| { |
| "sft_checkpoint": "tinker://398393e1-7182-555d-aa1b-7ddf23892338:train:0/sampler_weights/sft_final_sampler", |
| "rl_checkpoint": "tinker://a4ce5c47-f124-5cac-b700-d6071da84f98:train:0/sampler_weights/rl_final", |
| "sft_metrics": { |
| "any_match": 0.87, |
| "exact_match": 0.47, |
| "precision": 0.7266666666666667, |
| "recall": 0.75, |
| "mean_reward": 0.8087999999999999, |
| "format_valid": 1.0, |
| "f1": 0.7381489841986457 |
| }, |
| "rl_metrics": { |
| "any_match": 0.88, |
| "exact_match": 0.48, |
| "precision": 0.7383333333333334, |
| "recall": 0.77, |
| "mean_reward": 0.8179999999999998, |
| "format_valid": 1.0, |
| "f1": 0.7538342541436466 |
| }, |
| "rl_training_log": [ |
| { |
| "iteration": 0, |
| "mean_reward": 0.81125, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 72.97691488265991 |
| }, |
| { |
| "iteration": 1, |
| "mean_reward": 0.8003125, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 198.50843501091003 |
| }, |
| { |
| "iteration": 2, |
| "mean_reward": 0.81375, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 31.24391794204712 |
| }, |
| { |
| "iteration": 3, |
| "mean_reward": 0.91375, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 54.815149784088135 |
| }, |
| { |
| "iteration": 4, |
| "mean_reward": 0.8765624999999999, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 46.36493897438049 |
| }, |
| { |
| "iteration": 5, |
| "mean_reward": 0.7537499999999999, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 41.113487005233765 |
| }, |
| { |
| "iteration": 6, |
| "mean_reward": 0.714375, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 59.793246030807495 |
| }, |
| { |
| "iteration": 7, |
| "mean_reward": 0.721875, |
| "accuracy": 0.984375, |
| "num_rollouts": 64, |
| "time": 25.620058059692383 |
| }, |
| { |
| "iteration": 8, |
| "mean_reward": 0.72, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 32.78582692146301 |
| }, |
| { |
| "iteration": 9, |
| "mean_reward": 0.7989062499999999, |
| "accuracy": 1.0, |
| "num_rollouts": 64, |
| "time": 21.66036295890808 |
| } |
| ] |
| } |