Initial model weights

README.md

@@ -11,9 +11,9 @@ widget:
 
 # Simple Diffusion XS (sdxs-2b alpha version)
 
-*XS Size, Excess Quality (not for now)* 
+*XS Size, Excess Quality* 
 
-Train status: 1xRTX4090 / we need more gold / [support us please..](https://huggingface.co/AiArtLab/sdxs-1b#donations)
+Train status: 4xRTX4090 / we need more gold / [support us please..](https://huggingface.co/AiArtLab/sdxs-2b#donations)
 
 <img src="media/refined.jpg" height="256"/>
 
@@ -23,13 +23,16 @@ At AiArtLab, we strive to create a free, compact and fast model that can be trai
 - Qwen3.5:             0.8b parameters
 - Qwen vae:            16ch8x
 - Resolution:          Default 768x1152, trained from 768 to 1152 with step 64
-- Limitations:         trained on small datasets ~2kk, focused on art / illustrations / anime (with ~10% photo / use "photo" in negative prompt)
+- Limitations:         trained on small datasets ~2.5kk, focused on art / illustrations / anime and photo
 - Captions:            danbooru, natural (short/medium), trained with 250 max toks
 
 
 ### Key points
 
- - 26 Apr 2026: Started research on Cosmos Transformers https://wandb.ai/recoilme/transformer/runs/7kcr6fsa
+<img src="media/start.png" height="256"/>
+
+ - Apr 2026: Started research of Cosmos Transformer
+ - May 2026: 1st preview (Initial model weights)
 
 
 ### Samples with seed 0 and good prompts
@@ -55,39 +58,23 @@ pipe = DiffusionPipeline.from_pretrained(
 ).to(device)
 
 refined = "A blonde-haired Red Eyes girl with a hair ribbon, half-updo, and tsurime stands solo in a flower field holding a bouquet with a serene smile, wearing green overalls, a white shirt, rolled-up sleeves, and a straw hat with a flower while looking at the viewer under volumetric and natural lighting with a Dutch angle."
-print(refined)
+
+negative_prompt = "worst quality, low quality, loli, low details, blurry, jpeg artifacts, unfinished, sketch, sepia, missing limb, text, bad anatomy, bad proportions, bad hands, missing fingers"
+
 output = pipe(
     prompt=refined,
     negative_prompt=negative_prompt,
-    seed=43,
 )
 
 image = output.images[0]
-
 image.show()
 ```
 
 ### Prompt refiner
-```
-refined = pipe.refine_prompts("1girl, solo, lazypos, cinematic lighting, volumetric lighting, dutch angle,
-anime coloring, delicate, dreamy, natural shadow, serene atmosphere,
-<lora:hololive_kaela_kovalskia_redebut:0.65> kaela20, blonde hair, red eyes, hair ribbon, red ribbon, long hair, half updo, tsurime,
-BREAK
-overalls, green overalls, white shirt, straw hat, hat flower, daisy, sleeves rolled up, three-quarter sleeves,
-outdoors, flower field,
-bouquet, holding bouquet, looking at viewer, smile,
-m¨h0r_a97Q, refraction, prismatic light,
-<lora:WSSKX_WAI:0.35>
-<lora:ashima-il:0.35>
-<lora:MYRHA-¨00:0.75>
-<lora:Dramatic Lighting Slider:3.5>
-<lora:spo_sdxl_10ep_4k-data_lora_webui:1.3>")
+```python
+refined = pipe.refine_prompts("1girl")
 print(refined[0])
 ```
-Refined prompt:
-
-A blonde-haired Red Eyes girl with a hair ribbon, half-updo, and tsurime stands solo in a flower field holding a bouquet with a serene smile, wearing green overalls, a white shirt, rolled-up sleeves, and a straw hat with a flower while looking at the viewer under volumetric and natural lighting with a Dutch angle.
-![refine](media/refined.webp)
 
 ## Donations
 Donated on train: 0$ Thanks for your support!