[Inference] Add models CodeLlama-7b and falcon-7b #12
Conversation
* add testing scripts * remove temp-dir for worker * remove test files * add redpajama dp code * ignore all notebook files * update streaming code * add write-on-host for streaming * better line alignment * move files * rename folder * rename folder and add group_files * debug * add recovery test scripts * add additional python packages * add test flag * add README and some minor fixes * change the image name * change the directory back * add training stop for the second * fix typo * add data source support * clean up a bit * restructure folders * restructure files * add script headers * reorder and add READMEs * revert back due to file movements * fix typo * fix lib import * enable mounting localdisk * change name of cc * fix dtype * performance optimization for streaming * use the latest ray * change node * add new files * bug fix * add nltk * fix hdfs after re-order folders * set default to false * use variables instead of credentials * change the training config path * update README
|
Gentle ping @jiafuzha for review, thanks~ |
|
you need to add the two models in the include list so that they can be verified in PR. Otherwise, they will only be verified in nightly. include: |
Ic.. it passed in our previous repo. Let me trigger them here. |
|
Gentle ping @jiafuzha for another review, thanks! |
| - { model: "CodeLlama-7b-hf"} | ||
| - { model: "falcon-7b"} |
There was a problem hiding this comment.
please review them before merging it since we've verified them successfully.
There was a problem hiding this comment.
Ok, I will remove it as soon as it's ready to merge.
No description provided.