dage0127

Aquila-chat ：bash dist_trigger_docker.sh 调优 aquila-7b，报错：ImportError: numpy.core.multiarray failed to import

1

### System Info flagai 1.7.3 torch 1.13.1 numpy 1.22.4 numpy：1.15,1.19,1.22 都报同样错误 操作系统：ubuntu 18.04 Conda create python 3.8.16环境 错误信息： (gpt) llm@ht-PowerEdge-R740:/home/haitaiwork/fly/FlagAI/examples/Aquila/Aquila-chat$ tail /home/haitaiwork/fly/FlagAI/examples/Aquila/Aquila-chat/checkpoints_out/aquila_sft/2023062108/Aquila-chat.yaml.log.txt awk: line 1: runaway string constant "172.28.0.1 ......

ECC key format is not supported

1

运行报错，“ValueError: ECC key format is not supported”。 安装了pycryptodome的3.10.1,3.10.4,3.17版本，都报相同的错误。 (base) deepl@ht-PowerEdge-R740:~/explore/python-psi$ PSI_CONFIG=./config/server.config.yaml psi_run server 127.0.0.1:1234 2023-05-31 13:15:13.443 - INFO: start prepare Traceback (most recent call last): File "/home/haitaiwork/deepl/anaconda3/bin/psi_run", line 8, in...

Lora认知微调，微调成功，直接进行推理也正确。但merge之后，再推理，生成乱码，并报错“TypeError: argument 'tokens': 'NoneType' object cannot be converted to 'PyString'”

2

一、现象 1.使用lora 认知训练，成功。命令为： python llm_sft.py --model_type qwen1half-0_5b-chat --model_id_or_path /home/work/pbg/Qwen1.5-0.5B-Chat \ --dtype fp16 --sft_type lora --tuner_backend peft --output_dir output \ --dataset blossom-math-zh --train_dataset_sample 1000 --num_train_epochs 2 --max_length 512 \ --check_dataset_strategy warning...

一、推理错误 推理命令： swift infer --dtype fp16 --ckpt_dir '/home/haitaiwork/pbg/swift-main/examples/pytorch/llm/output/qwen1half-0_5b-chat/v9-20240509-172543/checkpoint-340' 或者使用： swift infer --ckpt_dir '/home/haitaiwork/pbg/swift-main/examples/pytorch/llm/output/qwen1half-0_5b-chat/v9-20240509-172543/checkpoint-340' 都出现乱码，并报错： [INFO:swift] generation_config: SamplingParams(n=1, best_of=1, presence_penalty=0.0, frequency_penalty=0.0, repetition_penalty=1.0, temperature=0.3, top_p=0.7, top_k=20, min_p=0.0, use_beam_search=False, length_penalty=1.0, early_stopping=False, stop=[], stop_token_ids=[],...