GPT-Image-Edit-T5-only: Fine-Tuned on GPT-Image-Edit-1.5M
Original Text encoders: T5 & CLIP text encoder (checkpoint bundle still includes Qwen-VL weights from codebase UniWorld-V1, but inference uses only T5 + CLIP text encoder)
Downloads last month
7
Safetensors
Model size
20.4B params
Tensor type
BF16
·
Collection including
UCSC-VLAA/gpt-image-edit-finetune-t5-only
