lint and update diffusers

Signed-off-by: vladmandic <mandic00@live.com>

CHANGELOG.md

@@ -1,6 +1,6 @@
 # Change Log for SD.Next
 
-## Update for 2025-01-10
+## Update for 2025-01-12
 
 - **Models**
   - [Qwen-Image-2512](https://huggingface.co/Qwen/Qwen-Image-2512)  
@@ -14,9 +14,13 @@
 - **Feaures**
   - **SDNQ** now has *19 int* based and *69 float* based quantization types  
     *note*: not all are exposed via ui purely for simplicity, but all are available via api and scripts  
-  - allow weights for wildcards, thanks @Tillerz  
+  - **wildcards**: allow weights, thanks @Tillerz  
-  - sampler: add laplace beta schedule  
+  - **sampler**: add laplace beta schedule  
     results in better prompt adherence and smoother infills  
+  - **prompt enhance**: improve handling and refresh ui, thanks @CalamitousFelicitousness  
+    new models such moondream-3 and xiaomo-mimo
+    add support for *thinking* mode where model can reason about the prompt  
+    add support for *vision* processing where prompt enhance can also optionally analyze input image  
 - **UI**
   - kanvas add send-to functionality  
   - kanvas improve support for standardui  
@@ -52,6 +56,7 @@
   - lora loading when using torch without distributed support  
   - google-genai auth, thanks @CalamitousFelicitousness  
   - reduce triton test verbosity  
+  - improve qwen i2i handling  
 
 ## Update for 2025-12-26
 

extensions-builtin/sdnext-modernui

@@ -1 +1 @@
-Subproject commit dff67ef8f40a300627b2f20d16484c65d1d43202
+Subproject commit d83de1303abc930d6a2e8570e6c929a0139510f1

html/locale_en.json

@@ -1224,7 +1224,7 @@
   {"id":"","label":"tdd","localized":"","reload":"","hint":"tdd"},
   {"id":"","label":"te","localized":"","reload":"","hint":"te"},
   {"id":"","label":"temperature","localized":"","reload":"","hint":"Controls randomness in token selection by reshaping the probability distribution.<br>Like adjusting a dial between cautious predictability (low values ~0.4) and creative exploration (higher values ~1). Higher temperatures increase willingness to choose less obvious options, but makes outputs more unpredictable.<br><br>Set to 0 to disable, resulting in silent switch to greedy decoding, disabling sampling."},
-  {"id":"","label":"Thinking mode","localized":"","reload":"","hint":"Enables thinking/reasoning, allowing the model to take more time to generate responses.<br>This can lead to more thoughtful and detailed answers, but will increase response time.<br>This setting affects both hybrid and thinking-only models, and in some may result in lower overall quality than expected. For thinking-only models like Qwen3-VL this setting might have to be combined with prefill to guarantee preventing thinking.<br><br>Models supporting this feature are marked with an \uf0eb icon."},
+  {"id":"","label":"Thinking Mode","localized":"","reload":"","hint":"Enables thinking/reasoning, allowing the model to take more time to generate responses.<br>This can lead to more thoughtful and detailed answers, but will increase response time.<br>This setting affects both hybrid and thinking-only models, and in some may result in lower overall quality than expected. For thinking-only models like Qwen3-VL this setting might have to be combined with prefill to guarantee preventing thinking.<br><br>Models supporting this feature are marked with an \uf0eb icon."},
   {"id":"","label":"Repetition penalty","localized":"","reload":"","hint":"Discourages reusing tokens that already appear in the prompt or output by penalizing their probabilities.<br>Like adding friction to revisiting previous choices. Helps break repetitive loops but may reduce coherence at aggressive values.<br><br>Set to 1 to disable."},
   {"id":"","label":"text guidance scale","localized":"","reload":"","hint":"text guidance scale"},
   {"id":"","label":"template","localized":"","reload":"","hint":"template"},

installer.py

@@ -648,7 +648,7 @@ def check_diffusers():
     t_start = time.time()
     if args.skip_all:
         return
-    sha = 'ed6e5ecf67ac762e878d3d6e25bda4e77b0c820a' # diffusers commit hash
+    sha = 'dad5cb55e6ade24fb397525eb023ad4eba37019d' # diffusers commit hash
     # if args.use_rocm or args.use_zluda or args.use_directml:
     #     sha = '043ab2520f6a19fce78e6e060a68dbc947edb9f9' # lock diffusers versions for now
     pkg = pkg_resources.working_set.by_key.get('diffusers', None)

modules/ltx/ltx_process.py

@@ -140,7 +140,7 @@ def run_ltx(task_id,
             "output_type": "latent",
         }
         if 'Condition' in shared.sd_model.__class__.__name__:
-            base_args["image_cond_noise_scale"] = image_cond_noise_scale    
+            base_args["image_cond_noise_scale"] = image_cond_noise_scale
         shared.log.debug(f'Video: cls={shared.sd_model.__class__.__name__} op=base {base_args}')
         if len(conditions) > 0:
             base_args["conditions"] = conditions