MOSS icon indicating copy to clipboard operation
MOSS copied to clipboard

出现问题 Setting `pad_token_id` to `eos_token_id`:106068 for open-end generation.

Open lin-cop opened this issue 1 year ago • 2 comments

========实例配置======== 核数:16 内存:127 GB 磁盘:49% 9.7G/20G 显卡:NVIDIA A100-SXM4-80GB, 1 输出内容: The attention mask and the pad token id were not set. As a consequence, you may observe unexpected behavior. Please pass your input's attention_mask to obtain reliable results. Setting pad_token_id to eos_token_id:106068 for open-end generation.

lin-cop avatar Apr 27 '23 05:04 lin-cop

95d6abbb9ebe244a0ac50efab4e42fe 使用模型:moss-moon-003-sft, 命令启动:streamlit run moss_web_demo_streamlit.py --server.port 23421 输入命令后能正常启动加载模型,还有我的模型在本地,没有使用线上的模型

改为 使用示例 后还是出现上述问题

单卡部署(适用于A100/A800) 以下是一个简单的调用moss-moon-003-sft生成对话的示例代码,可在单张A100/A800或CPU运行,使用FP16精度时约占用30GB显存:

    from transformers import AutoTokenizer, AutoModelForCausalLM
    tokenizer = AutoTokenizer.from_pretrained("fnlp/moss-moon-003-sft", trust_remote_code=True)
    model = AutoModelForCausalLM.from_pretrained("fnlp/moss-moon-003-sft", trust_remote_code=True).half().cuda()
   model = model.eval()

lin-cop avatar Apr 27 '23 05:04 lin-cop

我是用多卡推理时也出现了这个问题

llx-08 avatar Jun 13 '23 01:06 llx-08