Different Results from the demo

#30
by Ryanchynoweth - opened

I am working to deploy this model and have the following code:

tokenizer = AutoTokenizer.from_pretrained('stepfun-ai/GOT-OCR2_0', trust_remote_code=True, padding='max_length')
model = AutoModel.from_pretrained('stepfun-ai/GOT-OCR2_0', trust_remote_code=True, low_cpu_mem_usage=True, use_safetensors=True, pad_token_id=tokenizer.eos_token_id)

model = model.eval().cuda()
res = model.chat(tokenizer, image_file, ocr_type='ocr')

But the results are widely different from the demo website. What could be causing this difference?

Sign up or log in to comment