minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[Diffusion][Cookbook] Add Krea-2 cookbook (#29051)

Browse Source
This commit is contained in:
Yihao Wang
2026-06-23 08:02:08 -07:00
committed by GitHub
parent 31d71c47af
commit b5e0965b07
4 changed files with 218 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs_new/cookbook/diffusion/Krea/Krea-2.mdx
+205
View File
@@ -0,0 +1,205 @@
---
title: Krea-2
metatags:
description: "Deploy Krea-2 with SGLang - fast, high-quality text-to-image generation."
---
## 1. Model Introduction
[Krea-2](https://huggingface.co/krea/Krea-2-Turbo) is a high-quality text-to-image diffusion model from [Krea](https://www.krea.ai/). It ships in two variants that share the same backbone and differ only in their sampling recipe:
- **[Krea-2-Turbo](https://huggingface.co/krea/Krea-2-Turbo)** - a distilled, few-step model that produces photorealistic images in only **8 inference steps** with no classifier-free guidance (`guidance_scale = 1.0`), ideal for fast and interactive generation.
- **[Krea-2-Raw](https://huggingface.co/krea/Krea-2-Raw)** - the base (non-distilled) model that trades speed for maximum fidelity, using a longer schedule (~52 steps) with classifier-free guidance (`guidance_scale ≈ 4.5`).
Both variants are built on a single-stream MMDiT with a Qwen3-VL text encoder and the Qwen-Image VAE, and are distributed in the standard diffusers layout (a `model_index.json` plus sharded `transformer/`, `text_encoder/`, `vae/`, `tokenizer/`, and `scheduler/` folders). SGLang loads them **natively** - just point `--model-path` at the repo, no conversion step required.
**Key Features:**
- **Two variants, one pipeline**: switch between fast (Turbo) and high-fidelity (Raw) by changing only the model path and the sampling settings.
- **Photorealistic generation** at 1024x1024 and other resolutions.
- **Native diffusers loading**: components (DiT, text encoder, VAE, scheduler) are read straight from the repo's `model_index.json`.
For more details, see the [Krea-2-Turbo](https://huggingface.co/krea/Krea-2-Turbo) and [Krea-2-Raw](https://huggingface.co/krea/Krea-2-Raw) HuggingFace pages.
## 2. SGLang-diffusion Installation
SGLang-diffusion offers multiple installation methods. You can choose the most suitable installation method based on your hardware platform and requirements.
Please refer to the [official SGLang-diffusion installation guide](https://github.com/sgl-project/sglang/blob/main/python/sglang/multimodal_gen/docs/install.md) for installation instructions.
## 3. Model Deployment
This section covers deploying Krea-2-Turbo for fast, high-quality image generation.
### 3.1 Basic Configuration
Krea-2-Turbo generates high-quality images in only 8 inference steps. Launch the server with:
```bash Command
sglang serve \
--model-path krea/Krea-2-Turbo \
--num-gpus 1 \
--port 30000
```
The step count and guidance scale are **request-time** settings (see [API Usage](#4-api-usage)); Krea-2-Turbo defaults to 8 steps with `guidance_scale = 1.0`.
### 3.2 Configuration Tips
Currently supported optimizations are listed [here](https://github.com/sgl-project/sglang/blob/main/python/sglang/multimodal_gen/docs/support_matrix.md).
- `--num-gpus`: Number of GPUs to use.
- `--tp-size`: Tensor parallelism size (the recommended multi-GPU path for Krea-2). Its attention heads (48, with 12 KV heads) and text heads (20) are divisible by a tensor-parallel size of 1, 2, or 4.
## 4. API Usage
For complete API documentation, please refer to the [official API usage guide](https://github.com/sgl-project/sglang/blob/main/python/sglang/multimodal_gen/docs/openai_api.md).
### 4.1 Generate an Image
Generate an image with the OpenAI-compatible images API:
```python Example
import base64
from openai import OpenAI
client = OpenAI(api_key="EMPTY", base_url="http://localhost:30000/v1")
response = client.images.generate(
model="krea/Krea-2-Turbo",
prompt="a red fox sitting in fresh snow, golden hour, photorealistic",
n=1,
response_format="b64_json",
)
# Save the generated image
image_bytes = base64.b64decode(response.data[0].b64_json)
with open("output.png", "wb") as f:
f.write(image_bytes)
```
You can also generate a single image from the command line:
```bash Command
sglang generate --model-path krea/Krea-2-Turbo \
--prompt "a red fox sitting in fresh snow, golden hour, photorealistic" \
--num-inference-steps 8 --height 1024 --width 1024 --save-output
```
### 4.2 Advanced Usage
#### 4.2.1 CPU Offload
Krea-2's DiT is ~24 GB in bf16 (the bulk of the model). On memory-constrained GPUs you can keep less of it resident:
- `--dit-layerwise-offload`: stream the DiT's transformer blocks layer-by-layer with async host-to-device prefetch overlap, so only a small working set stays on the GPU. This is the primary way to fit Krea-2 on a single consumer / 32 GB-class card, at a modest latency cost. Tune the memory/latency trade-off with `--dit-offload-prefetch-size` (`0.0` prefetches one layer for the lowest memory; larger values prefetch more layers -- faster but more memory).
