| { |
| "tag": "ppr32-scope", |
| "k": 16, |
| "problems_per_round": 32, |
| "rounds_run": 60, |
| "budget": 256, |
| "world_size": 1, |
| "source": "sft", |
| "temperature": 1.0, |
| "top_k": 50, |
| "init_lr_frac": 0.05, |
| "device_batch_size": 8, |
| "raw_lrs": { |
| "unembedding": 0.004, |
| "embedding": 0.2, |
| "matrix": 0.02, |
| "scalar": 0.5 |
| }, |
| "error": null, |
| "solve_rate_first": 0.2871, |
| "solve_rate_last": 0.3496, |
| "solve_rate_max": 0.5254, |
| "gen_tok_per_s_med": 35170.6, |
| "train_s_med": 0.9, |
| "round_s_med": 2.1, |
| "total_s": 130.0, |
| "build_s": 3.5, |
| "warm_s": 81.9, |
| "peak_mem_gb": 43.6, |
| "kv_pool_gb": 26.0, |
| "buckets": "64,128,256,512", |
| "checklist": { |
| "decode_body_compiled": true, |
| "decode_cuda_graphs_captured": 4, |
| "prefill_varlen_packed_graph": true, |
| "prefill_body_compiled": true, |
| "prefix_sharing_active": true, |
| "train_forward_compiled": true, |
| "train_static_buckets": [ |
| 16384 |
| ], |
| "fullparam_inplace_no_recapture": true, |
| "same_process_gen_train": true, |
| "bf16_params_fp32_opt_state": true, |
| "kv_pool_permanent_gb": 26.0, |
| "chat_rl_aligned_algorithm": true, |
| "tool_use": false |
| }, |
| "run_dir": "/root/.cache/nanochat-ds/fastrl_runs/ppr32-scope" |
| } |
