Merge branch 'dev'

Author: william-murray1204

README.md

@@ -375,7 +375,7 @@ stable_diffusion = StableDiffusion(
 )
 output = stable_diffusion.generate_image(
       prompt="make the cat blue",
-      images=["input.png"],
+      ref_images=["input.png"],
       cfg_scale=1.0, # a cfg_scale of 1 is recommended for FLUX
 )
 ```
@@ -480,7 +480,6 @@ output = stable_diffusion.generate_image(
 You can use [PhotoMaker](https://github.com/TencentARC/PhotoMaker) to personalize generated images with your own ID.
 
 **NOTE**, currently PhotoMaker **ONLY** works with **SDXL** (any SDXL model files will work).
-The VAE in SDXL encounters NaN issues. You can find a fixed VAE here: [SDXL VAE FP16 Fix](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl_vae.safetensors).
 
 Download PhotoMaker model file (in safetensor format) [here](https://huggingface.co/bssrdf/PhotoMaker). The official release of the model file (in .bin format) does not work with `stablediffusion.cpp`.
 
@@ -535,6 +534,75 @@ An `id_embeds.bin` file will be generated in `input_images_dir`.
 
 ---
 
+### <u>QWEN Image</u>
+
+Download the weights from the links below:
+
+- Download `Qwen Image`
+  - safetensors: https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/tree/main/split_files/diffusion_models
+  - gguf: https://huggingface.co/QuantStack/Qwen-Image-GGUF/tree/main
+- Download `vae`
+  - safetensors: https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/tree/main/split_files/vae
+- Download `qwen_2.5_vl 7b`
+  - safetensors: https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/tree/main/split_files/text_encoders
+  - gguf: https://huggingface.co/mradermacher/Qwen2.5-VL-7B-Instruct-GGUF/tree/main
+
+```python
+from stable_diffusion_cpp import StableDiffusion
+
+stable_diffusion = StableDiffusion(
+      diffusion_model_path="../models/qwen-image-Q8_0.gguf",
+      qwen2vl_path="../models/Qwen2.5-VL-7B-Instruct.Q8_0.gguf",
+      vae_path="../models/qwen_image_vae.safetensors",
+      offload_params_to_cpu=True,
+      flow_shift=3,
+)
+
+output = stable_diffusion.generate_image(
+      prompt='一个穿着"QWEN"标志的T恤的中国美女正拿着黑色的马克笔面相镜头微笑。她身后的玻璃板上手写体写着 “一、Qwen-Image的技术路线： 探索视觉生成基础模型的极限，开创理解与生成一体化的未来。二、Qwen-Image的模型特色：1、复杂文字渲染。支持中英渲染、自动布局； 2、精准图像编辑。支持文字编辑、物体增减、风格变换。三、Qwen-Image的未来愿景：赋能专业内容创作、助力生成式AI发展。”',
+      cfg_scale=2.5,
+      sample_method='euler',
+)
+```
+
+#### <u>QWEN Image Edit</u>
+
+Download the weights from the links below:
+
+- Download `Qwen Image Edit`
+  - Qwen Image Edit
+    - safetensors: https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/tree/main/split_files/diffusion_models
+    - gguf: https://huggingface.co/QuantStack/Qwen-Image-Edit-GGUF/tree/main
+  - Qwen Image Edit 2509
+    - safetensors: https://huggingface.co/Comfy-Org/Qwen-Image-Edit_ComfyUI/tree/main/split_files/diffusion_models
+    - gguf: https://huggingface.co/QuantStack/Qwen-Image-Edit-2509-GGUF/tree/main
+- Download `vae`
+  - safetensors: https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/tree/main/split_files/vae
+- Download `qwen_2.5_vl 7b`
+  - safetensors: https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI/tree/main/split_files/text_encoders
+  - gguf: https://huggingface.co/mradermacher/Qwen2.5-VL-7B-Instruct-GGUF/tree/main
+
+```python
+from stable_diffusion_cpp import StableDiffusion
+
