Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
Lin-Bert 8470b5f601 | 5 months ago | |
---|---|---|
.. | ||
bert | 6 months ago | |
blip2 | 6 months ago | |
bloom | 5 months ago | |
clip | 6 months ago | |
codegeex2 | 6 months ago | |
general | 6 months ago | |
glm | 6 months ago | |
glm2 | 6 months ago | |
gpt2 | 6 months ago | |
llama | 5 months ago | |
llama2 | 5 months ago | |
mae | 6 months ago | |
pangualpha | 6 months ago | |
qa | 6 months ago | |
sam | 6 months ago | |
swin | 6 months ago | |
t5 | 6 months ago | |
tokcls | 6 months ago | |
txtcls | 6 months ago | |
vit | 6 months ago | |
README.md | 6 months ago |
configs统一在run_xxx.yaml中,排序按照修改频率的顺序和一般的模型训练流程顺序(数据集->模型->训练、评估、推理),具体顺序如下
需要满足实际运行的卡数 device_num = data_parallel × model_parallel × pipeline_stage
type: 模型参数配置类
checkpoint_name_or_path: 评估时不指定权重,模型默认加载的权重名
# 以下配置针对大规模语言模型推理
top_k: 从概率最大的top_k个tokens中采样
top_p: 从概率最大且概率累计不超过top_p的tokens中采样
do_sample: 使能top_k或top_p采样,为False时top_k和top_p均重置为1
use_past: 使能增量推理,为True时为增量推理,否则为自回归推理,使用时请参考模型支持列表
max_decode_length: 文本生成最大长度(输入长度统计在内)
repetition_penalty: 重复文本惩罚系数,该值不小于1,等于1时不惩罚
MindSpore Transformers套件的目标是构建一个大模型训练、推理、部署的全流程套件: 提供业内主流的Transformer类预训练模型, 涵盖丰富的并行特性。 期望帮助用户轻松的实现大模型训练。 文档:https://mindformers.readthedocs.io/zh-cn/latest/
Jupyter Notebook Python Markdown Shell
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》