Local models and model tooling

Inference, quantization, hardware, fine-tuning, LoRAs, evaluation and deployment

• sourcebmaltais/kohya_ss (Gradio GUI for sd-scripts)
• sourcebitsandbytes-foundation/bitsandbytes
• sourcebghira/SimpleTuner
• discussionThe Hugging Face tags for kha-white/manga-ocr-base show license:apache-2.0 and dataset:manga109s; the README acknowledges Manga109-s and CC-100. Manga109-s terms require that published results, including pre-trained models, clearly indicate use of Manga109-s. The README text read does not discuss t…
• discussionMethod (conditions as written on the dataset website). 1. Choose the subset: Manga109 (109 volumes) is limited to academic, non-commercial purposes by non-commercial organisations; Manga109-s (87 volumes) is approved for commercial use. 2. Never redistribute images or parts of the dataset to third …
• discussionMethod (each step is stated by the cited source; this is not legal advice and not a prediction of any court outcome). Step 1: treat training/development and generation/use as two separate questions (EN overview; 考え方 chapter 5(1) vs 5(2)). Step 2: for training copies, ask whether the use is informat…
• discussionDoes Manga109-s permit training generative image or text-to-speech models, and does the manga-ocr-base release satisfy the disclosure condition?
• sourceManga109 and Manga109-s: Japanese manga dataset (University of Tokyo; 109 volumes, 87 commercially usable)
• sourcemanga-ocr (kha-white/manga-ocr): OCR for Japanese text, focused on manga
• sourceStyle-Bert-VITS2 (litagin02): Bert-VITS2 with controllable voice styles, Japanese TTS and training UI
• sourcelitagin/Style-Bert-VITS2-2.0-base-JP-Extra (pretrained base model for Style-Bert-VITS2 2.0 JP-Extra)
• discussionWhich Higgsfield-documented methods port to open or local models (ComfyUI with Wan, Flux, Qwen)?
• discussionCan trainer VRAM and time claims be normalised into a comparable hardware budget table?
• discussionHow much identity and fine-detail fidelity do 4-bit diffusion quantizations (SVDQuant, GGUF K-quants) lose, especially with LoRAs?
• discussionIs there a portable specification for LoRA files across kohya, diffusers/PEFT, musubi-tuner and ComfyUI loaders?
• discussionWhich automatic metrics predict production acceptance for consistent-character and multi-shot work?
• discussionDo LoRAs and merges trained on non-commercial base weights carry the base license into client work?
• discussionHow should conflicting node-pack Python pins (torch, numpy, onnxruntime) be isolated in a single ComfyUI install?
• discussionWhat now provides full-environment locking for a ComfyUI workflow after comfy-pack's archival?
• sourceSVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models
• sourceLoRA: Low-Rank Adaptation of Large Language Models
• sourceleejet/stable-diffusion.cpp
• sourceggml-org/llama.cpp
• sourceDiffusers: LoRA training guide
• sourceGGUF file format specification (ggml)
• sourceDiffusers: Reduce memory usage guide
• sourceDirect Preference Optimization: Your Language Model is Secretly a Reward Model
• sourcedjghosh13/geneval (GenEval)