-
Hi,我按照modelscore教程,处理process.json数据,然后下载knowlm-13b-ie模型直接推理
CUDA_VISIBLE_DEVICES="5" **python src/inference.py** --model_name_or_path 'model/knowlm-13b-ie' --model_name 'llama' --input_file 'data/NER…
-
你好,cama参考链接(https://github.com/zjunlp/KnowLM)里除了全量的预训练权重(ZhiXi-13B-Diff),还有LoRA指令微调权重(ZhiXi-13B-LoRA),请问如果想使用QiZhen-CaMA-13B-Checkpoint-12400,是基于全量的预训练权重(ZhiXi-13B-Diff)合并得到的完整cama权重来做,还是也要用到cama的LoRA…
-
你好,非常感谢你们的工作。我在尝试knowlm后想要继续微调一下Qwen模型,使用的命令是
`CUDA_VISIBLE_DEVICES=0 accelerate launch finetune.py --data_path KnowLM-main/finetune/lora/knowlm/data/alpaca_data_cleaned.json --base_model Qwen1.5-7B-…
zhl98 updated
1 month ago
-
你好,非常感谢你们的工作。
我在使用lora微调你们的KnowLM-13B-Base模型,我使用的运行命令是
`CUDA_VISIBLE_DEVICES=0,1 accelerate launch finetune.py --data_path ./data/alpaca_data_cleaned.json --base_model /huggingface/knowlm-13b-v1.0`…
zhl98 updated
1 month ago
-
File "/home/keke/KnowLM/finetune/lora/knowlm/finetune.py", line 299, in
fire.Fire(train)
File "/home/keke/anaconda3/envs/knowlm/lib/python3.9/site-packages/fire/core.py", line 143, in Fire
…
-
使用命令:
`CUDA_VISIBLE_DEVICES=0 accelerate launch finetune.py --data_path ./data/introduction2_train.json --base_model /home/keke/KnowLM/knowlm-13b-zhixi --resume_from_checkpoint /home/keke/KnowLM/fine…
-
-
报错ImportError: cannot import name 'is_npu_available' from 'accelerate.utils'
求解决
-
如题,求指教!
-
操作系统环境:openEuler 22.03
显卡:两张各25GB的RTX A5000
环境变量设置:
export BNB_CUDA_VERSION=122
export PATH=/usr/local/cuda-12.2/bin:$PATH
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-12.2/lib64
依赖包版…