Qwen image edit (#37)

* New nodes (#35)

* Add 'none' option to LoRA_MODELS and improve reference image handling in Predictor class

* Update task choices in Predictor class to include 'none' option for improved flexibility in task selection.

* Refactor error handling in Predictor class to require only 'image' parameter, simplifying input validation for predictions.

* Remove unused character image files and update dataset loading functions to support spatial configurations.

* import fix

* Update .gitignore to improve file exclusion patterns and remove unused checkpoint files from the training directory.

* Remove unused filtered image files from character spatial reference directory to clean up the dataset.

* Add debug print statements for dataset loading and image optimization in FluxOminiKontextDataset class

* Enhance print statements in training and dataset classes to clarify LoRA weight usage and provide image size during optimization.

* Editor node rotation (#34)

* rotation support

* renaming

* fixes

* predit qwen

* Update diffusers package version to 0.35.1 in requirements and cog.yaml; remove direct installation from git in build process.

Author: Saquib764

cog.yaml

@@ -16,6 +16,7 @@ build:
 
   # a list of packages in the format <package-name>==<version>
   python_packages:
+    - diffusers==0.35.1
     - transformers
     - sentencepiece
     - protobuf
@@ -35,7 +36,6 @@ build:
 
   run:
     - curl -o /usr/local/bin/pget -L "https://github.com/replicate/pget/releases/download/v0.6.0/pget_linux_x86_64" && chmod +x /usr/local/bin/pget
-    - pip install git+https://github.com/huggingface/diffusers
 
 predict: "predict.py:Predictor"
 # train: "trainer.py:train"

requirements.txt

@@ -1,6 +1,6 @@
 torch>=2.0.0
 lightning>=2.0.0
-diffusers>=0.25.0
+diffusers>=0.35.1
 transformers>=4.35.0
 peft>=0.17.0
 prodigyopt>=1.0.0