anodyine/vllm-omni
1
0 Fork 0
Code Issues 4 Pull Requests 2 Actions Packages Projects Releases Wiki Activity

[Bugfix] Set VaeImageProcessor do_convert_rgb True (#1032)

Browse Source 
Signed-off-by: gcanlin <canlinguosdu@gmail.com>
This commit is contained in:
Canlin Guo authored and GitHub committed 2026-01-29 12:02:39 +00:00
1 parent dd518973de
commit 0c8b5ac2d5
2 files changed
+6 -6
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

vllm_omni/diffusion/models/qwen_image/pipeline_qwen_image_edit.py
+3 -3
View File
@@ -61,7 +61,7 @@ def get_qwen_image_edit_pre_process_func(
vae_config = json.load(f)
vae_scale_factor = 2 ** len(vae_config["temporal_downsample"]) if "temporal_downsample" in vae_config else 8
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2)
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2, do_convert_rgb=True)
latent_channels = vae_config.get("z_dim", 16)
def pre_process_func(
@@ -136,7 +136,7 @@ def get_qwen_image_edit_post_process_func(
vae_config = json.load(f)
vae_scale_factor = 2 ** len(vae_config["temporal_downsample"]) if "temporal_downsample" in vae_config else 8
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2)
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2, do_convert_rgb=True)
def post_process_func(
images: torch.Tensor,
@@ -261,7 +261,7 @@ class QwenImageEditPipeline(
self.vae_scale_factor = 2 ** len(self.vae.temperal_downsample) if getattr(self, "vae", None) else 8
self.latent_channels = self.vae.config.z_dim if getattr(self, "vae", None) else 16
self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor * 2)
self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor * 2, do_convert_rgb=True)
self.tokenizer_max_length = 1024
# Edit prompt template - different from generation template
self.prompt_template_encode = "<|im_start|>system\nDescribe the key features of the input image (color, shape, size, texture, objects, background), then explain how the user's text instruction should alter or modify the image. Generate a new image that meets the user's requirements while maintaining consistency with the original input where appropriate.<|im_end|>\n<|im_start|>user\n<|vision_start|><|image_pad|><|vision_end|>{}<|im_end|>\n<|im_start|>assistant\n" # noqa: E501
vllm_omni/diffusion/models/qwen_image/pipeline_qwen_image_edit_plus.py
+3 -3
View File
@@ -67,7 +67,7 @@ def get_qwen_image_edit_plus_pre_process_func(
vae_config = json.load(f)
vae_scale_factor = 2 ** len(vae_config["temporal_downsample"]) if "temporal_downsample" in vae_config else 8
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2)
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2, do_convert_rgb=True)
latent_channels = vae_config.get("z_dim", 16)
def pre_process_func(
@@ -158,7 +158,7 @@ def get_qwen_image_edit_plus_post_process_func(
vae_config = json.load(f)
vae_scale_factor = 2 ** len(vae_config["temporal_downsample"]) if "temporal_downsample" in vae_config else 8
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2)
image_processor = VaeImageProcessor(vae_scale_factor=vae_scale_factor * 2, do_convert_rgb=True)
def post_process_func(
images: torch.Tensor,
@@ -214,7 +214,7 @@ class QwenImageEditPlusPipeline(nn.Module, SupportImageInput):
self.vae_scale_factor = 2 ** len(self.vae.temperal_downsample) if getattr(self, "vae", None) else 8
self.latent_channels = self.vae.config.z_dim if getattr(self, "vae", None) else 16
self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor * 2)
self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor * 2, do_convert_rgb=True)
self.tokenizer_max_length = 1024
# Edit prompt template - different from generation template, supports multiple images
self.prompt_template_encode = (