Ting

about wgan-gp

2

i rewrite caogang/wgangp code to generate 512*512 images , i add 2-3 conv2d layers in Dnet and Gnet, but when i run that code ,the loss explode to 30000000, and...

Add dpo

2

### PR types ### PR changes ### Description 新增 dpo 训练框架及单测

from_aistudio supports chat_template

7

### PR types ### PR changes ### Description tokenizer.from_pretrained支持 aistuidio

Chatglm2 baichuan ci

4

### PR types ### PR changes ### Description 新增chatglm2 ci，baichuan 因 fused llama 不支持 alibi 暂时没加入 ci

gconfig support from aistudio

4

### PR types ### PR changes ### Description generaton config支持 from aistudio

bug fix

3

### PR types ### PR changes ### Description 修复 max_steps ＞ epoch_steps 的计算错误问题

[Unified Checkpoint] Checkpoint compression

3

### PR types ### PR changes ### Description checkpoint 压缩功能实现 新增参数 - --ckpt_quant_stage "O0"/"O1"/"O2" - O0：不压缩 - O1：channel-wise int8 压缩 - O2：group-wise int4 压缩 - --checkpoint_config "remove_master_weight" - amp O2开启此...