+stable_diffusion = StableDiffusion(
+      diffusion_model_path="../models/Qwen_Image_Edit-Q8_0.gguf",
+      qwen2vl_path="../models/Qwen2.5-VL-7B-Instruct.Q8_0.gguf",
+      vae_path="../models/qwen_image_vae.safetensors",
+      offload_params_to_cpu=True,
+      flow_shift=3,
+)
+
+output = stable_diffusion.generate_image(
+      prompt="make the cat blue",
+      ref_images=["input.png"],
+      cfg_scale=2.5,
+      sample_method='euler',
+)
+```
+
+---
+
 ### <u>Wan Video Generation</u>
 
 See [stable-diffusion.cpp Wan download weights](https://github.com/leejet/stable-diffusion.cpp/blob/master/docs/wan.md#download-weights) for a complete list of Wan models.

stable_diffusion_cpp/__init__.py

@@ -4,4 +4,4 @@
 
 # isort: on
 
-__version__ = "0.3.5"
+__version__ = "0.3.6"

stable_diffusion_cpp/_internals.py

@@ -23,6 +23,8 @@ def __init__(
         clip_g_path: str,
         clip_vision_path: str,
         t5xxl_path: str,
+        qwen2vl_path: str,
+        qwen2vl_vision_path: str,
         diffusion_model_path: str,
         high_noise_diffusion_model_path: str,
         vae_path: str,
@@ -35,13 +37,15 @@ def __init__(
         n_threads: int,
         wtype: int,
         rng_type: int,
+        prediction: int,
         offload_params_to_cpu: bool,
         keep_clip_on_cpu: bool,
         keep_control_net_on_cpu: bool,
         keep_vae_on_cpu: bool,
         diffusion_flash_attn: bool,
         diffusion_conv_direct: bool,
         vae_conv_direct: bool,
+        force_sdxl_vae_conv_scale: bool,
         chroma_use_dit_mask: bool,
         chroma_use_t5_mask: bool,
         chroma_t5_mask_pad: int,
@@ -56,6 +60,8 @@ def __init__(
             clip_g_path=clip_g_path.encode("utf-8"),
             clip_vision_path=clip_vision_path.encode("utf-8"),
             t5xxl_path=t5xxl_path.encode("utf-8"),
+            qwen2vl_path=qwen2vl_path.encode("utf-8"),
+            qwen2vl_vision_path=qwen2vl_vision_path.encode("utf-8"),
             diffusion_model_path=diffusion_model_path.encode("utf-8"),
             high_noise_diffusion_model_path=high_noise_diffusion_model_path.encode("utf-8"),
             vae_path=vae_path.encode("utf-8"),
@@ -69,13 +75,15 @@ def __init__(
             n_threads=n_threads,
             wtype=wtype,
             rng_type=rng_type,
+            prediction=prediction,
             offload_params_to_cpu=offload_params_to_cpu,
             keep_clip_on_cpu=keep_clip_on_cpu,
             keep_control_net_on_cpu=keep_control_net_on_cpu,
             keep_vae_on_cpu=keep_vae_on_cpu,
             diffusion_flash_attn=diffusion_flash_attn,
             diffusion_conv_direct=diffusion_conv_direct,
             vae_conv_direct=vae_conv_direct,
+            force_sdxl_vae_conv_scale=force_sdxl_vae_conv_scale,
             chroma_use_dit_mask=chroma_use_dit_mask,
             chroma_use_t5_mask=chroma_use_t5_mask,
             chroma_t5_mask_pad=chroma_t5_mask_pad,

stable_diffusion_cpp/_logger.py

@@ -27,7 +27,7 @@ def sd_log_callback(
     data: ctypes.c_void_p,
 ):
     if logger.level <= SD_LOG_LEVEL_TO_LOGGING_LEVEL[level]:
-        print(text.decode("utf-8"), end="", flush=True, file=sys.stderr)
+        print(text.decode("utf-8", errors="replace"), end="", flush=True, file=sys.stderr)
 
 
 stable_diffusion_cpp.sd_set_log_callback(sd_log_callback, ctypes.c_void_p(0))