LoRA Trainer Guide
Overview
Help users pick the right training configuration and understand what the parameters actually do. Ships with 38 real-world presets covering SD1.5, SDXL, SD3, and Flux1 — standard LoRA, LyCORIS variants (LoHA, LoKR, LoCon, iA3), finetuning, and dreambooth.
The presets live in the presets/ folder as JSON files. They're directly usable with Kohya SS, sd-scripts, and the Ktiseos-Nyx-Trainer.
When to Use
- User wants to train a LoRA and needs help picking settings
- User has a preset and wants to understand what the params mean
- User is troubleshooting training results ("my outputs look fried, what do I change?")
- User wants to compare LoRA types (standard vs LoHA vs LoKR etc.)
Opening Behavior
Always start by asking: "How familiar are you with LoRA training?"
Adjust your responses based on their answer:
- New to it — explain concepts, define terms, walk them through choices step by step
- Done it a few times — skip basics, focus on "which preset fits your use case" and key params to tweak
- Experienced — be concise, jump straight to recommendations, reference specific params
Preset Selection Flow
Walk the user through these questions to find the right preset:
1. What model architecture?
| Architecture |
Resolution |
Key differences |
| SD 1.5 |
512x512 |
Oldest, most presets available, clip_skip=2 common |
| SDXL |
1024x1024 |
Dual text encoder, needs sdxl: true, more VRAM |
| SD3 |
Varies |
Newer architecture, limited preset support |
| Flux1 |
512x512+ |
Requires fp8, T5 text encoder, different LoRA type |
2. What training type?
| Type |
Use when |
File size |
Flexibility |
| Standard LoRA |
General purpose, good balance of quality and size |
Medium |
High |
| LoHA |
Easier concepts, multi-concept, want better generalization |
Medium |
Very high |
| LoKR (small, factor=-1) |
Want tiny file size (<2.5MB) |
Very small |
Lower (model-specific) |
| LoKR (large, factor~8) |
Want LoRA-like quality with Kronecker math |
Medium-large |
High |
| LoCon |
Need convolution layer training specifically |
Medium |
Medium |
| iA3 |
Style learning, want ultra-tiny files (<1MB) |
Tiny |
Low (hard to transfer) |
| Finetune |
Full model training, best possible quality |
Very large |
Highest |
| Dreambooth |
Subject-specific training |
Very large |
High |
3. Recommend a preset
Filter the presets/ folder by model_type and LoRA_type from the JSON files. Each preset has this structure:
{
"name": "Human-readable name",
"description": "What this preset is optimized for",
"model_type": "SDXL",
"config": {
"LoRA_type": "Standard",
"optimizer": "AdamW8bit",
"network_dim": 32,
"network_alpha": 32,
"learning_rate": 0.0001,
...
}
}
Present matching presets as numbered options with a brief explanation of what makes each one different (usually the optimizer and dim/alpha choices).
What to Tweak
After selecting a preset, highlight the params the user is most likely to want to adjust. Don't dump every parameter — focus on what actually matters.
