ZIm/crates/eval/examples/metal_i64_support/diff_criteria.md
Nathan Sobo bab28560ef
Systematically optimize agentic editing performance (#28961)
Now that we've established a proper eval in tree, this PR is reboots of
our agent loop back to a set of minimal tools and simpler prompts. We
should aim to get this branch feeling subjectively competitive with
what's on main and then merge it, and build from there.

Let's invest in our eval and use it to drive better performance of the
agent loop. How you can help: Pick an example, and then make the outcome
faster or better. It's fine to even use your own subjective judgment, as
our evaluation criteria likely need tuning as well at this point. Focus
on making the agent work better in your own subjective experience first.
Let's focus on simple/practical improvements to make this thing work
better, then determine how we can craft our judgment criteria to lock
those improvements in.

Release Notes:

- N/A

---------

Co-authored-by: Max <max@zed.dev>
Co-authored-by: Antonio <antonio@zed.dev>
Co-authored-by: Agus <agus@zed.dev>
Co-authored-by: Richard <richard@zed.dev>
Co-authored-by: Max Brunsfeld <maxbrunsfeld@gmail.com>
Co-authored-by: Antonio Scandurra <me@as-cii.com>
Co-authored-by: Michael Sloan <mgsloan@gmail.com>
2025-04-19 02:47:59 +00:00

1.1 KiB

  1. The changes improve the configurability of the TextGeneration struct and its initialization by refactoring generation parameters (temperature, top_p) to use non-optional types with default values, simplifying their use throughout the codebase.
  2. The argument parser is updated to enhance usability: verbose_prompt is renamed to a more general verbose flag, several arguments are given default values (e.g., temperature, top_p, sample_len), and optional arguments like cache_path and weight_path are now properly handled with conditional logic and fallbacks.
  3. The code loading the model configuration is updated to support deserializing from a JSON config file using Serde, and the Config struct is extended with a new rope_ratio field with a default value via a helper function, improving flexibility for different model setups.
  4. Import statements and general code layout are cleaned up for clarity and consistency, including reorganizing imports and removing unnecessary unwraps or panics, while maintaining the same core functionality of the text generation pipeline.