Repository navigation
Conversation
The prompt sampler only reads prompt.system_message, and load_config then copied that value to every model, so a system message set under llm was replaced by the default template without a warning. The Rust and R examples set theirs there, so their prompts never reached the model. Config loading (and the controller, for configs built in code) now uses llm.system_message for the prompt when prompt.system_message is unset, warns that the key belongs under prompt, and warns when both are set. The two examples move their messages under prompt.
|
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The prompt sampler only reads
prompt.system_message, andload_configthen copies that value to every model. So a system message set underllm:gets replaced by the default template, with no warning.The Rust and R examples both set theirs under
llm:, so their instructions never reach the model. On main, withexamples/rust_adaptive_sort/config.yaml:llm.system_message: "You are an expert Rust systems programmer specializing in high-performance algorithms..."I ran into this while using the Rust example as a template for a Java experiment.
Changes:
Config.from_dictusesllm.system_messageasprompt.system_messagewhen the prompt one is unset, and warns that the key belongs underprompt:. If both are set,prompt.system_messagewins and the warning saysllm.system_messageis ignored.validate_evolution_mode(). A second call does nothing, so the warning shows once.prompt:, like the other 42 configs that set one.Tests: new
tests/test_system_message_config.py, 5 of its 6 tests fail on main. Full suite (python -m unittest discover -s tests -p "test_*.py"): 1082 OK, 7 skipped.