Running training command                            
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/.venv/bin/
python3 -m policy_models.cli.run_tasks train_from_episodes --episodes_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/persistent
-data/state-service/episodes --checkpoint_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/human --pv_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/persistent
-data/trainer/models --device cpu --backbone lg_transformer --bc_epochs 20 
--ppo_epochs 40 --h_max 160 --w_max 120 --tb_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/tb/human --run_label human
 starting run 'human'
checkpoint dir: 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/human | tensorboard dir: 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/tb/human
Failed to push training telemetry event: <urlopen error [Errno 61] Connection refused>
Skipping zero-style recordings: record count <= 0
          Waiting for state service tester queue to drain (zero-style)          
tester queue size=0; is_empty=True
Tester queue drained; continuing
                            Running training command                            
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/.venv/bin/
python3 -m policy_models.cli.run_tasks train_from_episodes --episodes_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/persistent
-data/state-service/shallow-zero-style-episodes --checkpoint_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/zero_style --pv_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/persistent
-data/trainer/models --device cpu --backbone lg_transformer --bc_epochs 0 
--ppo_epochs 20 --h_max 160 --w_max 120 --tb_dir 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/tb/zero_style --run_label zero_style --init-from-pv
 starting run 'zero_style'
checkpoint dir: 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/zero_style | tensorboard dir: 
/Users/michaelsigamani/Documents/DevelopmentCode/2025-fall/codeassist/logs/_trai
ning_artifacts/tb/zero_style
Loaded assistant policy weights from persisted models
Loaded human policy weights from persisted models
Loaded featurizer weights from persisted models
Failed to push training telemetry event: <urlopen error [Errno 61] Connection refused>