Skip to content

RUN_LOG.md

Latest commit

 

History

History
109 lines (90 loc) · 3.12 KB

RUN_LOG.md

File metadata and controls

109 lines (90 loc) · 3.12 KB
Raw

RUN LOG

Stage 0

  • Workspace: /tmp/work/swarmagentic/SwarmAgenticCode
  • Repo: yaoz720/SwarmAgenticCode

Stage 1 - Environment inventory

sw_vers

ProductName: macOS ProductVersion: 26.3 BuildVersion: 25D125

uname -m

arm64

sysctl -n machdep.cpu.brand_string

which git && git --version

/opt/homebrew/bin/git git version 2.39.2

which python3 && python3 --version

/Users/linruihe/anaconda3/bin/python3 Python 3.10.9

which conda && conda --version

/Users/linruihe/anaconda3/bin/conda conda 23.9.0

Stage 2 - Conda env

  • Created env: swarm (python 3.11, osx-arm64)

python/arch in env

/Users/linruihe/anaconda3/envs/swarm/bin/python arm64

Stage 3 - Dependencies

  • Installed requirements with minimal fixes
  • Change 1: requirements.txt pydantic==2.12.3 -> pydantic==2.11.9
  • Change 2: pip override pandas to 2.3.3 (compatible with numpy 2.3.4 and gradio<3)

sanity import check

imports_ok

Stage 4 - API key load

API_keys.txt permissions

-rw------- API_keys.txt

  • detected_format=OpenAI API:VALUE OPENAI_API_KEY_set= True len= 164 prefix= sk-pro...

Stage 5 - Smoke test (travelplanner/swarm)

PSO command

python pso.py
--max_iteration 1
--settings 0.2 0.8
--sample_step 45
--max_workers 1
--save_dir evaluation_smoke
--dataset data/train_45.jsonl
--ref_info data/train_ref_info.jsonl

PSO result

  • Loaded 1 examples (sampled every 45 items)
  • Optimization completed
  • Global best fitness: 0.0
  • Global best trend: [0.0]

Artifact check

  • travelplanner/swarm/save.jsonl exists
  • travelplanner/swarm/evaluation_smoke/results-0-0.jsonl
  • travelplanner/swarm/evaluation_smoke/results-0-1.jsonl

Eval command

python test.py
--particle_idx -1
--start_index 0
--end_index 2
--max_workers 1
--save_dir evaluation_smoke/test
--dataset data/validation.jsonl
--ref_info data/validation_ref_info.jsonl

Eval printed metrics

  • Commonsense Constraint Pass Rate: {'Commonsense Constraint Micro Pass Rate': 0.0, 'Hard Constraint Micro Pass Rate': 0.0}
  • Hard Constraint Pass Rate: {'fallback_reason': "[Errno 2] No such file or directory: '../database/flights/clean_Flights_2022.csv'"}

Eval artifacts

  • travelplanner/swarm/evaluation_smoke/test/results.jsonl

Stage 6 - Normal run commands (heavier)

From repo root:

  • cd travelplanner/swarm
  • python pso.py --max_iteration 5 --dataset data/train_45.jsonl --ref_info data/train_ref_info.jsonl
  • python test.py --particle_idx -1
  • Default model: gpt-4o-mini (unless --model is provided)

Stage 7 - Deliverables

  • RUN_LOG.md (this file)
  • pip_freeze.txt
  • conda_env.txt

Code/Dependency changes applied

  1. requirements.txt
  • pydantic==2.12.3 -> pydantic==2.11.9 (gradio pin compatibility)
  1. Runtime package override (kept in environment)
  • pandas upgraded to 2.3.3 (to resolve numpy/pandas ABI issue with numpy 2.3.4)
  1. travelplanner/swarm/eval.py
  • Added fallback evaluator path when travelplanner/database assets are missing
  • Restored cwd after evaluator import attempt to avoid side effects
  1. travelplanner/swarm/pso.py
  • Added save_particles(particles) before save_state(...) so save.jsonl is created for test.py