Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / l294265421/alpaca-rlhf issues and pull requests

#15 - 训练问题

Issue - State: open - Opened by wanghao-007 about 1 year ago

#13 - step2和step3中padding side似乎不一样？

Issue - State: open - Opened by qiancheng99 over 1 year ago - 1 comment

#12 - A question about setting tokens

Issue - State: open - Opened by hepj987 over 1 year ago - 1 comment

#11 - element 0 of tensors does not require grad and does not have a grad_fn

Issue - State: open - Opened by Bill-Orz over 1 year ago - 5 comments

#10 - Fix pad_token_id bug

Issue - State: closed - Opened by Ablustrund over 1 year ago - 2 comments

#9 - 关于Step3中是否需要把生成的answer中eos后面token mask掉

Issue - State: closed - Opened by Ablustrund over 1 year ago - 1 comment

#8 - deepspeed.initialize的一些疑惑

Issue - State: closed - Opened by iamsile over 1 year ago - 8 comments

#7 - how to run it, need more details

Issue - State: open - Opened by SeekPoint over 1 year ago - 2 comments

#6 - v100 step3 oom

Issue - State: closed - Opened by iamsile over 1 year ago - 12 comments

#5 - stop at step2 evaluation_reward

Issue - State: open - Opened by murphypei over 1 year ago - 4 comments

#4 - reward model在v100上训练时会卡住不动

Issue - State: closed - Opened by iamsile over 1 year ago - 2 comments

#3 - v100训练时显存oom

Issue - State: closed - Opened by iamsile over 1 year ago - 2 comments

#2 - Steps

Issue - State: open - Opened by syngokhan over 1 year ago - 1 comment

#1 - 训练效果怎么样

Issue - State: closed - Opened by Curious-chen over 1 year ago - 3 comments