THUDM / ChatGLM2-6B

ChatGLM2-6B: An Open Bilingual Chat LLM | 开源双语对话语言模型
Other
15.74k stars 1.85k forks source link

为什么ChatGLM2-6B的序列长度只有1024呢?? #541

Open evyy1234 opened 1 year ago

evyy1234 commented 1 year ago

Is there an existing issue for this?

Current Behavior

用的ChatGLM2-6B-32k模型,报错如下: Token indices sequence length is longer than the specified maximum sequence length for this model (2430 > 1024). Running this sequence through the model will result in indexing errors 官网不是宣传对话阶段使用 32K 的上下文长度训练,为什么这里才支持1024的token呢?我看别人好像都是2048,这个值可以如何设置呢?

Expected Behavior

更长的输入tokens支持

Steps To Reproduce

。。。

Environment

- OS:Ubuntu
- Python:3.10
- Transformers:
- PyTorch:4.26.1
- CUDA Support (`python -c "import torch; print(torch.cuda.is_available())"`) :true

Anything else?

nothing