Commit 8424ebab authored by Gaetano Scandariato's avatar Gaetano Scandariato
Browse files

chore: rerun the TOI-5398 RM runs on the laptop GPU after the TF32 fix; queue for the H100 runs 

Co-Authored-By: default avatarClaude Sonnet 5.5 <noreply@anthropic.com>
parent 4296ee36

paper_benchmarks/machines/round6_rm_gpu_server.sh

0 → 100755
+19 −0
Changes for paper_benchmarks/machines/round6_rm_gpu_server.sh: 19 added lines, 0 removed lines.
Original line number Diff line number Diff line 
#!/bin/bash

# 2026-10-04: rerun the TOI-5398 RM runs made on the H100 after the fix of the TF32 matrix products in the RM disc

# integrals (CHANGELOG, "Fixed before the release"; GPU results were biased by up to 2 m/s near ingress and egress).

# Waits for heavy.lock, then replaces the sources by the staged ones (<work dir>/staging_src, copied from the

# laptop) so that a queue running before is not affected, and reruns the two H100 runs.

# usage: round6_rm_gpu_server.sh <work dir>

set -u

W=$(realpath "$1"); B=$W/PyAETNA/paper_benchmarks; LOG=$W/logs/round6.log

exec 9> "$W/heavy.lock"; flock 9

export PATH=$W/envs/pyaetna/bin:$PATH GNU_TIME=$W/tools/gnutime/bin/time

say() { echo "$(date '+%F %T') $*" >> "$LOG"; }

rsync -a --delete "$W/staging_src/PyAETNA/" "$W/PyAETNA/src/PyAETNA/"

say "sources replaced; fix present: $(grep -c _rm_disc_terms_tf32_unsafe "$W/PyAETNA/src/PyAETNA/body_models.py")"

for r in toi5398_rm_pyorbit/runs/ccf_h100 toi5398_rm_pyorbit/runs/template_h100; do

  say "start $r"

  (cd "$B" && nice -n 19 ./run.sh "$r" cuda,cpu 64G > /dev/null)

  say "end $r $(cat "$B/$r/done.txt") $(grep 'Elapsed (wall' "$B/$r/time.out")"

done

say "done"

paper_benchmarks/toi5398_rm_pyorbit/results/summary.json

+292 −264

File changed.

Preview size limit exceeded, changes collapsed.

paper_benchmarks/toi5398_rm_pyorbit/runs/ccf/PyAETNAouts/toi5398_ccf_results.json

+111 −111
Changes for paper_benchmarks/toi5398_rm_pyorbit/runs/ccf/PyAETNAouts/toi5398_ccf_results.json: 111 added lines, 111 removed lines.
Original line number Diff line number Diff line
@@ -2,15 +2,15 @@ 
 "format": "pyaetna-results",

