diff --git a/scripts/dataset_tag_editor/interrogators/blip2_captioning.py b/scripts/dataset_tag_editor/interrogators/blip2_captioning.py
index 4d37ab3..b500efe 100644
--- a/scripts/dataset_tag_editor/interrogators/blip2_captioning.py
+++ b/scripts/dataset_tag_editor/interrogators/blip2_captioning.py
@@ -13,10 +13,10 @@ class BLIP2Captioning:
     def load(self):
         if self.model is None or self.processor is None:
             self.processor = Blip2Processor.from_pretrained(
-                self.MODEL_REPO, cache_dir=paths.model_path / "aesthetic"
+                self.MODEL_REPO, cache_dir=paths.model_path
             )
             self.model = Blip2ForConditionalGeneration.from_pretrained(
-                self.MODEL_REPO, cache_dir=paths.model_path / "aesthetic"
+                self.MODEL_REPO, cache_dir=paths.model_path
             ).to(devices.device)
 
     def unload(self):
diff --git a/scripts/dataset_tag_editor/interrogators/git_large_captioning.py b/scripts/dataset_tag_editor/interrogators/git_large_captioning.py
index 4c8680a..bf9f223 100644
--- a/scripts/dataset_tag_editor/interrogators/git_large_captioning.py
+++ b/scripts/dataset_tag_editor/interrogators/git_large_captioning.py
@@ -1,6 +1,8 @@
 from transformers import AutoProcessor, AutoModelForCausalLM
 from modules import shared, devices, lowvram
 
+from scripts.paths import paths
+
 
 # brought from https://huggingface.co/docs/transformers/main/en/model_doc/git and modified
 class GITLargeCaptioning:
@@ -12,10 +14,12 @@ class GITLargeCaptioning:
 
     def load(self):
         if self.model is None or self.processor is None:
-            self.processor = AutoProcessor.from_pretrained(self.MODEL_REPO)
-            self.model = AutoModelForCausalLM.from_pretrained(self.MODEL_REPO).to(
-                shared.device
+            self.processor = AutoProcessor.from_pretrained(
+                self.MODEL_REPO, cache_dir=paths.model_path
             )
+            self.model = AutoModelForCausalLM.from_pretrained(
+                self.MODEL_REPO, cache_dir=paths.model_path
+            ).to(shared.device)
         lowvram.send_everything_to_cpu()
 
     def unload(self):
diff --git a/scripts/dataset_tag_editor/interrogators/waifu_diffusion_tagger.py b/scripts/dataset_tag_editor/interrogators/waifu_diffusion_tagger.py
index 75156d5..b66d41b 100644
--- a/scripts/dataset_tag_editor/interrogators/waifu_diffusion_tagger.py
+++ b/scripts/dataset_tag_editor/interrogators/waifu_diffusion_tagger.py
@@ -4,6 +4,8 @@ from typing import List, Tuple
 from modules import shared, devices
 import launch
 
+from scripts.paths import paths
+
 
 class WaifuDiffusionTagger:
     # brought from https://huggingface.co/spaces/SmilingWolf/wd-v1-4-tags/blob/main/app.py and modified
@@ -24,7 +26,7 @@ class WaifuDiffusionTagger:
 
         if not self.model:
             path_model = huggingface_hub.hf_hub_download(
-                self.MODEL_REPO, self.MODEL_FILENAME
+                self.MODEL_REPO, self.MODEL_FILENAME, cache_dir=paths.model_path
             )
             if (
                 "all" in shared.cmd_opts.use_cpu