User contributions for Civitai
Jump to navigation
Jump to search
11 October 2023
- 07:5307:53, 11 October 2023 diff hist +323 N Noise Offset Created page with "First described by researchers at Crosslabs. A method of introducing true "darkness" (and highlights) into models at the training stage. Note that Noise Offset will increase dampening. A Setting of 0.1 will make a LoRA's colors more vivid. Default is 0. Category:Training" current Tag: Visual edit
- 07:5207:52, 11 October 2023 diff hist +303 N Network Rank Dimension Created page with "Also expressed as Net Dim or Rank. This setting affects the “power” of the model in displaying the concepts trained within. Higher values result in a larger LoRA and more training time, but may capture the element to be trained with better fidelity. Category:Training" current Tag: Visual edit
- 07:5007:50, 11 October 2023 diff hist +417 N Network Alpha Created page with "Closely related to the Network Rank (Dimension), the smaller the Network Alpha value, the larger the LoRA neural net weights. Can be used to Dampen, or "slow down" learning. Alpha of 16 and Network Rank (Dim) of 32 halves the Learning Rate. If Alpha and Network Rank are set to the same value, there will be no effect on Learning Rate. Category:Training" current Tag: Visual edit
- 07:4807:48, 11 October 2023 diff hist +388 N Mixed Precision Created page with "Weight data is originally 32bit units, but we can gain considerable VRAM savings by training with 16bit precision. LoRA can be successfully trained with FP16 (16bit precision). BF16 is a format devised to provide the VRAM savings of FP16 with the accuracy of FP32 (32bit). BF16 may only work on the latest generation GPUs. Category:Training" current Tag: Visual edit
- 07:4707:47, 11 October 2023 diff hist +157 N Memory Efficient Attention Created page with "When enabled, results in greatly lowered VRAM consumption while training, but is slower than Xformers. Default is OFF. Category:Training" current Tag: Visual edit
- 07:4307:43, 11 October 2023 diff hist +214 N Learning Rate Scheduler Created page with "Learning Rate is a parameter of the Optimizer. A Learning Rate Scheduler adjusts the learning rate according to a predefined schedule during the training process. Category:Training" current Tag: Visual edit
- 07:4207:42, 11 October 2023 diff hist +267 N Learning Rate Created page with "The larger the Learning Rate value, the faster the LoRA training, but the more details are missed during training. Low Learning Rates are typically desirable, to retain flexibility in the LoRA. Category:Training" current Tag: Visual edit
- 07:4107:41, 11 October 2023 diff hist +22 Learn Dampening No edit summary current Tag: Visual edit
- 07:4007:40, 11 October 2023 diff hist +364 N Learn Dampening Created page with "The goal of training is (generally) to fit the most number of Steps in, without Overcooking. Certain settings, by design, or coincidentally, "dampen" learning, allowing us to train more steps before the LoRA appears Overcooked. Some settings which affect Dampening include Network Alpha and Noise Offset." Tag: Visual edit
- 07:3707:37, 11 October 2023 diff hist +102 N Hugging Face Created page with "A community/data science platform providing tools to build, train, and deploy machine learning models." Tag: Visual edit
- 07:3607:36, 11 October 2023 diff hist +153 N Gradio Created page with "A web-browser based interface framework, specifically for Machine Learning applications. Auto1111 WebUI runs in a Gradio interface." current Tag: Visual edit
- 07:3507:35, 11 October 2023 diff hist +170 N Hallucinations Created page with "Sometimes LLM models like ChatGPT produce information that sounds plausible but is nonsensical or entirely false. This is called a Hallucination." current Tag: Visual edit
- 07:3507:35, 11 October 2023 diff hist +82 N Inpainting Created page with "The practice of removing or replacing objects in an image based on a painted mask." current Tag: Visual edit
- 07:3307:33, 11 October 2023 diff hist +56 N LAION Created page with "See Large-scale Artificial Intelligence Open Network" Tag: Visual edit
- 07:3207:32, 11 October 2023 diff hist +97 N Large-scale Artificial Intelligence Open Network Created page with "A non-profit organization, providing data sets, tools, and models, for machine learning research." Tag: Visual edit
- 07:3207:32, 11 October 2023 diff hist +146 N Lanczos Created page with "An interpolation method used to compute new values for sampled data. In this case, used to upscale images. Named after creator, Cornelius Lanczos." Tag: Visual edit
- 07:3107:31, 11 October 2023 diff hist +247 N Latent Diffusion Created page with "A type of diffusion model that contains compressed image representations instead of the actual images. This type of model allows the storage of a large amount of data that can be used by encoders to reconstruct images from textual or image inputs." Tag: Visual edit
- 07:3107:31, 11 October 2023 diff hist +131 N Latent Mirroring Created page with "Applies mirroring to the latent images mid-generation to produce anything from subtly balanced compositions to perfect reflections." current Tag: Visual edit
