============================================================
GRPO E2E TEST REPORT: test_e2e_training_loop_mock
============================================================
Status:               passed
Duration:             0.04s
Training steps:       2
Rollouts per step:    4
Total rollouts:       8

--- Model Weight Change ---
Weights changed:      True
Total delta L2 norm:  0.185388
  embedding.weight: delta=0.025263 (norm 90.3909 -> 90.3893)
  head.weight: delta=0.180862 (norm 9.1759 -> 9.1803)
  head.bias: delta=0.031929 (norm 1.6532 -> 1.6515)

--- Training Metrics ---
  Step 0: loss=0.0000 reward_mean=0.50 grad_norm=1.0000 valid_terms=8
  Step 1: loss=0.0000 reward_mean=0.50 grad_norm=1.0000 valid_terms=8

--- Checkpoint ---
Checkpoint path:      test_artifacts/grpo_e2e/20261016_215856/checkpoint/step_2
Checkpoint loadable:  True

============================================================