remove oldest items from model cache first

api/onnx_web/chain/correct_gfpgan.py

@@ -1,6 +1,5 @@
 from logging import getLogger
 from os import path
-from typing import Optional
 
 import numpy as np
 from gfpgan import GFPGANer

api/onnx_web/chain/upscale_resrgan.py

@@ -48,7 +48,10 @@ def load_resrgan(
     elif params.format == "onnx":
         # use ONNX acceleration, if available
         model = OnnxNet(
-            server, model_file, provider=device.provider, provider_options=device.options
+            server,
+            model_file,
+            provider=device.provider,
+            provider_options=device.options,
         )
     elif params.format == "pth":
         model = RRDBNet(
@@ -71,7 +74,9 @@ def load_resrgan(
     logger.debug("loading Real ESRGAN upscale model from %s", model_path)
 
     # TODO: shouldn't need the PTH file
-    model_path_pth = path.join(server.model_path, ".cache", ("%s.pth" % params.upscale_model))
+    model_path_pth = path.join(
+        server.model_path, ".cache", ("%s.pth" % params.upscale_model)
+    )
     upsampler = RealESRGANer(
         scale=params.scale,
         model_path=model_path_pth,

api/onnx_web/chain/upscale_stable_diffusion.py

@@ -15,7 +15,6 @@ from ..utils import ServerContext, run_gc
 logger = getLogger(__name__)
 
 
-
 def load_stable_diffusion(
     server: ServerContext, upscale: UpscaleParams, device: DeviceParams
 ):

api/onnx_web/server/hacks.py

@@ -121,4 +121,10 @@ def apply_patches(ctx: ServerContext):
     apply_patch_basicsr(ctx)
     apply_patch_codeformer(ctx)
     apply_patch_facexlib(ctx)
-    unload(["basicsr.utils.download_util", "codeformer.facelib.utils.misc", "facexlib.utils"])
+    unload(
+        [
+            "basicsr.utils.download_util",
+            "codeformer.facelib.utils.misc",
+            "facexlib.utils",
+        ]
+    )

api/onnx_web/server/model_cache.py

@@ -43,6 +43,6 @@ class ModelCache:
             logger.info(
                 "Removing models from cache, %s of %s", (total - self.limit), total
             )
-            self.cache[:] = self.cache[: self.limit]
+            self.cache[:] = self.cache[-self.limit :]
         else:
             logger.debug("Model cache below limit, %s of %s", total, self.limit)