zhaoyang0618

Results 4 comments of zhaoyang0618

[@HwwwwwwwH](https://github.com/HwwwwwwwH) 我也出现了在v0.7.1上运行vllm 跑不起来的问题,下面是我的环境和运行命令,提示AttributeError: 'MiniCPMOProcessor' object has no attribute 'get_audio_placeholder'的问题。 更新一下最新的HF仓库代码应该就可以了 RuntimeError: Failed to apply MiniCPMOProcessor on data={'text': '(./)', 'images': []} with kwargs={}, 我这边更新后会报这个错误 同样的错误!是否有解决方案?

[rank0]: Traceback (most recent call last): [rank0]: File "/opt/anaconda3/envs/vllm/lib/python3.12/site-packages/vllm/inputs/registry.py", line 160, in call_hf_processor [rank0]: return hf_processor(**data, **merged_kwargs, return_tensors="pt") [rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ [rank0]: File "/root/.cache/huggingface/modules/transformers_modules/processing_minicpmo.py", line 77, in __call__ [rank0]: image_inputs =...

我把transformers修改成4.48.2版本之后,一开始会报错,AttributeError: 'MiniCPMOProcessor' object has no attribute 'get_audio_placeholder',搜索之后,从HF下载文件(processing_minicpmo.py和modeling_minicpmo.py)进行更新,成功! 说明modelscope上的模型不是最新的!最新的最好还是在HF下载! 多谢!

minicpm-v-2_6模型一开始也有类似问题,更换transformers版本之后,一开始也是报错: The model's max seq len (4096) is larger than the maximum number of tokens that can be stored in KV cache (1856). Try increasing `gpu_memory_utilization` or decreasing `max_model_len` when...