- `--dit-cpu-offload`: keep the whole DiT in host memory. Combine it with `--dit-layerwise-offload` for the lowest peak GPU memory (weights stay on host and only the layers needed for the current step are brought on-device).
- `--text-encoder-cpu-offload`: offload the Qwen3-VL text encoder (it is idle during the denoise loop).
- `--vae-cpu-offload`: offload the VAE.
- `--pin-cpu-memory`: pin host memory for offload. Add only as a temporary workaround if you hit `CUDA error: invalid argument`.
On large-VRAM GPUs (e.g. H200), keep everything resident (offloads off) for the fastest latency.
#### 4.2.2 Style LoRAs
Krea publishes diffusers-format style LoRAs for Krea-2-Turbo. Load one with `--lora-path` (a HuggingFace repo id or local path) and `--lora-weight-name`, set its strength with `--lora-scale` (recommended 0.8-1.0), and include the LoRA's **trigger phrase** in the prompt.
| LoRA | Trigger phrase | Scale |
| --- | --- | --- |
| `krea/Krea-2-LoRA-coolblue` | `Teal watercolor illustration style` | 0.8 |
| `krea/Krea-2-LoRA-warmpastel` | `Muted minimalist sketch style` | 0.8 |
| `krea/Krea-2-LoRA-plasmoid` | `Ethereal shimmering light style` | 0.8 |
| `krea/Krea-2-LoRA-darkbrush` | `monochrome ink wash style` | 1.0 |
```bash Command
sglang generate --model-path krea/Krea-2-Turbo \
--lora-path krea/Krea-2-LoRA-coolblue --lora-weight-name coolblue.safetensors --lora-scale 0.8 \
--prompt "A deer grazing in the forest, Teal watercolor illustration style" \
--num-inference-steps 8 --guidance-scale 0.0 --height 1024 --width 1024 --save-output
```
The model-card previews are rendered on Turbo at 8 steps with `guidance_scale 0.0`. The same `--lora-path`/`--lora-weight-name`/`--lora-scale` flags also apply to `sglang serve`.
> **Note:** LoRA requires the DiT to be GPU-resident or streamed with `--dit-layerwise-offload`; it is not compatible with full `--dit-cpu-offload` (the offloaded base weights are not streamed for LoRA-wrapped layers). On memory-constrained GPUs use `--dit-layerwise-offload true`.
## 5. Benchmark
Test Environment:
- Hardware: NVIDIA H200 GPU (1x)
- Model: krea/Krea-2-Turbo (8 inference steps)
- sglang diffusion version: 0.5.13
**Server Command** (used for both benchmarks below):
```shell Command
sglang serve --model-path krea/Krea-2-Turbo --port 30000
```
### 5.1 Generate an image
**Benchmark Command**:
```shell Command
python3 -m sglang.multimodal_gen.benchmarks.bench_serving \
--model krea/Krea-2-Turbo --dataset vbench --task text-to-image \
--num-prompts 1 --max-concurrency 1
```
**Result**:
```text Output
================= Serving Benchmark Result =================
Task: text-to-image
Model: krea/Krea-2-Turbo
Dataset: vbench
--------------------------------------------------
Benchmark duration (s): 1.56
Request rate: inf
Max request concurrency: 1
Successful requests: 1/1
--------------------------------------------------
Request throughput (req/s): 0.64
Latency Mean (s): 1.5600
Latency Median (s): 1.5600
Latency P99 (s): 1.5600
--------------------------------------------------
Peak Memory Max (MB): 37466.00
Peak Memory Mean (MB): 37466.00
Peak Memory Median (MB): 37466.00
============================================================
```
### 5.2 Generate images with high concurrency
**Benchmark Command**:
```shell Command
python3 -m sglang.multimodal_gen.benchmarks.bench_serving \
--model krea/Krea-2-Turbo --dataset vbench --task text-to-image \
--num-prompts 20 --max-concurrency 20
```
**Result**:
```text Output
================= Serving Benchmark Result =================
Task: text-to-image
Model: krea/Krea-2-Turbo
Dataset: vbench
--------------------------------------------------
Benchmark duration (s): 31.47
Request rate: inf
Max request concurrency: 20
Successful requests: 20/20
--------------------------------------------------
Request throughput (req/s): 0.64
Latency Mean (s): 16.5000
Latency Median (s): 16.5200
Latency P99 (s): 31.1300
--------------------------------------------------
Peak Memory Max (MB): 37468.00
Peak Memory Mean (MB): 37466.40
Peak Memory Median (MB): 37466.00
============================================================
```
docs_new/cookbook/diffusion/intro.mdx
+6
View File
@@ -53,6 +53,12 @@ Offline models generate each image or video request as a bounded denoising job.
href="/cookbook/diffusion/Z-Image/Z-Image-Turbo"
img="/cards/logos/zimage.png"
/>
<Card
title="Krea"
mode="card"
href="/cookbook/diffusion/Krea/Krea-2"
img="/cards/logos/krea.png"
/>
<Card
title="ERNIE-Image"
mode="card"