This repository implements DreamBooth subject fine-tuning for Stable Diffusion using textual inversion.
Open source
24 alternativesexamples/stable-diffusion-finetuning at main · LambdaLabsML/examples
github.com/LambdaLabsML/examples/tree/main/sta…
ML Training & Optimization
Open-source projectOpen source
This repository includes examples and a guide for fine-tuning Stable Diffusion models.
About examples/stable-diffusion-finetuning at main · LambdaLabsML/examples
This repository explains how to fine-tune Stable Diffusion on a specific image style as an alternative to prompt engineering. It uses a Pokémon example to train a text-to-image model that generates custom Pokémon from text prompts, with a fine-tuning notebook, a modified training code fork, a base YAML configuration, a captioned dataset, and trained weights.
Training requires images paired with text captions. The example uses almost a thousand Pokémon images captioned with BLIP and describes running about 15,000 steps, needing more than 24GB of VRAM with adjustments to batch size and gradient accumulation on different GPUs.
Key features
- Fine-tune Stable Diffusion on custom image-caption dataset
- Example notebook for Pokémon fine-tuning workflow
- Automatic caption generation using BLIP model
- Pre-captioned Pokémon dataset on Hugging Face
- Base YAML configuration for fine-tuning training
- Access to trained model weights in Diffusers format
Best for: People with GPU access who want to adapt Stable Diffusion to a specific image style.
Good for
- Includes training code, captioned dataset, and trained weights for the example
- Shows how to modify the YAML data configuration for other datasets
- Documents example hardware setup and adjustments for different GPUs
Watch out for
- Requires images paired with text captions; the base Pokémon set had none
- Needs more than 24GB of VRAM and manual batch-size tuning on smaller setups
- Auto-generated BLIP captions are described as reasonably accurate but not perfect
examples/stable-diffusion-finetuning at main · LambdaLabsML/examples at a glance
- Type
- Open-source project
- Pricing
- Not confirmed. Check the site
- Open source
- Yes · Source code
- Language
- English
- Last checked
- Oct 5, 2026 · Online when checked
- Listed
- Oct 5, 2026
examples/stable-diffusion-finetuning at main · LambdaLabsML/examples alternatives
Tools and sites you could use instead, ranked by how closely they match.
DreamBooth describes fine-tuning text-to-image diffusion models to generate images of specific subjects.
PyTorch Lightning extension that accelerates and enhances foundation model experimentation with flexible fine-tuning schedules.
Open source
24 alternatives
LightlyTrain supports pretraining, fine-tuning, and distillation for computer vision models.
Open source
24 alternatives
P-tuning v2 is a deep prompt tuning strategy designed as an alternative to full model fine-tuning.
Open source
24 alternatives
oumi supports fine-tuning, evaluation, and deployment of open-weight language models.
Open source
24 alternativesPopular in ML Training & Optimization
Well-known listings from the same topic.
Edwardedwardlib.orgAerosolveaerosolve.airbnb.techDataturedatature.ioTogether AItogether.aiOptunaoptuna.orgCatBoostcatboost.ai
examples/stable-diffusion-finetuning at main · LambdaLabsML/examples FAQ
What hardware does the example use?
2xA6000 GPUs on Lambda GPU Cloud for about 15,000 steps in about 6 hours. Training may run on a single or lower-spec GPU with more than 24GB of VRAM after adjusting batch size and gradient accumulation steps.
What dataset is used and where is the captioned version?
A Pokémon dataset from FastGAN with almost a thousand images, captioned with BLIP and shared as lambdalabs/pokemon-blip-captions on Hugging Face.
Can I generate Pokémon images without training the model myself?
Yes, the page points to using the notebook or trying it on Replicate, and lists separate downloads for model weights and code.
Is something wrong or out of date?
Tell us and an editor will check it.
Own this website?
Claim this listing
Claim the listing and suggest updates yourself. An editor checks every claim first.