Emaonly

This is a visual technique where a trader applies multiple EMAs of varying lengths (e.g., 10, 20, 30, 40, 50) to the chart.

You should avoid the EMA-only version and use the "full" or "non-pruned" model only if you intend to perform , such as training a new Dreambooth or LoRA model from scratch. For everyday image generation, EMA-only is the industry standard. emaonly

| Trader Type | EMA Settings | Purpose | | :--- | :--- | :--- | | | 9 EMA & 21 EMA | Fast reactions for short-term trades. | | Day Trader | 9 EMA, 21 EMA, 50 EMA | Identifying intraday momentum and pullbacks. | | Swing Trader | 50 EMA & 200 EMA | Identifying major macro trends. | This is a visual technique where a trader

When a model is trained, its weights (the internal numbers that determine how it creates an image) fluctuate as it learns from new data. To produce a more stable and reliable result, developers often maintain a running average of these weights. The "emaonly" version of a model contains only these averaged weights, stripped of the extra data used for ongoing training. Key Benefits of emaonly Checkpoints | Trader Type | EMA Settings | Purpose

There are two primary ways traders utilize an EMA-only strategy:

For 99% of AI artists, is the gold standard. It provides the full creative power of the Stable Diffusion architecture in a leaner, more efficient package, leaving you more room for your growing collection of specialized styles and characters.

While emaonly is perfect for using a model, it is not ideal if you plan to train a model further. If you are a developer looking to perform or Dreambooth training, you may need the "full" or non-ema version. The extra data stripped from emaonly files contains the "momentum" of the training process, which helps the model continue learning from a specific point without "forgetting" what it previously knew. How to Use emaonly in Your Workflow