[ST][MS][2.3][mindformers][readme][llama2][大模型] 文档描述不清晰，启动方式需改成msrun启动，单卡推理失败

| name       | about	                             | labels   |
| ---------- | ------------------------------------- | -------- |
| Bug Report | Use this template for reporting a bug | kind/bug |

### Describe the current behavior / 问题描述 (Mandatory / 必填)
文档描述不清晰，启动方式需改成msrun启动，单卡推理失败
 
 
### Environment / 环境信息 (Mandatory / 必填)
- **Hardware Environment(`Ascend`/`GPU`/`CPU`)  / 硬件环境**: 
> Please delete the backend not involved / 请删除不涉及的后端:
> /device ascend/GPU/CPU/kirin/等其他芯片

- **Software Environment / 软件环境 (Mandatory / 必填)**:
-- MindSpore version (e.g., 1.7.0.Bxxx) :
-- Python version (e.g., Python 3.7.5) :
-- OS platform and distribution (e.g., Linux Ubuntu 16.04):
-- GCC/Compiler version (if compiled from source): 
ms版本：__commit_id__ = '[sha1]:39ac2284,[branch]:(HEAD,origin/master,origin/HEAD,master)'
mf版本：
dev
commit 9d8c186
Merge: a709b59 d836b61
Author: i-robot <huawei_ci_bot@163.com>
Date:   Tue Apr 30 09:12:29 2024 +0000

!2892 【dev】Baichuan2-13B动态shape推理精度问题修复
    Merge pull request !2892 from renyujin/baichuan2_13b_dynamic_fix

- **Excute Mode / 执行模式 (Mandatory / 必填)(`PyNative`/`Graph`)**: 
> Please delete the mode not involved / 请删除不涉及的模式:
> /mode pynative
> /mode graph

### Related testcase / 关联用例 (Mandatory / 必填)
不涉及
 
### Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)
不涉及

### Describe the expected behavior / 预期结果 (Mandatory / 必填)
文档描述清晰，用户看了知道怎么操作
多卡启动方式需换成msrun启动
推理正常
 
### Related log / screenshot / 日志 / 截图 (Mandatory / 必填)
# 问题1：推理-基本介绍里，没说明配置文件是哪个
推理时将配置文件中param_init_type修改为float32；若为Atlas 800T A2推理，则加入ascend_config配置。

# 问题2：多卡运行自定义多batch推理的脚本需换成msrun启动
\# >>> `run_predict.sh`文件
export RANK_TABLE_FILE=$1
CHECKPOINT_PATH=$2
YAML_FILE=$3
MODEL_TYPE=$4
\# define variable
export RANK_SIZE=$5
export START_RANK=$6 # this server start rank
let END_RANK=START_RANK+RANK_SIZE # this server end rank

\# run
for((i=${START_RANK}; i<${END_RANK}; i++))
do
    export RANK_ID=$((i-START_RANK))
    export DEVICE_ID=$i
    echo "Start distribute running for rank $RANK_ID, device $DEVICE_ID"
    python3 ./predict_custom.py --checkpoint_path $CHECKPOINT_PATH --yaml_file ${YAML_FILE} --model_type ${MODEL_TYPE} &> mindformers_$RANK_ID.log &
done

# 问题3：单卡推理失败，ckpt下载失败
2024-05-06 15:39:38,050 - mindformers[mindformers/tools/download_tools.py:72] - ERROR - Connect error, please download https://ascend-repo-modelzoo.obs.cn-east-2.myhuaweicloud.com/MindFormers/llama2/tokenizer.model to ./checkpoint_download/llama2/tokenizer.model.
Traceback (most recent call last):
  File "predict_custom.py", line 82, in <module>
    main(args)
  File "predict_custom.py", line 39, in main
    tokenizer = AutoTokenizer.from_pretrained(args.model_type)
  File "/home/miniconda3/envs/ci/lib/python3.7/site-packages/mindformers/models/auto/tokenization_auto.py", line 347, in from_pretrained
    instanced_class = cls.get_class_from_origin_mode(yaml_name_or_path, **kwargs)
  File "/home/miniconda3/envs/ci/lib/python3.7/site-packages/mindformers/models/auto/tokenization_auto.py", line 412, in get_class_from_origin_mode
    instanced_class = dynamic_class.from_pretrained(yaml_name_or_path, **kwargs)
  File "/home/miniconda3/envs/ci/lib/python3.7/site-packages/mindformers/models/tokenization_utils_base.py", line 1836, in from_pretrained
    tokenizer = cls.from_origin_pretrained(name_or_path, **kwargs)
  File "/home/miniconda3/envs/ci/lib/python3.7/site-packages/mindformers/models/tokenization_utils_base.py", line 1872, in from_origin_pretrained
    config, cache_path = cls._download_using_name(name_or_path)
  File "/home/miniconda3/envs/ci/lib/python3.7/site-packages/mindformers/models/tokenization_utils_base.py", line 1959, in _download_using_name
    try_sync_file(file_path)
  File "/home/miniconda3/envs/ci/lib/python3.7/site-packages/mindformers/tools/utils.py", line 315, in try_sync_file
    with open(file_name, 'r') as fp:
FileNotFoundError: [Errno 2] No such file or directory: './checkpoint_download/llama2/tokenizer.model'
 
### Special notes for this issue/备注 (Optional / 选填)
走给 冯浩

GVP MindSpore/mindformers
关闭

内容风险标识

评论 (5)

GVPMindSpore/mindformers关闭

内容风险标识