:main"
 

scripts/configure_invokeai.py

@@ -2,8 +2,8 @@
 # Copyright (c) 2022 Lincoln D. Stein (https://github.com/lstein)
 
 import warnings
-from ldm.invoke.config import configure_invokeai
+from ldm.invoke.config import invokeai_configure
 
 if __name__ == '__main__':
-    warnings.warn("configire_invokeai.py is deprecated, please run 'invoke'", DeprecationWarning)
+    warnings.warn("configure_invokeai.py is deprecated, please run 'invokai-configure'", DeprecationWarning)
     configure_invokeai.main()

scripts/pypi_helper.py

@@ -0,0 +1,27 @@
+import requests as request
+
+import ldm.invoke._version as version
+
+local_version = str(version.__version__)
+
+
+def get_pypi_versions(package_name="InvokeAI") -> list[str]:
+    """Get the versions of the package from PyPI"""
+    url = f"https://pypi.org/pypi/{package_name}/json"
+    response = request.get(url).json()
+    versions: list[str] = list(response["releases"].keys())
+    return versions
+
+
+def local_on_pypi(package_name="InvokeAI", local_version=local_version) -> bool:
+    """Compare the versions of the package from PyPI and the local package"""
+    pypi_versions = get_pypi_versions(package_name)
+    return local_version in pypi_versions
+
+
+if __name__ == "__main__":
+    package_name = "InvokeAI"
+    if local_on_pypi():
+        print(f"Package {package_name} is up to date")
+    else:
+        print(f"Package {package_name} is not up to date")