Code

Agent optimization improvements#142

Code

Open

jpt-sn wants to merge 34 commits intomainServiceNow/TapeAgents:mainfrom optimize-reviewServiceNow/TapeAgents:optimize-reviewCopy head branch name to clipboard

+815-166

Commits

Commits on Nov 25, 2024

• Fix readme, lint, update res path

  jpt-sn

  committed
  129f4e6

  Copy full SHA for 129f4e6

• model_validate

  jpt-sn

  committed
  81ee234

  Copy full SHA for 81ee234

Commits on Dec 5, 2024

• Merge branch 'main' into optimize-review

  jpt-sn

  committed
  dfd6a19

  Copy full SHA for dfd6a19

• fix imports

  jpt-sn

  committed
  f2f9967

  Copy full SHA for f2f9967

• revert test

  jpt-sn

  committed
  f4042a0

  Copy full SHA for f4042a0

Commits on Dec 6, 2024

• add optimize_demos and improve robustness

  jpt-sn

  committed
  5da3c2b

  Copy full SHA for 5da3c2b

Commits on Dec 10, 2024

• metric as a function param, update dataset, use test data, log errors, llama support

  jpt-sn

  committed
  a1d2b4d

  Copy full SHA for a1d2b4d

Commits on Dec 11, 2024

• optimization with llama

  jpt-sn

  committed
  a062a2a

  Copy full SHA for a062a2a

• Merge branch 'main' into optimize-review

  jpt-sn

  committed
  ffef8db

  Copy full SHA for ffef8db

• update optimize test config

  jpt-sn

  committed
  0b25cec

  Copy full SHA for 0b25cec

• fix metric_fn param typing

  jpt-sn

  committed
  aeb4a01

  Copy full SHA for aeb4a01

Commits on Dec 27, 2024

• import HydraConfig

  jpt-sn

  committed
  341b9d3

  Copy full SHA for 341b9d3

• update optimize readme

  jpt-sn

  committed
  a2d57e6

  Copy full SHA for a2d57e6

• use litellm instead of trainablellm. Better error handling for litellm.

  jpt-sn

  committed
  0ef5e1b

  Copy full SHA for 0ef5e1b

• Merge branch 'main' into optimize-review

  jpt-sn

  committed
  1436f5f

  Copy full SHA for 1436f5f

• do not randomize number of demos

  jpt-sn

  committed
  440d684

  Copy full SHA for 440d684

• add docstring

  jpt-sn

  committed
  f7896da

  Copy full SHA for f7896da

• rename metric_mean_retrieval_answer to compute_weighted_accuracy

  jpt-sn

  committed
  eb1163f

  Copy full SHA for eb1163f

Commits on Dec 30, 2024

• raise error

  jpt-sn

  committed
  0d66181

  Copy full SHA for 0d66181

• update optimize tests for gpt35 and gpt4omini

  jpt-sn

  committed
  072c552

  Copy full SHA for 072c552

• llm_function uses output variables to extract information from llm output

  jpt-sn

  committed
  1b49308

  Copy full SHA for 1b49308

• add unit tests for llm function

  jpt-sn

  committed
  6bf2f42

  Copy full SHA for 6bf2f42

Commits on Jan 6, 2025

• Merge branch 'main' into optimize-review

  jpt-sn

  committed
  c2c9cad

  Copy full SHA for c2c9cad

• add llama3.3 optimization tests

  jpt-sn

  committed
  0fea981

  Copy full SHA for 0fea981

• fix gitignore

  jpt-sn

  committed
  1dc8677

  Copy full SHA for 1dc8677

• fix gitignore for sqlite tests

  jpt-sn

  committed
  f8936c0

  Copy full SHA for f8936c0

• reduce hydra logs

  jpt-sn

  committed
  668d538

  Copy full SHA for 668d538

• use max_tokens=1024 instead

  jpt-sn

  committed
  1b47073

  Copy full SHA for 1b47073

• improve CachedLLM logging

  jpt-sn

  committed
  ef2a752

  Copy full SHA for ef2a752

• updae make_test_data with tests groups

  jpt-sn

  committed
  b1f2fd5

  Copy full SHA for b1f2fd5

• update optimize tests

  jpt-sn

  committed
  7b55d5d

  Copy full SHA for 7b55d5d

• update optimize tests resources

  jpt-sn

  committed
  2f533dc

  Copy full SHA for 2f533dc

Commits on Jan 7, 2025

• simplified llm_function.generate_steps algo and rename RationaleOutput -> ReasoningOutput

  jpt-sn

  committed
  5c525fd

  Copy full SHA for 5c525fd

Commits on Jan 10, 2025

• Merge branch 'main' into optimize-review

  jpt-sn

  authored
  3aa5967

  Copy full SHA for 3aa5967