- 07:3007:30, 11 October 2023 diff hist +174 N Latent Space Created page with "The information-dense space where the diffusion model’s image representation, attention, and transformation are merged and form the initial noise for the diffusion process." current Tag: Visual edit
- 07:3007:30, 11 October 2023 diff hist +41 N LDSR Created page with "See Latent Diffusion Super Resolution" Tag: Visual edit
- 07:3007:30, 11 October 2023 diff hist +54 N Latent Diffusion Super Resolution Created page with "A method to increase the dimensions/quality of images." current Tag: Visual edit
- 07:2907:29, 11 October 2023 diff hist +228 N Kohya Page created Tag: Visual edit
- 07:2707:27, 11 October 2023 diff hist +161 N Gradient Checkpointing Created page with "Enables intermediate saving of gradients, reduces overall training speed but uses less VRAM. Has no effect on the training results. Category:Training" current Tag: Visual edit
- 07:2607:26, 11 October 2023 diff hist +82 N GIT Created page with "See Generative Image-to-Text Transformer." Tag: Visual edit
- 07:2407:24, 11 October 2023 diff hist +211 N Generative Image-to-Text Transformer Created page with "Generative Image-to-Text (GIT), first discussed in [https://arxiv.org/abs/2205.14100 this paper], was trained on 20 million image-text pairs, and further fine-tuned on TextCaps. A robust image-to-text processor." Tag: Visual edit
- 07:2307:23, 11 October 2023 diff hist +253 N Flip Augmentation Created page with "Artificially double the number of training images by performing a horizontal flip. If your data is not left-right symmetrical (which will be the case when training humans, subjects), this option should be avoided. Category:Training" current Tag: Visual edit
- 07:2007:20, 11 October 2023 diff hist +294 N Color Augmentation Created page with "Artificially increases the number of training image variations by changing the image hue during learning; supposedly improving model fidelity. When enabled, Cache latents cannot be used due to the training images changing dynamically during training. Category:Training" current Tag: Visual edit
- 07:1907:19, 11 October 2023 diff hist +457 N Clip Skip Created page with "The Text Encoder uses a mechanism called "CLIP", made up of 12 layers (corresponding to the 12 layers of the Stable Diffusion neural network). Clip Skip specified the layer number Xth from the end. Clip Skip of 2 will send the penultimate layer's output vector to the Attention block. Unless the base model you're training against was trained (or Mixed) with Clip Skip 2, you can use 1. SDXL does not benefit from Clip Skip 2." current Tag: Visual edit
- 07:1807:18, 11 October 2023 diff hist +211 N Captioning Created page with "The process of describing your input training images to help Stable Diffusion understand what it's looking at. Captioning can be done by hand, or via a number of Caption creation tools. Category:Training" Tag: Visual edit
- 07:1507:15, 11 October 2023 diff hist +12 Cache latents No edit summary current Tag: Visual edit
- 07:1407:14, 11 October 2023 diff hist +58 N Cache latents to disk Created page with "See Cache Latents. Category:Training" Tag: Visual edit
- 07:1307:13, 11 October 2023 diff hist +187 N Cache latents Created page with "To speed up the LoRA training process, pre-generate the latent representation of the training images in advance, and save into system memory. Category:Training" Tag: Visual edit
- 07:1107:11, 11 October 2023 diff hist +22 Loss No edit summary Tag: Visual edit
- 07:1007:10, 11 October 2023 diff hist +22 Overfitting No edit summary Tag: Visual edit
- 07:1007:10, 11 October 2023 diff hist +22 Underfitting No edit summary Tag: Visual edit
- 07:1007:10, 11 October 2023 diff hist +22 Training Data No edit summary Tag: Visual edit
- 07:0807:08, 11 October 2023 diff hist +22 Epoch No edit summary Tag: Visual edit
- 07:0807:08, 11 October 2023 diff hist +43 N Category:Training Page created current Tag: Visual edit
- 07:0707:07, 11 October 2023 diff hist +22 Bucket No edit summary current Tag: Visual edit
- 07:0107:01, 11 October 2023 diff hist +27 Civitai Trainer No edit summary Tag: Visual edit
- 06:5906:59, 11 October 2023 diff hist −178 Samplers Replaced content with "See Sampler." Tags: Replaced Visual edit
- 06:5706:57, 11 October 2023 diff hist +23 Low-Rank Adaptation No edit summary Tag: Visual edit
- 06:5606:56, 11 October 2023 diff hist −1,832 LoRA Replaced content with "See Low-Rank Adaptation. Category:Generative AI" Tags: Replaced Visual edit
- 06:5506:55, 11 October 2023 diff hist +1,854 N Low-Rank Adaptation Page created Tag: Visual edit
- 06:5406:54, 11 October 2023 diff hist +348 N Bucket Page created Tag: Visual edit
- 06:5306:53, 11 October 2023 diff hist −66 BLIP No edit summary Tag: Visual edit
- 06:5206:52, 11 October 2023 diff hist +173 N Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation Page created Tag: Visual edit
- 06:4806:48, 11 October 2023 diff hist +173 N BLIP Created page with "Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation (BLIP), created by Salesforce, is a solid, simple image-to-text processor." Tag: Visual edit
- 06:4406:44, 11 October 2023 diff hist +69 Image Restoration Using Shifted Window Transformer Message No edit summary current Tag: Visual edit
- 06:4206:42, 11 October 2023 diff hist +41 Vladmandic No edit summary Tag: Visual edit