Why do errors occur when running templates?

[bbhxwl]

Why do errors occur when running templates?

[starve3meals]

解包子工作流 检查输入输出节点是否有未连接的，这应该是模板工作流有遗漏，不是comfyui的问题

[bbhxwl]

在苹果设备上 运行fp16的模型 需要设备内存大于32GB m5是统一内存 没有显存 GPU使用统一内存

我128g 内存的，那看来我想跑起来，得先学怎么自己弄工作流了，默认模版都fp8，好难。

[starve3meals]

并没有多难 下载模型 放到对应的文件夹就可以了

[starve3meals]

你可以尝试下面的链接下载
unet(放到models/diffusion_models文件夹)

• https://huggingface.co/wangkanai/wan22-fp16-i2v/tree/main/diffusion_models/wan
  lighting lora放到loras文件夹
• High-noise FP16: https://huggingface.co/Kijai/WanVideo_comfy/blob/main/LoRAs/Wan22-Lightning/old/Wan2.2-Lightning_I2V-A14B-4steps-lora_HIGH_fp16.safetensors
• Low-noise FP16: https://huggingface.co/Kijai/WanVideo_comfy/blob/main/LoRAs/Wan22-Lightning/old/Wan2.2-Lightning_I2V-A14B-4steps-lora_LOW_fp16.safetensors
  vae放到models/vae文件夹
• https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors
  clip(放到models/diffusion_models文件夹)
• https://huggingface.co/wangkanai/wan22-fp16-encoders/tree/main/text_encoders

[bbhxwl]

你可以尝试下面的链接下载 unet：

• https://huggingface.co/wangkanai/wan22-fp16-i2v/tree/main/diffusion_models/wan
  lighting lora：
• High-noise FP16: https://huggingface.co/Kijai/WanVideo_comfy/blob/main/LoRAs/Wan22-Lightning/old/Wan2.2-Lightning_I2V-A14B-4steps-lora_HIGH_fp16.safetensors
• Low-noise FP16: https://huggingface.co/Kijai/WanVideo_comfy/blob/main/LoRAs/Wan22-Lightning/old/Wan2.2-Lightning_I2V-A14B-4steps-lora_LOW_fp16.safetensors
  vae:
• https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/vae/wan_2.1_vae.safetensors
  clip:
• https://huggingface.co/wangkanai/wan22-fp16-encoders/tree/main/text_encoders

有交流群么，comfyui的，
我学学怎么找对应的模型，您刚发的摩达， 模型文件里面，Wan2.2-I2V-A14B/high_noise_model
目录也没看到fp16相关字样，一会我再学学huggingface 怎么找。

[starve3meals]

目前没交流群 我也是独狼

[gershu-ar]

After re-reading the issue, and seeing you have an M5: Apple M5 chip is fully compatible with Wan 2.2, but the "no link found" error usually indicates a version mismatch between your ComfyUI core and the native template.

The error no link found in parent for id (xxx) slot [x] (cfg) means the workflow is looking for a CFG (Classifier Free Guidance) input that was renamed or moved in a recent update. This happens when the UI is newer/older than the internal template logic.

Try

1. Hard Reset the Node:
   -> Enable IDs in Settings > Badge: ID.
   -> Find the node ID from the error message (likely the Sampler).
   -> Delete and re-add that specific node manually from the search menu. This forces ComfyUI to re-register the links using the M5-compatible logic.

2. Clear UI Cache:
   -> Refresh your browser (Cmd+R) after updating. This ensures the frontend JSON matches the backend's new node definitions.

3. Update ComfyUI Core:
   -> Run git pull in your terminal. For the M5, you must be on the latest March 2026 builds to support the new GPU subgraphs used by Wan 2.2.

Note: If the error persists, manually typing the CFG value into the node (instead of using an external "Float" node) often bypasses this specific link bug on Apple Silicon.

[bbhxwl]

After re-reading the issue, and seeing you have an M5: Apple M5 chip is fully compatible with Wan 2.2, but the "no link found" error usually indicates a version mismatch between your ComfyUI core and the native template.

The error no link found in parent for id (xxx) slot [x] (cfg) means the workflow is looking for a CFG (Classifier Free Guidance) input that was renamed or moved in a recent update. This happens when the UI is newer/older than the internal template logic.

Try

1. Hard Reset the Node:
   -> Enable IDs in Settings > Badge: ID.
   -> Find the node ID from the error message (likely the Sampler).
   -> Delete and re-add that specific node manually from the search menu. This forces ComfyUI to re-register the links using the M5-compatible logic.
2. Clear UI Cache:
   -> Refresh your browser (Cmd+R) after updating. This ensures the frontend JSON matches the backend's new node definitions.
3. Update ComfyUI Core:
   -> Run git pull in your terminal. For the M5, you must be on the latest March 2026 builds to support the new GPU subgraphs used by Wan 2.2.

Note: If the error persists, manually typing the CFG value into the node (instead of using an external "Float" node) often bypasses this specific link bug on Apple Silicon.

https://www.comfy.org/
Can I download the latest here? I didn't use Git to pull the latest version. I am using it for the first time, so I still don't understand the terms, so I want to find a template that can run it. I have tested that both wan2.2 and ltx2.3 cannot be used properly, and ltx2.3 prompts that fp8 is not supported

[gershu-ar]

Correct, sir.

Latest Apple version version (with everything up to date and packed in a nice dmg), should always be downloaded from the official page: https://www.comfy.org/download

Keep in mind ComfyUI updates very often. To keep track of latest version and changes:
https://github.com/Comfy-Org/desktop/releases
and
https://github.com/Comfy-Org/ComfyUI/releases (don't download any of those releases, they are for Windows, just check the link for specific update references)

[bbhxwl]

Correct, sir.

Latest Apple version version (with everything up to date and packed in a nice dmg), should always be downloaded from the official page: https://www.comfy.org/download

Keep in mind ComfyUI updates very often. To keep track of latest version and changes: https://github.com/Comfy-Org/desktop/releases and https://github.com/Comfy-Org/ComfyUI/releases (don't download any of those releases, they are for Windows, just check the link for specific update references)

I tested downloading the latest version and still have this error.
ComfyUI 0.8.27 - arm64.dmg

[gershu-ar]

Might be an issue with the base template offered by ComfyUI?
Try checking the subgraph of the "image to video (wan2.2)" node so you can pin point which is the node that is giving you the story; if there're none (only that error message), I'd fill an issue as a potential bug for the base template: https://github.com/Comfy-Org/ComfyUI/issues