From 1e098d61327e1c02c1a47b2626514474aa8e3c7e Mon Sep 17 00:00:00 2001 From: comfyanonymous <121283862+comfyanonymous@users.noreply.github.com> Date: Fri, 26 Sep 2025 15:34:17 -0700 Subject: [PATCH 1/3] Don't add template to qwen2.5vl when template is in prompt. (#10043) Make the hunyuan image refiner template_end 36. --- comfy/text_encoders/hunyuan_image.py | 8 ++++- comfy/text_encoders/qwen_image.py | 48 +++++++++++++++++----------- 2 files changed, 36 insertions(+), 20 deletions(-) diff --git a/comfy/text_encoders/hunyuan_image.py b/comfy/text_encoders/hunyuan_image.py index 699eddc3336f..ff04726e11ba 100644 --- a/comfy/text_encoders/hunyuan_image.py +++ b/comfy/text_encoders/hunyuan_image.py @@ -63,7 +63,13 @@ def __init__(self, byt5=True, device="cpu", dtype=None, model_options={}): self.byt5_small = None def encode_token_weights(self, token_weight_pairs): - cond, p, extra = super().encode_token_weights(token_weight_pairs) + tok_pairs = token_weight_pairs["qwen25_7b"][0] + template_end = -1 + if tok_pairs[0][0] == 27: + if len(tok_pairs) > 36: # refiner prompt uses a fixed 36 template_end + template_end = 36 + + cond, p, extra = super().encode_token_weights(token_weight_pairs, template_end=template_end) if self.byt5_small is not None and "byt5" in token_weight_pairs: out = self.byt5_small.encode_token_weights(token_weight_pairs["byt5"]) extra["conditioning_byt5small"] = out[0] diff --git a/comfy/text_encoders/qwen_image.py b/comfy/text_encoders/qwen_image.py index 6646b1003b43..40fa6793739c 100644 --- a/comfy/text_encoders/qwen_image.py +++ b/comfy/text_encoders/qwen_image.py @@ -18,13 +18,22 @@ def __init__(self, embedding_directory=None, tokenizer_data={}): self.llama_template_images = "<|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" def tokenize_with_weights(self, text, return_word_ids=False, llama_template=None, images=[], **kwargs): - if llama_template is None: - if len(images) > 0: - llama_text = self.llama_template_images.format(text) - else: - llama_text = self.llama_template.format(text) + skip_template = False + if text.startswith('<|im_start|>'): + skip_template = True + if text.startswith('<|start_header_id|>'): + skip_template = True + + if skip_template: + llama_text = text else: - llama_text = llama_template.format(text) + if llama_template is None: + if len(images) > 0: + llama_text = self.llama_template_images.format(text) + else: + llama_text = self.llama_template.format(text) + else: + llama_text = llama_template.format(text) tokens = super().tokenize_with_weights(llama_text, return_word_ids=return_word_ids, disable_weights=True, **kwargs) key_name = next(iter(tokens)) embed_count = 0 @@ -47,22 +56,23 @@ class QwenImageTEModel(sd1_clip.SD1ClipModel): def __init__(self, device="cpu", dtype=None, model_options={}): super().__init__(device=device, dtype=dtype, name="qwen25_7b", clip_model=Qwen25_7BVLIModel, model_options=model_options) - def encode_token_weights(self, token_weight_pairs): + def encode_token_weights(self, token_weight_pairs, template_end=-1): out, pooled, extra = super().encode_token_weights(token_weight_pairs) tok_pairs = token_weight_pairs["qwen25_7b"][0] count_im_start = 0 - for i, v in enumerate(tok_pairs): - elem = v[0] - if not torch.is_tensor(elem): - if isinstance(elem, numbers.Integral): - if elem == 151644 and count_im_start < 2: - template_end = i - count_im_start += 1 - - if out.shape[1] > (template_end + 3): - if tok_pairs[template_end + 1][0] == 872: - if tok_pairs[template_end + 2][0] == 198: - template_end += 3 + if template_end == -1: + for i, v in enumerate(tok_pairs): + elem = v[0] + if not torch.is_tensor(elem): + if isinstance(elem, numbers.Integral): + if elem == 151644 and count_im_start < 2: + template_end = i + count_im_start += 1 + + if out.shape[1] > (template_end + 3): + if tok_pairs[template_end + 1][0] == 872: + if tok_pairs[template_end + 2][0] == 198: + template_end += 3 out = out[:, template_end:] From 196954ab8c55bc4ac48113686a57ce250677c7b5 Mon Sep 17 00:00:00 2001 From: Jedrzej Kosinski Date: Fri, 26 Sep 2025 19:55:03 -0700 Subject: [PATCH 2/3] Add 'input_cond' and 'input_uncond' to the args dictionary passed into sampler_cfg_function (#10044) --- comfy/samplers.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/comfy/samplers.py b/comfy/samplers.py index b3202cec6f2c..c59e296a1ad3 100755 --- a/comfy/samplers.py +++ b/comfy/samplers.py @@ -360,7 +360,7 @@ def calc_cond_uncond_batch(model, cond, uncond, x_in, timestep, model_options): def cfg_function(model, cond_pred, uncond_pred, cond_scale, x, timestep, model_options={}, cond=None, uncond=None): if "sampler_cfg_function" in model_options: args = {"cond": x - cond_pred, "uncond": x - uncond_pred, "cond_scale": cond_scale, "timestep": timestep, "input": x, "sigma": timestep, - "cond_denoised": cond_pred, "uncond_denoised": uncond_pred, "model": model, "model_options": model_options} + "cond_denoised": cond_pred, "uncond_denoised": uncond_pred, "model": model, "model_options": model_options, "input_cond": cond, "input_uncond": uncond} cfg_result = x - model_options["sampler_cfg_function"](args) else: cfg_result = uncond_pred + (cond_pred - uncond_pred) * cond_scale @@ -390,7 +390,7 @@ def sampling_function(model, x, timestep, uncond, cond, cond_scale, model_option for fn in model_options.get("sampler_pre_cfg_function", []): args = {"conds":conds, "conds_out": out, "cond_scale": cond_scale, "timestep": timestep, "input": x, "sigma": timestep, "model": model, "model_options": model_options} - out = fn(args) + out = fn(args) return cfg_function(model, out[0], out[1], cond_scale, x, timestep, model_options=model_options, cond=cond, uncond=uncond_) From 0572029fee48741a8cf34a8e4d485898c5ab5dfd Mon Sep 17 00:00:00 2001 From: ComfyUI Wiki Date: Sat, 27 Sep 2025 12:18:16 +0800 Subject: [PATCH 3/3] Update template to 0.1.88 (#10046) --- requirements.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/requirements.txt b/requirements.txt index 2980bebdd909..b3f81e8fa0a1 100644 --- a/requirements.txt +++ b/requirements.txt @@ -1,5 +1,5 @@ comfyui-frontend-package==1.26.13 -comfyui-workflow-templates==0.1.86 +comfyui-workflow-templates==0.1.88 comfyui-embedded-docs==0.2.6 torch torchsde