LlavaNextProcessor.__init__() got an unexpected keyword argument 'image_token'

[M-Fannilla]

System Info

transformers 4.41.2
Python 3.10.12
Nvidia A40 / Runpod.io

Whe working with LlavaNextProcessor I get errors that were not there.

Error:
LlavaNextProcessor.init() got an unexpected keyword argument 'image_token'

Which previously did not happen (~week ago).

Who can help?

No response

Information

• The official example scripts
• My own modified scripts

Reproduction

model_name = "llava-hf/llava-v1.6-vicuna-7b-hf"
processor = LlavaNextProcessor.from_pretrained(
model_name, padding_side='left'
)
model = LlavaNextForConditionalGeneration.from_pretrained(
model_name, torch_dtype=torch.float16, low_cpu_mem_usage=True
).to("cuda").eval()

Expected behavior

I would expect to initiate the objects properly

[zucchini-nlp]

@M-Fannilla hi, I just updates llava-weights in the hub which caused the error. I will revert the changes soon

[M-Fannilla]

@zucchini-nlp Great, Thanks!

[zucchini-nlp]

@M-Fannilla should be working now, closing the issue as resolved!

[M-Fannilla]

There is new issue:

Failed to import transformers.models.llama.modeling_llama because of the following error (look up to see its traceback):
/usr/local/lib/python3.10/dist-packages/flash_attn_2_cuda.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZN2at4_ops5zeros4callEN3c108ArrayRefINS2_6SymIntEEENS2_8optionalINS2_10ScalarTypeEEENS6_INS2_6LayoutEEENS6_INS2_6DeviceEEENS6_IbEE

I did not have this one before.

[zucchini-nlp]

@M-Fannilla this one isn't related to the model checkpoint, but rather to the installed packages you have. Seems like flash-attn wasn't installed properly or has some dependency issues. Try to uninstall flash-attn and load again

If it doesn't work feel free to open a new issue :)