sugarquark/sd15-text-encoder-t5g-2b-ul2-it

Inference

from transformers.models import AutoTokenizer, T5GemmaEncoderModel
import torch

if __name__ == '__main__':
    model = T5GemmaEncoderModel.from_pretrained(t5gemma_path, torch_dtype=torch.bfloat16)
    tokenizer = AutoTokenizer.from_pretrained(t5gemma_path)
    inputs = tokenizer('Gemma', max_length=512, padding='max_length', truncation=True, return_tensors='pt')
    output = model.forward(**inputs).last_hidden_state

SD1.5 and Gemma

from diffusers import StableDiffusionPipeline
from gemma_encoder import Encoder

if __name__ == '__main__':
    pipeline = StableDiffusionPipeline.from_pretrained('NovelAI/nai-anime-v2')
    pipeline.enable_model_cpu_offload()
    encoder = Encoder(adapter_model, t5gemma_path, device='cpu')
    load_model(adapter_model, 'adapter.safetensors')
    image = pipeline(
        None,
        prompt_embeds=encoder.encode(pipeline, text).to('cpu'),
        negative_prompt='bad quality, low quality, worst quality'
    ).images[0]
    image.save('preview.png')

Datasets

• alfredplpl/artbench-pd-256x256
• danbooru2023-florence2-caption (verb, action clauses)
• spatial-caption
• SPRIGHT-T2I/spright_coco
• sugarquark/colormix (synthetic color, fashion dataset)
• trojblue/danbooru2025-metadata