Runtime 1m 30s   Processes 3   Roles actor x2, learner x1
Host golden-host

RESOURCES
──────────────────────────────────────────────────────────────
                            used     available      per proc
RAM  peak RSS           677.1 MB                    225.7 MB

  (machine capacity was not recorded for this run)
  per-proc figures are over 3 process(es) (actor x2, learner x1)
  heaviest process held 476.8 MB RSS against a 225.7 MB mean
  (no CPU readings in this run's samples; install psutil to record them)

ACTOR  (2 processes, imbalance 1.11)
  % of phase wall time at the first branching level, summed over 2 processes
──────────────────────────────────────────────────────────────
mcts                           75.0%       675.0ms (Σ2 proc)
env_step                       24.0%       216.0ms (Σ2 proc)
Other                           1.0%         9.0ms (Σ2 proc)

DOMINANT PHASES          entries        self    wait       p50       p99
iteration/mcts                40     675.0ms     91%    15.7ms    18.8ms
    + mcts_simulations          2,560     3,792.6/s  263.7us/ea
iteration/env_step            40     216.0ms     81%     5.2ms     6.3ms
    + env_steps                    40       185.2/s    5.4ms/ea
iteration                     40       9.0ms     91%    21.0ms    25.1ms

ITERATIONS  (40 entries)
  mean     22.5ms   p50     21.0ms   p95     25.0ms   p99     25.1ms

LEARNER  (1 process, imbalance 1.00)
  % of phase wall time at the first branching level
──────────────────────────────────────────────────────────────
train_step                     80.0%       400.0ms
checkpoint_to_object_store     19.0%        95.0ms
Other                           1.0%         5.0ms

DOMINANT PHASES          entries        self    wait       p50       p99
iteration/train_step          10     400.0ms      2%    39.8ms    41.9ms
    + train_samples             1,280     3,200.0/s  312.5us/ea
…ckpoint_to_object_store       2      95.0ms     89%    48.2ms    50.3ms
iteration                     10       5.0ms      4%    48.2ms    50.3ms

ITERATIONS  (10 entries)
  mean     50.0ms   p50     48.2ms   p95     50.1ms   p99     50.3ms

I/O BY PHASE (measured exactly)
──────────────────────────────────────────────────────────────
  …eckpoint_to_object_store r        0 B   w    40.0 MB    421.1 MB/s

I/O
──────────────────────────────────────────────────────────────
Read (from disk)                       0 B           0 B/s
Write                              40.0 MB        8.0 MB/s

  iteration/train_step      r        0 B   w    40.0 MB

  read                                                    peak 0 B/s
  write █        █         █        █         █           peak 76.8 MB/s

  (bytes come from the OS process counters; attribution to a phase has a
   resolution of one sample interval. Per-operation attribution needs eBPF.)

GPU
──────────────────────────────────────────────────────────────
  No GPU data was collected. If this run used one, install nvidia-ml-py and
  re-read the report without --no-samples; a phase around device work reports
  submission time until something proves the device was busy.

MEMORY (RSS)
──────────────────────────────────────────────────────────────
Current                           677.1 MB
Peak                              677.1 MB
Growth over run                     9.5 MB
Trend                             1.9 MB/s

  growing under iteration/mcts                1.9 MB/s