Commits · 331a4901115f44937116830860a65f8db8eacb6e · Grießhaber Daniel / evoprompt · GitLab

Snippets Groups Projects

Aug 19, 2024
- simplify history handling in `ChatModel.create_completion` · 331a4901
  Grießhaber Daniel authored 7 months ago
  
  331a4901
- refactor llm model abstraction · a78c704c
  Grießhaber Daniel authored 7 months ago
  
  a78c704c
- respect --max-tokens cli parameter · 66338b96
  Grießhaber Daniel authored 7 months ago
  
  66338b96
- remove OpenAI (non-chat) model · ad25c5a5
  Grießhaber Daniel authored 7 months ago
  
  ad25c5a5
- fix build_model_input for OpenAI models to use same layout as the LLama models · 33f58a73
  Grießhaber Daniel authored 7 months ago
  
  33f58a73
- refactor build_model_input to not modify its parameters · b247f106
  Grießhaber Daniel authored 7 months ago
  
  b247f106
- replace path separators in model_name with underscores · 46b178e4
  Grießhaber Daniel authored 7 months ago
  
  46b178e4
- Apply 1 suggestion(s) to 1 file(s) · 1750086e
  Grießhaber Daniel authored 7 months ago
  
  Co-authored-by: Max Kimmich <maximilian.kimmich@ims.uni-stuttgart.de>
  1750086e
- Apply suggestions to 1 file · ce71ba8d
  Grießhaber Daniel authored 7 months ago
  
  ce71ba8d
- refactor _call_model_cached function implementation · 13b82620
  Grießhaber Daniel authored 7 months ago
  
  13b82620
- make OpenAI.build_model_input return all arguments as dict · 7420ba00
  Grießhaber Daniel authored 7 months ago
  
  7420ba00
- fix error for openai model after refactoring · b47cdc45
  Grießhaber Daniel authored 7 months ago
  
  b47cdc45
Aug 16, 2024
- remove max_tokens parameter from create_completion · 1083a08d
  Grießhaber Daniel authored 7 months ago
  
  1083a08d
- Add CLI option for maximum number of tokens being generated · b4f9fd21
  Max Kimmich authored 7 months ago and Grießhaber Daniel committed 7 months ago
  
  b4f9fd21
- remove need for cache key computation during runtime · 3ae988c3
  Grießhaber Daniel authored 7 months ago
  
  3ae988c3
- remove is_chat argument · 2df18f7f
  Grießhaber Daniel authored 7 months ago
  
  2df18f7f
- Add CLI option for maximum number of tokens being generated · 691ced52
  Max Kimmich authored 7 months ago
  
  691ced52
- Add option to disable caching and use separate caches on a model basis · 8ba5ca7f
  Max Kimmich authored 7 months ago
  
  8ba5ca7f
- Add caching functionality for model calls · a68e6415
  Max Kimmich authored 7 months ago
  
  a68e6415
- Refactor models · d685a040
  Max Kimmich authored 7 months ago
  
  d685a040
Aug 14, 2024
- Fix cache not taking into account model · bc8b5acf
  Max Kimmich authored 7 months ago
  
  bc8b5acf
- Improve prompt extraction · 03728b20
  Max Kimmich authored 7 months ago
  
  03728b20
Aug 09, 2024
- Fix prompt extraction for DE evolution · 36eef613
  Max Kimmich authored 7 months ago
  
  36eef613
- Fix bug where prompt was not evolved correctly due to too few tokens being generated · 1c97f13a
  Max Kimmich authored 7 months ago
  
  1c97f13a
- Allow to set random seed to allow for reproducible experiments · 580b49d7
  Max Kimmich authored 7 months ago
  
  580b49d7
- Add CoT for DE · d3db25c6
  Max Kimmich authored 8 months ago
  
  d3db25c6
- Enforce abstract class check · faf3be25
  Max Kimmich authored 8 months ago
  
  faf3be25
- Remove system message from DE prompt template · 46117c4b
  Max Kimmich authored 7 months ago
  
  46117c4b
- Perform some refactoring · d045eee1
  Max Kimmich authored 7 months ago
  
  d045eee1
- log and return correct prompts in get_initial_prompts · d8a5f185
  Grießhaber Daniel authored 7 months ago
  
  d8a5f185
- implement para_topk algorithm from https://github.com/beeevita/EvoPrompt for initial population · db61291e
  Grießhaber Daniel authored 7 months ago
  
  db61291e
- load initial prompts from json file for tasks where applicable · 328be630
  Grießhaber Daniel authored 7 months ago
  
  328be630
- refactored base_prompts property to support multiple initial prompts · 5851162e
  Grießhaber Daniel authored 7 months ago
  
  5851162e
- add .cache_dir to gitignore · f9391084
  Grießhaber Daniel authored 7 months ago
  
  f9391084
- add initial prompts from https://github.com/beeevita/EvoPrompt/ · db803426
  Grießhaber Daniel authored 7 months ago
  
  db803426
- refactor initial prompt generation · 29cb6b17
  Grießhaber Daniel authored 7 months ago
  
  29cb6b17
- add persistent caching to `_evaluate_sample` · b9311d44
  Grießhaber Daniel authored 7 months ago
  
  b9311d44
Aug 08, 2024
- Add methods for using demonstrations in evaluation · aa9e4327
  Max Kimmich authored 7 months ago
  
  aa9e4327
- Add demonstration for DE · dad15285
  Max Kimmich authored 7 months ago
  
  dad15285
Aug 06, 2024
- Always select subset for evolution · de38cf26
  Max Kimmich authored 7 months ago
  
  de38cf26