open-compass
version of transfomers
which version of transformers should i apply when employing qwen2vl_series and mplug_owl3 respectively. i encounter the following problems: for qwen2vl：
ImportError: cannot import name 'Qwen2VLForConditionalGeneration' from 'transformers'
for mplug_owl3:
File "/data1/VLMEvalKit/vlmeval/vlm/mplug_owl3.py", line 146, in __init__ self.tokenizer = AutoTokenizer.from_pretrained( File "/usr/local/lib/python3.10/dist-packages/transformers/models/auto/tokenization_auto.py", line 733, in from_pretrained raise ValueError( ValueError: Tokenizer class Qwen2Tokenizer does not exist or is not currently imported.
Hi @The-Git-Rain. For Qwen2VL, we recommend that you use transformers>=4.48.0 For mPLUG-Owl3, it seems to be the same problem as Qwen2vl. You can also try to upgrade the transformers
For Qwen2VL, I use the latest transformers version and solve the problem
For mPLUG-Owls, i use the latest version but I still have the problem below:
File "/data1/VLMEvalKit/vlmeval/vlm/mplug_owl3.py", line 320, in generate_inner if listinstr(['MVBench'], dataset): File "/data1/VLMEvalKit/vlmeval/smp/misc.py", line 121, in listinstr if item in s: TypeError: argument of type 'NoneType' is not iterable
it seems the configuration is not compatible to mPLUG-Owl3