Rain

Results 4 comments of Rain

new_beams.append( Beam( beam.text, beam.next_word, **beam.partial_word + " " + char,** char, beam.text_frames, new_part_frames, beam.logit_score + p_char, ) ) line 528, decoder.py, I add space in: beam.partial_word + " " +...

modify:ChatGLM-Instruct-Tuning/modeling_chatglm.py", line 836 to MASK, gMASK = 130000, 130001

> > modify:ChatGLM-Instruct-Tuning/modeling_chatglm.py", line 836 to MASK, gMASK = 130000, 130001 > > 还有两个文件更新了,[tokenization_chatglm.py](https://huggingface.co/THUDM/chatglm-6b/blob/main/tokenization_chatglm.py)和[tokenizer_config.json](https://huggingface.co/THUDM/chatglm-6b/blob/main/tokenizer_config.json)这两个替换了吗 谢谢你的信息,这两个文件没有直接在tuning项目中引用,我估计已经直接使用了缓存中的最新的huggingface这两个文件。你那边运行成功了吗?我这边跑起来的时候一下就吃完了30G显存,把模型改成load_in_8bit则出各种问题,还未跑起来,请问有何建议,谢谢~