P2 of qwen edit model. (#9412)

* P2 of qwen edit model.

* Typo.

* Fix normal qwen.

* Fix.

* Make the TextEncodeQwenImageEdit also set the ref latent.

If you don't want it to set the ref latent and want to use the
ReferenceLatent node with your custom latent instead just disconnect the
VAE.

comfy_extras/nodes_qwen.py

@@ -0,0 +1,63 @@
+import node_helpers
+import comfy.utils
+
+PREFERRED_QWENIMAGE_RESOLUTIONS = [
+    (672, 1568),
+    (688, 1504),
+    (720, 1456),
+    (752, 1392),
+    (800, 1328),
+    (832, 1248),
+    (880, 1184),
+    (944, 1104),
+    (1024, 1024),
+    (1104, 944),
+    (1184, 880),
+    (1248, 832),
+    (1328, 800),
+    (1392, 752),
+    (1456, 720),
+    (1504, 688),
+    (1568, 672),
+]
+
+
+class TextEncodeQwenImageEdit:
+    @classmethod
+    def INPUT_TYPES(s):
+        return {"required": {
+            "clip": ("CLIP", ),
+            "prompt": ("STRING", {"multiline": True, "dynamicPrompts": True}),
+            },
+            "optional": {"vae": ("VAE", ),
+                         "image": ("IMAGE", ),}}
+
+    RETURN_TYPES = ("CONDITIONING",)
+    FUNCTION = "encode"
+
+    CATEGORY = "advanced/conditioning"
+
+    def encode(self, clip, prompt, vae=None, image=None):
+        ref_latent = None
+        if image is None:
+            images = []
+        else:
+            images = [image]
+            if vae is not None:
+                width = image.shape[2]
+                height = image.shape[1]
+                aspect_ratio = width / height
+                _, width, height = min((abs(aspect_ratio - w / h), w, h) for w, h in PREFERRED_QWENIMAGE_RESOLUTIONS)
+                image = comfy.utils.common_upscale(image.movedim(-1, 1), width, height, "lanczos", "center").movedim(1, -1)
+                ref_latent = vae.encode(image[:, :, :, :3])
+
+        tokens = clip.tokenize(prompt, images=images)
+        conditioning = clip.encode_from_tokens_scheduled(tokens)
+        if ref_latent is not None:
+            conditioning = node_helpers.conditioning_set_values(conditioning, {"reference_latents": [ref_latent]}, append=True)
+        return (conditioning, )
+
+
+NODE_CLASS_MAPPINGS = {
+    "TextEncodeQwenImageEdit": TextEncodeQwenImageEdit,
+}