From 0e5d3510996b028ec2ce9e00bb8e6c8df834c053 Mon Sep 17 00:00:00 2001
From: CalamitousFelicitousness <iowasovereign@gmail.com>
Date: Thu, 9 Apr 2026 03:21:50 +0100
Subject: [PATCH] add google gemma-4-e2b and gemma-4-e4b support

- add to prompt enhance models, img2img, and models_cls
- add to vlm/caption models
- update _load_gemma class selection for gemma 4
- bump transformers pin to 5.5.0.dev for gemma4 architecture support
---
 installer.py                  | 2 +-
 modules/caption/models_def.py | 2 ++
 modules/caption/vqa.py        | 6 ++++--
 scripts/prompt_enhance.py     | 5 +++++
 4 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/installer.py b/installer.py
index 70fc19cb9..bb9aa0349 100644
--- a/installer.py
+++ b/installer.py
@@ -513,7 +513,7 @@ def check_transformers():
     pkg_tokenizers = package_spec('tokenizers')
     # target_commit = '753d61104116eefc8ffc977327b441ee0c8d599f' # transformers commit hash == 4.57.6
     # target_commit = "aad13b87ed59f2afcfaebc985f403301887a35fc" # transformers commit hash == 5.3.0
-    target_commit = "2dba8e0495974930af02274d75bd182d22cc1686" # transformers commit hash == 5.3.0.dev-0331
+    target_commit = "380e3cc5d59912a48508cb6d4959a31cd460e12e" # transformers commit hash == 5.5.0.dev-0409
     if args.use_directml:
         target_transformers = '4.52.4'
         target_tokenizers = '0.21.4'
diff --git a/modules/caption/models_def.py b/modules/caption/models_def.py
index 7e53e2e76..f91b2e5f2 100644
--- a/modules/caption/models_def.py
+++ b/modules/caption/models_def.py
@@ -5,6 +5,8 @@ vlm_models = {
     "Google Gemma 3 4B": "google/gemma-3-4b-it",
     "Google Gemma 3n E2B": "google/gemma-3n-E2B-it", # 1.5GB
     "Google Gemma 3n E4B": "google/gemma-3n-E4B-it", # 1.5GB
+    "Google Gemma 4 E2B": "google/gemma-4-E2B-it",
+    "Google Gemma 4 E4B": "google/gemma-4-E4B-it",
     "Nidum Gemma 3 4B Uncensored": "nidum/Nidum-Gemma-3-4B-it-Uncensored",
     "Allura Gemma 3 Glitter 4B": "allura-org/Gemma-3-Glitter-4B",
     # Qwen3.5
diff --git a/modules/caption/vqa.py b/modules/caption/vqa.py
index 807479144..74cc3cdca 100644
--- a/modules/caption/vqa.py
+++ b/modules/caption/vqa.py
@@ -619,11 +619,13 @@ class VQA:
         return response
 
     def _load_gemma(self, repo: str):
-        """Load Gemma 3 model and processor."""
+        """Load Gemma model and processor."""
         if self.model is None or self.loaded != repo:
             log.debug(f'Caption load: vlm="{repo}"')
             self._unload_current()
-            if '3n' in repo:
+            if 'gemma-4' in repo.lower():
+                cls = transformers.Gemma4ForConditionalGeneration
+            elif '3n' in repo:
                 cls = transformers.Gemma3nForConditionalGeneration  # pylint: disable=no-member
             else:
                 cls = transformers.Gemma3ForConditionalGeneration
diff --git a/scripts/prompt_enhance.py b/scripts/prompt_enhance.py
index b0d46bac8..e919081f9 100644
--- a/scripts/prompt_enhance.py
+++ b/scripts/prompt_enhance.py
@@ -109,6 +109,8 @@ class Options:
         'google/gemma-3-4b-it',
         'google/gemma-3n-E2B-it',
         'google/gemma-3n-E4B-it',
+        'google/gemma-4-E2B-it',
+        'google/gemma-4-E4B-it',
         # Gemma Finetunes
         'nidum/Nidum-Gemma-3-4B-it-Uncensored',
         'allura-org/Gemma-3-Glitter-4B',
@@ -173,6 +175,8 @@ class Options:
         'google/gemma-3-4b-it': {},
         'google/gemma-3n-E2B-it': {},
         'google/gemma-3n-E4B-it': {},
+        'google/gemma-4-E2B-it': {},
+        'google/gemma-4-E4B-it': {},
         # Gemma Finetunes
         'nidum/Nidum-Gemma-3-4B-it-Uncensored': {},
         'allura-org/Gemma-3-Glitter-4B': {},
@@ -267,6 +271,7 @@ class Options:
         'qwen2_5_vl': 'Qwen2_5_VLForConditionalGeneration',
         'qwen2_vl': 'Qwen2VLForConditionalGeneration',
         'mistral3': 'Mistral3ForConditionalGeneration',
+        'gemma4': 'Gemma4ForConditionalGeneration',
     }
 
     # default = list(models)[1] # gemma-3-4b-it