feat(mm): wip port main models to new api

Author: psychedelicious

invokeai/app/api/routers/model_manager.py

@@ -29,10 +29,7 @@
 )
 from invokeai.app.util.suppress_output import SuppressOutput
 from invokeai.backend.model_manager import BaseModelType, ModelFormat, ModelType
-from invokeai.backend.model_manager.config import (
-    AnyModelConfig,
-    MainCheckpointConfig,
-)
+from invokeai.backend.model_manager.config import AnyModelConfig, SD_1_2_XL_XLRefiner_CheckpointConfig
 from invokeai.backend.model_manager.load.model_cache.cache_stats import CacheStats
 from invokeai.backend.model_manager.metadata.fetch.huggingface import HuggingFaceMetadataFetch
 from invokeai.backend.model_manager.metadata.metadata_base import ModelMetadataWithFiles, UnknownMetadataException
@@ -741,9 +738,10 @@ async def convert_model(
         logger.error(str(e))
         raise HTTPException(status_code=424, detail=str(e))
 
-    if not isinstance(model_config, MainCheckpointConfig):
-        logger.error(f"The model with key {key} is not a main checkpoint model.")
-        raise HTTPException(400, f"The model with key {key} is not a main checkpoint model.")
+    if isinstance(model_config, SD_1_2_XL_XLRefiner_CheckpointConfig):
+        msg = f"The model with key {key} is not a main SD 1/2/XL checkpoint model."
+        logger.error(msg)
+        raise HTTPException(400, msg)
 
     with TemporaryDirectory(dir=ApiDependencies.invoker.services.configuration.models_path) as tmpdir:
         convert_path = pathlib.Path(tmpdir) / pathlib.Path(model_config.path).stem

invokeai/app/services/model_install/model_install_default.py

@@ -41,7 +41,6 @@
     InvalidModelConfigException,
     ModelConfigFactory,
 )
-from invokeai.backend.model_manager.legacy_probe import ModelProbe
 from invokeai.backend.model_manager.metadata import (
     AnyModelRepoMetadata,
     HuggingFaceMetadataFetch,
@@ -601,22 +600,11 @@ def _probe(self, model_path: Path, config: Optional[ModelRecordChanges] = None):
         hash_algo = self._app_config.hashing_algorithm
         fields = config.model_dump()
 
-        # WARNING!
-        # The legacy probe relies on the implicit order of tests to determine model classification.
-        # This can lead to regressions between the legacy and new probes.
-        # Do NOT change the order of `probe` and `classify` without implementing one of the following fixes:
-        # Short-term fix: `classify` tests `matches` in the same order as the legacy probe.
-        # Long-term fix: Improve `matches` to be more specific so that only one config matches
-        #   any given model - eliminating ambiguity and removing reliance on order.
-        # After implementing either of these fixes, remove @pytest.mark.xfail from `test_regression_against_model_probe`
-        try:
-            return ModelProbe.probe(model_path=model_path, fields=deepcopy(fields), hash_algo=hash_algo)  # type: ignore
-        except InvalidModelConfigException:
-            return ModelConfigFactory.from_model_on_disk(
-                mod=model_path,
-                overrides=deepcopy(fields),
-                hash_algo=hash_algo,
-            )
+        return ModelConfigFactory.from_model_on_disk(
+            mod=model_path,
+            overrides=deepcopy(fields),
+            hash_algo=hash_algo,
+        )
 
     def _register(
         self, model_path: Path, config: Optional[ModelRecordChanges] = None, info: Optional[AnyModelConfig] = None