I am in the process of reproducing the evaluation of the seed-llama-8b-sft model on SeedBench-Image and COCO-test (text2image). To ensure accuracy and fidelity in my reproduction, I would be grateful if you could provide the following information:
For text-to-image evaluation: the Diffusion guidance, inference steps, and the LLM CFG guidance used when generating image tokens.
For SeedBench evaluation: the prompt template used when evaluating seed-llama 8b-instruct.
I‘m assured that this information will be used strictly for academic purposes.
I am in the process of reproducing the evaluation of the seed-llama-8b-sft model on SeedBench-Image and COCO-test (text2image). To ensure accuracy and fidelity in my reproduction, I would be grateful if you could provide the following information:
I‘m assured that this information will be used strictly for academic purposes.