Always worth checking
| Parameter |
What it does |
Typical range |
Watch out for |
network_dim |
LoRA rank — higher = more expressive, bigger file |
8–128 |
Higher isn't always better; 32 is a solid default |
network_alpha |
Scales the learning rate relative to dim |
1 to same-as-dim |
alpha=1 vs alpha=dim behave VERY differently — see pitfalls |
learning_rate |
How fast the model learns |
1e-5 to 1e-3 |
Depends heavily on optimizer choice |
train_batch_size |
Images per step |
1–8 |
Higher = more VRAM, may need to adjust LR |
max_train_epochs / max_train_steps |
When to stop |
Varies |
More isn't better — watch for overtraining |
optimizer |
Learning algorithm |
See reference.md |
Each optimizer wants different LR ranges |
save_every_n_epochs |
Checkpoint frequency |
1–5 |
Save often so you can pick the best epoch |
Per-architecture specifics
SD 1.5:
clip_skip: usually 2 for anime models, 1 for realisticmax_resolution: 512,512mixed_precision: fp16 or bf16
SDXL:
sdxl: must be truemax_resolution: 1024,1024sdxl_no_half_vae: usually true (prevents NaN)clip_skip: usually 1- Consider
min_snr_gamma: 5 helps stabilize training
Flux1:
fp8_base: must be true (Flux needs fp8)flux1_checkbox: trueLoRA_type: must be "Flux1"t5xxl_max_token_length: 512timestep_sampling: "sigmoid"discrete_flow_shift: 3model_prediction_type: "raw"- Cache text encoder outputs to disk (saves VRAM)
Common Pitfalls
| Problem |
Likely cause |
Fix |
| Loss goes NaN |
LoHA + high dim, or LR too high |
Lower LR, reduce dim, or switch to LoRA |
| Output looks nothing like training data |
Undertrained or LR too low |
More epochs, higher LR, check captions |
| Output is copy-paste of training images |
Overtrained |
Fewer epochs, lower dim, use an earlier checkpoint |
| Style bleeds into everything |
Model learned style too hard |
Use LoHA (dampens style), lower dim, add regularization images |
| Colors are washed out / wrong |
VAE issues on SDXL |
Set sdxl_no_half_vae: true |
| Flux training OOMs immediately |
Not using fp8 |
Set fp8_base: true, cache text encoder outputs |
| alpha=1 trains differently than expected |
alpha scales effective LR |
alpha=1 with dim=32 means effective LR is multiplied by 1/32. Set alpha=dim for "normal" behavior, or adjust LR accordingly |
| Increasing batch size makes results worse |
LR wasn't scaled |
When you increase batch size, increase LR proportionally |
See Also
reference.md in this skill folder — full parameter definitions, algorithm comparison tables, optimizer reference- Presets are in
presets/ — each is a self-contained JSON file ready to load
1---2name: lora-trainer-guide3description: Use when helping users configure LoRA training — select presets, understand parameters, and pick the right settings for their model type. Covers SD1.5, SDXL, SD3, and Flux1 with real-world presets.4---5
6# LoRA Trainer Guide
7
8## Overview
9
10Help users pick the right training configuration and understand what the parameters actually do. Ships with 38 real-world presets covering SD1.5, SDXL, SD3, and Flux1 — standard LoRA, LyCORIS variants (LoHA, LoKR, LoCon, iA3), finetuning, and dreambooth.
11
12The presets live in the `presets/` folder as JSON files. They're directly usable with Kohya SS, sd-scripts, and the Ktiseos-Nyx-Trainer.
13
14## When to Use
15
16- User wants to train a LoRA and needs help picking settings
17- User has a preset and wants to understand what the params mean
18- User is troubleshooting training results ("my outputs look fried, what do I change?")
19- User wants to compare LoRA types (standard vs LoHA vs LoKR etc.)
20
21## Opening Behavior
22
23**Always start by asking:** "How familiar are you with LoRA training?"
24
25Adjust your responses based on their answer:
26- **New to it** — explain concepts, define terms, walk them through choices step by step
27- **Done it a few times** — skip basics, focus on "which preset fits your use case" and key params to tweak
28- **Experienced** — be concise, jump straight to recommendations, reference specific params
29
30## Preset Selection Flow
31
32Walk the user through these questions to find the right preset:
33
34### 1. What model architecture?
35
36| Architecture | Resolution | Key differences |
37|-------------|-----------|-----------------|
38| SD 1.5 | 512x512 | Oldest, most presets available, clip_skip=2 common |
39| SDXL | 1024x1024 | Dual text encoder, needs `sdxl: true`, more VRAM |
40| SD3 | Varies | Newer architecture, limited preset support |
41| Flux1 | 512x512+ | Requires fp8, T5 text encoder, different LoRA type |
42
43### 2. What training type?
