Design
Which solid notes this experiment tests, and how.
Results
20260901-094918-qwen3_4b_gsm8k_decomposed_lrdir025_lrmag10_jean_zay-b9847c9
commit: b9847c9
| metric | value |
|---|---|
| center_eval_do_sample | 0 |
| center_eval_every | 5 |
| center_eval_temperature | 1 |
| center_eval_top_k | 0 |
| completed | 0 |
| decomposed_lr_direction | 0.25 |
| decomposed_lr_magnitude | 1 |
| decomposed_sigma_direction | 0.001 |
| decomposed_sigma_magnitude | 0.001 |
| decomposed_tangent_project_direction | 0 |
| eggroll_lr | 1 |
| eggroll_sigma | 0.001 |
| epochs | 15 |
| eval_batch_size | 8 |
| eval_prompts | 128 |
| generations_per_prompt | 6 |
| max_new_tokens | 1024 |
| method_decomposed_enabled | 1 |
| method_plain_enabled | 0 |
| noise_reuse | 4 |
| num_center_eval_records | 3 |
| num_completed_runs | 0 |
| num_epoch_records | 15 |
| num_methods | 1 |
| num_ranks | 1 |
| num_seeds | 1 |
| num_target_module_patterns | 8 |
| parseability_shaping | 0 |
| population | 126 |
| prompts_per_epoch | 21 |
| temperature | 1 |
| top_k | 0 |
| train_batch_size | 6 |
| wall_clock_seconds | 35821.46464041597 |
Interpretation
Run 20260901-094918-qwen3_4b_gsm8k_decomposed_lrdir025_lrmag10_jean_zay-b9847c9
is a partial calibration point, not a completed result: Slurm killed it at the
time limit after all epoch records had been flushed but before the final
center evaluation and run summary could complete. The available center evals
still show that this larger magnitude learning rate was not obviously unstable
early; it reached a strong center-model eval score by the last recorded center
checkpoint while using exact-answer reward without parseability shaping.
The boring alternative explanation is that this is another fixed-batch fit
whose apparent eval gain is sensitive to the small deterministic eval slice.
Because the run timed out, it should not be ranked ahead of completed points
without rerunning or reducing evaluation cost. For now it mainly says the
lr_direction=0.25, lr_magnitude=1.0 setting is worth keeping in the
candidate set if the rest of the grid does not dominate it cleanly.