Asuka-FM: 200M Dual-Stream SPRINT DiT (Flow Matching)

An ultra-efficient, 200M-parameter Dual-Stream Diffusion Transformer (DiT) trained from scratch on consumer hardware (1x NVIDIA RTX 4090 in ~50 hours) using Rectified Flow / Flow Matching, SPRINT token-dropping, and Continuous Aspect-Normalized RoPE.

Despite its compact size, the model generates high-fidelity anime art with accurate character reproduction, prompt alignment, and zero-shot camera control (panning and zoom) at inference time.

Developed by: aipracticecafe
Repository: deeplearningcafe/asuka-fm
Dataset: curated-danbooru-2026 (~337K samples)
Architecture: Dual-Stream DiT (SPRINT-aligned, SwiGLU, Sandwich RMSNorm)
Parameters: ~200M (Backbone) + 230M (Text Encoder)
Objective: Conditional Flow Matching (Logit-Normal timestep sampling)
License: OpenRAIL++

Prompt

1girl, souryuu asuka langley, neon genesis evangelion, rebuild of evangelion, evangelion: 3.0 you can $not$ redo, momoko $momopoco$, toosaka asagi, ebifurya, sensitive, masterpiece, good score, absurdres, highres, best quality, amazing quality, great quality, solo, long hair, blue eyes, orange hair, breasts, parted lips, official alternate costume, pilot suit, red bodysuit, black eye patch, animal hat, cabbie hat, cat hat, black hat, gloves, hat, hat horns, jacket, mecha pilot suit, bodysuit under clothes, red gloves, red jacket, track jacket, turtleneck, turtleneck jacket, interface headset, multicolored clothes, interface headset $evangelion$, upper body, simple background

Prompt

1girl, hoshino ai, oshi no ko, ebifurya, sensitive, masterpiece, good score, absurdres, highres, best quality, amazing quality, great quality, dress, gloves, hair ornament, idol, idol clothes, long hair, looking at viewer, microphone, one eye closed, open mouth, pink dress, pink gloves, pointing, purple eyes, purple hair, solo, star-shaped pupils, star, star in eye, symbol-shaped pupils, symbol in eye, anime coloring, black shirt, brooch, frilled gloves, frilled shirt, frilled sleeves, frills, hair between eyes, heart, heart brooch, jewelry, messy hair, one side up, pink shirt, rabbit hair ornament, shiny clothes, shiny skin, shirt, short sleeves, solo, star $symbol$, turtleneck, wing brooch, wings, yellow shirt, upper body, simple background

Prompt

2girls, claire francois, rae taylor, watashi no oshi wa akuyaku reijou, sensitive, masterpiece, good score, absurdres, highres, best quality, amazing quality, great quality, :d, ^ ^, blonde hair, blue eyes, blush, white ascot, bow, brown hair, closed eyes, collared shirt, cropped jacket, drill hair, drill sidelocks, floating hair, frills, gem, hair bow, half updo, holding another's arm, jacket, long hair, long sleeves, multiple girls, red bow, red jacket, school uniform, shirt, short hair, smile, yuri, brown eyes, multiple girls, white shirt, ascot, brooch, frilled jacket, jewelry, medium breasts, very long hair, upper body

Key Architecture & Features

1. SPRINT Dual-Stream Transformer Backbone

The base architecture is based on the results from (i1: A Simple and Fully Open Recipe for Strong Text-to-Image Models)[https://arxiv.org/abs/2606.11289]

Minimalist Conditioning: Uses prepended time tokens rather than heavy per-block AdaLN modulation, reducing parameter overhead.
Sandwich Normalization & SwiGLU: RMSNorm placed before and after attention and FFN operations ensures training stability in bfloat16.
SPRINT Acceleration: Employs structured group-wise 2x2 patch dropping (75% drop ratio during early pretraining), cutting compute demands by half while retaining global semantic context.

2. HDM Continuous Aspect-Normalized RoPE (Camera Control)

Based on the Patch Diffusion: Faster and More Data-Efficient implementation proposed in the HDM paper. Early experiments with discrete integer position IDs caused severe multi-head replication when deviating from trained aspect ratios. We resolve this by adopting continuous aspect-normalized coordinate mapping which was the original idea proposed in HDM:

$r_H = \sqrt{\frac{H}{W}}, \quad r_W = \sqrt{\frac{W}{H}}$

$y_{\text{pos}} = \left(\frac{y_{\text{center}}}{H}\right) \cdot (2 r_H) - r_H, \quad x_{\text{pos}} = \left(\frac{x_{\text{center}}}{W}\right) \cdot (2 r_W) - r_W$

Manipulating $(x_{\text{pos}}, y_{\text{pos}})$ directly controls the visual framing at inference time:

X-Shift / Y-Shift: Translates the virtual camera viewport horizontally or vertically.
Zoom: Scales spatial coordinates around the focal center.