 "format_version": 1,

 "evidence": {

  "log_evidence": -159.99468963911812,

  "log_evidence_err": 0.1424691145459429,

  "information_nats": 10.671224135591586,

  "log_evidence": -160.4913298318188,

  "log_evidence_err": 0.14283947217486376,

  "information_nats": 10.779646476969532,

  "converged": true,

  "n_iterations": 145,

  "n_dead": 7750,

  "n_likelihood_evals_approx": 934459,

  "effective_sample_size": 2782.9857339285336,

  "elapsed_s": 9.417337417602539,

  "n_iterations": 146,

  "n_dead": 7800,

  "n_likelihood_evals_approx": 928345,

  "effective_sample_size": 2828.5547580563,

  "elapsed_s": 8.797070980072021,

  "prior_rejected_fraction": 0.0,

  "n_dim": 12,

  "num_inner_steps": 48,
@@ -28,64 +28,64 @@ 
  },

  "posterior_summary": {

   "delta_period_b": {

    "median": -2.159204057558643e-06,

    "p16": -1.4144445910196197e-05,

    "p84": 9.237386499835044e-06

    "median": -1.3620162077180183e-06,

    "p16": -1.3426471699823085e-05,

    "p84": 1.0642820659138833e-05

   },

   "delta_tau_b": {

    "median": -1.917992707743339e-05,

    "p16": -0.00023205103064095615,

    "p84": 0.00020419205118929888

    "median": -1.659065235055351e-05,

    "p16": -0.00022311028172309401,

    "p84": 0.0002106864166021556

   },

   "impact_b": {

    "median": 0.22628426301176194,

    "p16": 0.1476452048278822,

    "p84": 0.2882437062258545

    "median": 0.22653094186803405,

    "p16": 0.14908310573189062,

    "p84": 0.2894641500123522

   },

   "kRV_b": {

    "median": 15.777915147880066,

    "p16": 14.331618385080802,

    "p84": 17.22522301218314

    "median": 15.771787219299096,

    "p16": 14.328834283993505,

    "p84": 17.270133975151637

   },

   "lambda_spin_b": {

    "median": -0.003627305578704921,

    "p16": -0.09454488121842741,

    "p84": 0.08419552439912176

    "median": 0.004872057772958143,

    "p16": -0.08369551947209045,

    "p84": 0.09497087101889251

   },

   "log_jitter": {

    "median": -1.6129917093670039,

    "p16": -3.701388751770221,

    "p84": 0.5322771595464679

    "median": -1.4086148214986758,

    "p16": -3.5680925618052988,

    "p84": 0.5328984576728384

   },

   "norm": {

    "median": -9.417746370063648,

    "p16": -9.418682193861917,

    "p84": -9.416734198484216

    "median": -9.417615836829569,

    "p16": -9.418627596760526,

    "p84": -9.416710030875969

   },

   "q1": {

    "median": 0.48706105499721486,

    "p16": 0.43762092403162756,

    "p84": 0.5337745734682977

    "median": 0.4847008724360507,

    "p16": 0.4346673265283473,

    "p84": 0.5348607580756441

   },

   "q2": {

    "median": 0.3998816247272913,

    "p16": 0.38241711208333434,

    "p84": 0.4165019858735322

    "median": 0.39946098254972023,

    "p16": 0.3814434742534384,

    "p84": 0.4165224642393515

   },

   "rho_star": {

    "median": 1.3057974240287376,

    "p16": 1.246869572604999,

    "p84": 1.3692797179230005

    "median": 1.3047167399294046,

    "p16": 1.2470599525678594,

    "p84": 1.3648213993062783

   },

   "rp_by_rs_b": {

    "median": 0.09001128271601587,

    "p16": 0.08926355162163656,

    "p84": 0.09059072111489394

    "median": 0.08990743059456484,

    "p16": 0.08918478455032833,

    "p84": 0.09056902148613193

   },

   "v_sin_i": {

    "median": 6.925445691006281,

    "p16": 6.719720880540746,

    "p84": 7.143209296918225

    "median": 6.964845909197844,

    "p16": 6.749339109182637,

    "p84": 7.169780990976243

   }

  },

  "device": "gpu",
@@ -95,10 +95,10 @@ 
  "config_sha256": "56f5e7121ec9995b14edfdcfa235d88fc80f03db9672ff57c472cdebdf72d55e",

  "pyaetna_version": "3.8.0",

  "jax_version": "0.11.0",

  "timestamp_utc": "2026-10-02T11:56:43Z",

  "timestamp_utc": "2026-10-04T21:21:07Z",

  "provenance": {

   "pyaetna_version": "3.8.0",

   "git_commit": "7de93d64bae90e17f88ba7e025f8aebd625f2bf2",

   "git_commit": "4296ee362757201004bdea2c9843835db06e2ec7",

   "git_dirty": false,

   "git_diff_sha256": null,

   "python": "3.14.4",
@@ -122,7 +122,7 @@ 
     "memory": "12227 MiB"

    }

   ],

   "compilation_cache_entries": 165011

   "compilation_cache_entries": 186403

  },

  "blackjax_version": "1.6.2"

 },
@@ -134,13 +134,13 @@ 
  "num_chains": 1,

  "num_draws_per_chain": 4000,

  "num_warmup": 0,

  "sampling_elapsed_s": 9.417337417602539,

  "sampling_elapsed_s": 8.797070980072021,

  "random_seed": 1,

  "timestamp_utc": "2026-10-02T11:56:43Z"

  "timestamp_utc": "2026-10-04T21:21:07Z"

 },

 "provenance": {

  "pyaetna_version": "3.8.0",

  "git_commit": "7de93d64bae90e17f88ba7e025f8aebd625f2bf2",

  "git_commit": "4296ee362757201004bdea2c9843835db06e2ec7",

  "git_dirty": false,

  "git_diff_sha256": null,

  "python": "3.14.4",
@@ -164,92 +164,92 @@ 
    "memory": "12227 MiB"