44
45| Type | Use when | File size | Flexibility |
46|------|----------|-----------|-------------|
47| Standard LoRA | General purpose, good balance of quality and size | Medium | High |
48| LoHA | Easier concepts, multi-concept, want better generalization | Medium | Very high |
49| LoKR (small, factor=-1) | Want tiny file size (<2.5MB) | Very small | Lower (model-specific) |
50| LoKR (large, factor~8) | Want LoRA-like quality with Kronecker math | Medium-large | High |
51| LoCon | Need convolution layer training specifically | Medium | Medium |
52| iA3 | Style learning, want ultra-tiny files (<1MB) | Tiny | Low (hard to transfer) |
53| Finetune | Full model training, best possible quality | Very large | Highest |
54| Dreambooth | Subject-specific training | Very large | High |
55
56### 3. Recommend a preset
57
58Filter the `presets/` folder by `model_type` and `LoRA_type` from the JSON files. Each preset has this structure:
59
60```json
61{
62 "name": "Human-readable name",
63 "description": "What this preset is optimized for",
64 "model_type": "SDXL",
65 "config": {
66 "LoRA_type": "Standard",
67 "optimizer": "AdamW8bit",
68 "network_dim": 32,
69 "network_alpha": 32,
70 "learning_rate": 0.0001,
71 ...
72 }
73}
74```
75
76Present matching presets as numbered options with a brief explanation of what makes each one different (usually the optimizer and dim/alpha choices).
77
78## What to Tweak
79
80After selecting a preset, highlight the params the user is most likely to want to adjust. Don't dump every parameter — focus on what actually matters.
81
82### Always worth checking
83
84| Parameter | What it does | Typical range | Watch out for |
85|-----------|-------------|---------------|---------------|
86| `network_dim` | LoRA rank — higher = more expressive, bigger file | 8–128 | Higher isn't always better; 32 is a solid default |
87| `network_alpha` | Scales the learning rate relative to dim | 1 to same-as-dim | alpha=1 vs alpha=dim behave VERY differently — see pitfalls |
88| `learning_rate` | How fast the model learns | 1e-5 to 1e-3 | Depends heavily on optimizer choice |
89| `train_batch_size` | Images per step | 1–8 | Higher = more VRAM, may need to adjust LR |
90| `max_train_epochs` / `max_train_steps` | When to stop | Varies | More isn't better — watch for overtraining |
91| `optimizer` | Learning algorithm | See reference.md | Each optimizer wants different LR ranges |
92| `save_every_n_epochs` | Checkpoint frequency | 1–5 | Save often so you can pick the best epoch |
93
94### Per-architecture specifics
95
96**SD 1.5:**
97- `clip_skip`: usually 2 for anime models, 1 for realistic
98- `max_resolution`: 512,512
99- `mixed_precision`: fp16 or bf16
100
101**SDXL:**
102- `sdxl`: must be `true`
103- `max_resolution`: 1024,1024
104- `sdxl_no_half_vae`: usually `true` (prevents NaN)
105- `clip_skip`: usually 1
106- Consider `min_snr_gamma`: 5 helps stabilize training
107
108**Flux1:**
109- `fp8_base`: must be `true` (Flux needs fp8)
110- `flux1_checkbox`: `true`
111- `LoRA_type`: must be "Flux1"
112- `t5xxl_max_token_length`: 512
113- `timestep_sampling`: "sigmoid"
114- `discrete_flow_shift`: 3
115- `model_prediction_type`: "raw"
116- Cache text encoder outputs to disk (saves VRAM)
117
118## Common Pitfalls
119
120| Problem | Likely cause | Fix |
121|---------|-------------|-----|
122| Loss goes NaN | LoHA + high dim, or LR too high | Lower LR, reduce dim, or switch to LoRA |
123| Output looks nothing like training data | Undertrained or LR too low | More epochs, higher LR, check captions |
124| Output is copy-paste of training images | Overtrained | Fewer epochs, lower dim, use an earlier checkpoint |
125| Style bleeds into everything | Model learned style too hard | Use LoHA (dampens style), lower dim, add regularization images |
126| Colors are washed out / wrong | VAE issues on SDXL | Set `sdxl_no_half_vae: true` |
127| Flux training OOMs immediately | Not using fp8 | Set `fp8_base: true`, cache text encoder outputs |
128| alpha=1 trains differently than expected | alpha scales effective LR | alpha=1 with dim=32 means effective LR is multiplied by 1/32. Set alpha=dim for "normal" behavior, or adjust LR accordingly |
129| Increasing batch size makes results worse | LR wasn't scaled | When you increase batch size, increase LR proportionally |
130
131## See Also
132
133- `reference.md` in this skill folder — full parameter definitions, algorithm comparison tables, optimizer reference
134- Presets are in `presets/` — each is a self-contained JSON file ready to load
majiayu000