hjc3613

predict result repeat the same result

3

train on 30000000 parallel english and chinese sentences, after trained 100000 steps, the predit result:  content of config.yaml: # wmt14_en_de.yaml save_data: data/wmt/run/example # Corpus opts: data: corpus_1: path_src: corpus/corpus_1/en-zh.en.bpe...

您好，请问能提供nyt10数据集的json格式下载连接吗？官网下的都是pb文件

Can't find adapter_config.json

3

微调后，没有生成adaptor_config.json文件，也没有config.json文件，所以在合并时，报错了：Can't find adapter_config.json - [√ ] **基础模型**： Alpaca-Plus - [ √] **运行系统**： Linux - [ √] **问题分类**：模型转换和合并 / 模型训练与精调 / - [ √] （必选）由于相关依赖频繁更新，请确保按照[Wiki](https://github.com/ymcui/Chinese-LLaMA-Alpaca/wiki)中的相关步骤执行 - [ √] （必选）我已阅读[FAQ章节](https://github.com/ymcui/Chinese-LLaMA-Alpaca/wiki/常见问题)并且已在Issue中对问题进行了搜索，没有找到相似问题和解决方案 - [...

weight must be 2-D error

5

### 详细描述问题 执行 run_sft.sh时，报错：  注：我用的deepspeed stage 3，配置项：  run_sft.sh脚本内容为：  run_sft.sh与原始内容的差异在于我去掉了peft_path，让模型重新初始化，其它的基本没变动 ### 必查项目（前三项只保留你要问的） - [ √] **基础模型**：Alpaca-Plus - [ √] **运行系统**：Linux - [ √] **问题分类**： 模型训练与精调 - [ √]...

freezing layers have differenct behaves for different models

2

### System Info Dockerfile:  ### Information - [ ] The official example scripts - [ ] My own modified scripts ### 🐛 Describe the bug when freezing the top...

does deepspeed support pure bf16 training?

9

**training 70B cost too large gpu memory** when training 70+B model, the cost of gpu memory is too large, as mentioned in the deepspeed deepspeed-readthedocs-io-en-stable.pdf, the total memory was about...

could you please provide details about the implementation of the trice algorithm? the code in https://github.com/google-research/cascades/blob/main/cascades/examples/notebooks/trice.ipynb has few functions unimplemented, I want to refer back to the original implementation, thanks...

[Feature]: speed up maskgct

5

## Is your feature request related to a problem? Please describe. (A clear and concise description of what the problem is.) when I generate 20s audio, the time cost is...

Notice: In order to resolve issues more efficiently, please raise issue following the template. （注意：为了更加高效率解决您遇到的问题，请按照模板提问，补充细节） ## ❓ Questions and Help ### Before asking: 1. search the issues. 2. search the...