Thank you for your awesome work! When I using the MLLM with UniTok, I met some problems:
When run:
with torch.inference_mode():
output_ids = model.generate(
input_ids,
image=image_tensor.unsqueeze(0).half().cuda(),
image_sizes=[image.size],
do_sample=True if args.temperature > 0 else False,
temperature=args.temperature,
top_p=args.top_p,
num_beams=args.num_beams,
# no_repeat_ngram_size=3,
max_new_tokens=1024,
use_cache=True)
Got:
ValueError: The following model_kwargs are not used by the model: ['images', 'image_sizes'] (note: typos in the generate arguments will also show up in this list)
Could you give me some idea?Thank!