Sana
Sana copied to clipboard
NVlabs
SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer
@torch.no_grad() def sample(self, shape): self.eval() zt = torch.randn(shape).cuda() ## init the noise noise_schedule = NoiseScheduleFlow(schedule='discrete_flow') model_fn_continuous = model_wrapper( self.model, noise_schedule, model_type="flow", model_kwargs={}, guidance_type="uncond" , condition= None, classifier_kwargs={}, interval_guidance = [0,...
Hi, when the face region is relatively small, it tends to become distorted. Is this due to the high compression ratio of dc-ae? Is there a solution to this problem?...
Hi. I didn't pay attention to it before, but I am currently testing a training base on the "Sana_600M_img512.yaml" config file with a dataset ok 20k images (single 4090 GPU...
Unfortunately, due to its hardware requirements, many people cannot run it on their own computers. That's why I created a Runpod.io template for it so that anyone can use it...
The generated 4K images appear somewhat blurry and out of focus. I have partially enlarged all the images, including the comparison images, below for easier comparison. For example, this image:...
After taking a look at benchmarks doing comparisons, wondering if there are ways to show that Sana is really flexible or not - Flux https://blackforestlabs.ai/announcing-black-forest-labs/ - Playground v3 https://arxiv.org/html/2409.10695v2 -...
Your team is doing a great job. If possible, it would be greatly appreciated if you could mention somewhere in your documentation that when the PAG (Perturbed-Attention Guidance) Guidance Scale...
- Correct "border-raduis" to "border-radius" - Insert missing article "a" in "modern giant diffusion model" - Insert missing article "a" in "decoder-only small LLM"
Hello, I would like to inquire about Figure 3 in SANA-Srint. In the two subplots, does the "training gradient norm" mentioned refer to the gradients of trainable parameters \(\theta\) during...
Hi, First of all, thank you so much for this great research work. I’m using this model as the backbone for my own research. With a major conference deadline approaching,...
