From 2f98c243603aaf461eda72b86e447c1b047623c2 Mon Sep 17 00:00:00 2001 From: comfyanonymous Date: Thu, 30 Jan 2025 02:12:43 -0500 Subject: [PATCH 1/2] Update Readme with link to instruction for Nvidia 50 series. --- README.md | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index ab589f4e0..311dbb294 100644 --- a/README.md +++ b/README.md @@ -130,6 +130,8 @@ Simply download, extract with [7-Zip](https://7-zip.org) and run. Make sure you If you have trouble extracting it, right click the file -> properties -> unblock +If you have a 50 series Blackwell card like a 5090 or 5080 see [this discussion thread](https://github.com/comfyanonymous/ComfyUI/discussions/6643) + #### How do I share models between another UI and ComfyUI? See the [Config file](extra_model_paths.yaml.example) to set the search paths for models. In the standalone windows build you can find this file in the ComfyUI directory. Rename this file to extra_model_paths.yaml and edit it with your favorite text editor. @@ -186,7 +188,7 @@ Additional discussion and help can be found [here](https://github.com/comfyanony Nvidia users should install stable pytorch using this command: -```pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu124``` +```pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu126``` This is the command to install pytorch nightly instead which might have performance improvements: From 8d8dc9a262bfee9f7eb3a2fc381b79a69274f225 Mon Sep 17 00:00:00 2001 From: comfyanonymous Date: Thu, 30 Jan 2025 06:49:52 -0500 Subject: [PATCH 2/2] Allow batch of different sigmas when noise scaling. --- comfy/model_sampling.py | 3 +++ 1 file changed, 3 insertions(+) diff --git a/comfy/model_sampling.py b/comfy/model_sampling.py index 4370516b9..ff27b09a8 100644 --- a/comfy/model_sampling.py +++ b/comfy/model_sampling.py @@ -31,6 +31,7 @@ class EPS: return model_input - model_output * sigma def noise_scaling(self, sigma, noise, latent_image, max_denoise=False): + sigma = sigma.view(sigma.shape[:1] + (1,) * (noise.ndim - 1)) if max_denoise: noise = noise * torch.sqrt(1.0 + sigma ** 2.0) else: @@ -61,9 +62,11 @@ class CONST: return model_input - model_output * sigma def noise_scaling(self, sigma, noise, latent_image, max_denoise=False): + sigma = sigma.view(sigma.shape[:1] + (1,) * (noise.ndim - 1)) return sigma * noise + (1.0 - sigma) * latent_image def inverse_noise_scaling(self, sigma, latent): + sigma = sigma.view(sigma.shape[:1] + (1,) * (latent.ndim - 1)) return latent / (1.0 - sigma) class ModelSamplingDiscrete(torch.nn.Module):