3. Components

Text Encoder: LiquidAI/LFM2.5-230M (unidirectional causal LM hidden states passed via a 2-layer Transformer Text Adapter).
Autoencoder: kaiyuyue/FLUX.2-dev-vae (C=32, f=8, patch size p=2, effective image token size 16x16).

Training Recipe (1x RTX 4090, ~50h)

The model was pretrained on 337,000 curated Danbooru samples at 256px base resolution with shifted square cropping and cached in-RAM latents.

Phase	Epochs	Token Drop	RoPE Format	Description
Phase 1	100 ep	0.75 (SPRINT)	Discrete	Fast semantic convergence
Phase 2	20 ep	0.75 (SPRINT)	Continuous	RoPE coordinates alignment
Phase 3	25 ep	0.00 (Dense)	Continuous	High-frequency detail recovery
Phase 4	25 ep	0.00 (Dense)	Continuous	Cosine LR decay to minimum LR

Total Epochs: 175
Optimizer: 8-bit AdamW (bitsandbytes), $\beta=(0.9, 0.95)$, WD=0.01
Learning Rate: 5e-4 with warmup and cosine annealing only for last stage
Effective Batch Size: 256 (Batch size 64x4 gradient accumulation)
Precision: Mixed Precision (torch.bfloat16 with torch.compile)

Inference and WebUI

The training repository has a simple Gradio UI to perform sampling with camera controls. ui_app.py

Recommended Settings

Sampling Algorithm: Forward Euler ODE Solver (Rectified Flow)
Inference Steps: 25 – 35
CFG Scale: 4.5 – 6.0
Time Shift: 1.0 – 3.0 (higher values enhance contrast/saturation)

Tag Formatting Template

1girl/1boy, character name, series name, artists, rating, quality tags, actions/clothing

Negative prompt:

sketch, monochrome, english text, realistic, traditional media, very displeasing, displeasing, bad score, worse score, bad quality,worst quality, worst detail, watermark, signature, blurry, bad hands, bad anatomy, bad fingers, extra fingers, extra limbs, deformed limbs, comic, speech bubble, lowres

Python Sampling Script

import torch
from omegaconf import OmegaConf
from src.models.factory import load_trainable_model
from src.diffusion.schedules import LinearSchedule
from src.diffusion.sampling import generate_samples

cfg = OmegaConf.load("config.yaml")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
dtype = torch.bfloat16

unet, text_encoder, vae, tokenizer, _ = load_trainable_model(
    models_path=cfg.paths.models,
    device=device,
    dtype=dtype,
    resume_from_checkpoint="path_to_checkpoint",
    model_type="sprint_dual",
    model_cfg=cfg.models,
    autocast_dtype=dtype,
)

schedule = LinearSchedule(device=device)

sample_configs = [
    {
        "prompt": "1girl, souryuu asuka langley, neon genesis evangelion, "
                  "masterpiece, absurdres",
        "negative_prompt": "very displeasing, displeasing, bad score",
        "height": 256,
        "width": 384,
        "sample_steps": 30,
        "cfg_scale": 6.0,
        "shift": 1.0,
        "zoom": 1.0,
        "x_shift": 0.0,
        "y_shift": 0.0,
        "seed": 42,
    }
]

images = generate_samples(
    unet=unet,
    text_encoder=text_encoder,
    tokenizer=tokenizer,
    vae=vae,
    schedule=schedule,
    sample_configs=sample_configs,
    global_batch_size=1,
    diffusion_type="flow_matching",
    device=device,
    dtype=dtype,
    autocast_dtype=dtype,
    use_unet_mult=False,
    vae_mean=cfg.models.vae_mean,
    vae_std=cfg.models.vae_std,
    in_channels=cfg.models.in_channels,
    coord_system="aspect_norm",
)

images[0].save("output.png")

Content Advisory & Limitations

Resolution: Trained on 256px shifted crops. While continuous RoPE allows sampling at 384px and 512px, high resolutions may exhibit slight softness until a dedicated 512px/1024px fine-tuning pass is completed.
Complex Hands: Small 200M parameter count prioritizes overall composition and character concepts over fine micro-structures like fingernails.
NSFW & Sensitive Content: The model was trained on a dataset that contains uncensored anime-style art, including sensitive, questionable, and explicit material (g, s, q, e ratings). Use appropriate tag filtering during training if a SFW-only model is required.
Legal Disclaimer: Images are sourced from Danbooru for research, educational, and generative modeling purposes under fair use. Copyright of all underlying illustrations belongs to their respective creators.

References

i1: i1: A Simple and Fully Open Recipe for Strong Text-to-Image Models
Patch Diffusion: Patch Diffusion: Faster and More Data-Efficient
Home-made Diffusion Model from Scratch to Hatch: HDM
Danbooru Dataset Toolkit: Dataset processed and formatted using the Danbooru Dataset Toolkit.

License

This model is licensed under the CreativeML Open RAIL++-M License. You are free to use, modify, and distribute the model weights provided you include the license text and respect the behavioral use restrictions.

Downloads last month: -

Datasets used to train aipracticecafe/dual-stream-dit

Papers for aipracticecafe/dual-stream-dit