sisl / astra-rl / 17971784237

36%

Build Type

push

github

Committed by web-flow

Commit Message

Documentation and initial experiments (#16)

* first trial of gpt2 v gpt2

* Add custom moderator guide, llamaguard implementation, and updates

* updated documentation

* updated docs

* Delete docs/tutorials/quick_start_training_old.md

* Delete docs/tutorials/custom_moderators

* fixed out of index error in gpt2 run

* fixed out of index error in gpt2 run

* fixed out of index error in gpt2 run, init tokenizer diff

* training docs

* updated example with eval

* Delete docs/tutorials/customize_training/prob.md

* training/dev/test reddit prompts

Initial prompts to rollout adv/defender conversations. They are cleaned and sorted to be non-toxic.

* Delete docs/tutorials/customize_training directory

* pushing updated custom docs

* fixing quick start

* updating layout and gitignore

* bookkeeping

* update gitignore

* initial prompts

* update quick start

* typo

* quickstart updates

* quickstart fixes

* fix

* adding hf_ast trainer class plus updating ASTEnvironment

* update ast hf trainer class'

* update trainer docs

* updated docs

* updated training docs

* fixed training docs

* backup

* graph example

* graph example

* graph example

* backup works yay

* stub the eval funciton

* llama eval script

* actually call the eval script

* De/trail1+docs+fixes (#15)

* Address mkdocs warnings. Add installation and dev_setup pages. Update base documentation page

* Add footnote citations. Fix environment note formatting

* ast_problem updates

* small fix

* doc updates for llama3 baseline

---------

Co-authored-by: alliegriffith <allie11@stanford.edu>
Co-authored-by: Duncan Eddy <6956476+duncaneddy@users.noreply.github.com>

Run Details

10 of 138 new or added lines in 4 files covered. (7.25%)

3 existing lines in 2 files now uncovered.

267 of 680 relevant lines covered (39.26%)

0.79 hits per line