   }

  ],

  "compilation_cache_entries": 165011

  "compilation_cache_entries": 186403

 },

 "parameters": {

  "delta_period_b": {

   "mean": -2.6027966985643547e-06,

   "std": 1.210270676028522e-05,

   "ess": 3648.63068898108,

   "rhat": 1.0000939062114609,

   "ess_per_sec": 387.4376086558388

   "mean": -1.6064150632728923e-06,

   "std": 1.2104875586400467e-05,

   "ess": 4150.461461168109,

   "rhat": 0.999756763254715,

   "ess_per_sec": 471.8003833969439

  },

  "delta_tau_b": {

   "mean": -1.5134573756303806e-05,

   "std": 0.00021467410545373313,

   "ess": 4025.9070016754877,

   "rhat": 0.9998216107624004,

   "ess_per_sec": 427.49949621114894

   "mean": -1.0517706349619697e-05,

   "std": 0.00022030594659791465,

   "ess": 3972.7653437823615,

   "rhat": 0.9997708162511594,

   "ess_per_sec": 451.6009195312684

  },

  "impact_b": {

   "mean": 0.21693604329635927,

   "std": 0.07495403716800378,

   "ess": 4196.23664529764,

   "rhat": 0.9997715454195706,

   "ess_per_sec": 445.58631163137363

   "mean": 0.2190158542614672,

   "std": 0.07378918846189997,

   "ess": 3858.952574913977,

   "rhat": 0.9999705130256871,

   "ess_per_sec": 438.6633441580329

  },

  "kRV_b": {

   "mean": 15.784924312081959,

   "std": 1.4858512362284482,

   "ess": 4038.6659421047507,

   "rhat": 0.9998071181329492,

   "ess_per_sec": 428.8543314329829

   "mean": 15.78764934146664,

   "std": 1.5028453124495968,

   "ess": 4127.915727364721,

   "rhat": 1.0013215228020094,

   "ess_per_sec": 469.2375151588154

  },

  "lambda_spin_b": {

   "mean": -0.018323183925391083,

   "std": 0.15505264904177904,

   "ess": 3670.353232511265,

   "rhat": 0.999834781833506,

   "ess_per_sec": 389.74426313437345

   "mean": 0.012461620681242647,

   "std": 0.14884843678155293,

   "ess": 4190.877700129723,

   "rhat": 0.999771538601084,

   "ess_per_sec": 476.3946670003352

  },

  "log_jitter": {

   "mean": -1.5746831279479028,

   "std": 1.7813495107773774,

   "ess": 3809.552302236049,

   "rhat": 0.999938092645033,

   "ess_per_sec": 404.525412365004

   "mean": -1.446890424816179,

   "std": 1.730558853662311,

   "ess": 3806.202504619091,

   "rhat": 0.999795148194861,

   "ess_per_sec": 432.6670221533133

  },

  "norm": {

   "mean": -9.417718967457844,

   "std": 0.0009807911881579627,

   "ess": 3725.3839930531053,

   "rhat": 0.999763815546046,

   "ess_per_sec": 395.5878214674304

   "mean": -9.417643222854553,

   "std": 0.0009753959138275626,

   "ess": 4325.147662286562,

   "rhat": 0.9998574430652849,

   "ess_per_sec": 491.65769744092165

  },

  "q1": {

   "mean": 0.4870649400299899,

   "std": 0.04842160764985508,

   "ess": 4002.6799880588087,

   "rhat": 0.9997906949157724,

   "ess_per_sec": 425.0330863771693

   "mean": 0.4837134482282904,

   "std": 0.050302763577919005,

   "ess": 4055.228787408568,

   "rhat": 1.0001171628663312,

   "ess_per_sec": 460.9748854584515

  },

  "q2": {

   "mean": 0.39970398841901145,

   "std": 0.017351021237952298,

   "ess": 3862.0751509218658,

   "rhat": 0.9998005772799115,

   "ess_per_sec": 410.102662744463

   "mean": 0.3993078422945057,

   "std": 0.017883257958600286,

   "ess": 4099.416570105668,

   "rhat": 0.9999564677859667,

   "ess_per_sec": 465.9978962761656

  },

  "rho_star": {

   "mean": 1.308298673684167,

   "std": 0.06398773142121812,

   "ess": 3992.620082856063,

   "rhat": 0.9999722678489764,

   "ess_per_sec": 423.96485395046005

   "mean": 1.3063618545819344,

   "std": 0.06279271324088309,

   "ess": 3824.4779949907515,

   "rhat": 1.0004128196869462,

   "ess_per_sec": 434.7444738884488

  },

  "rp_by_rs_b": {

   "mean": 0.08995413029990394,

   "std": 0.0006660751474720803,

   "ess": 4119.272179115296,

   "rhat": 0.9999774687576114,

   "ess_per_sec": 437.41367612205386

   "mean": 0.08989354250165062,

   "std": 0.0006825657678887709,

   "ess": 3810.30577786444,

   "rhat": 1.0001651844607946,

   "ess_per_sec": 433.1334584540598

  },

  "v_sin_i": {

   "mean": 6.934510310238247,

   "std": 0.21195231518690905,

   "ess": 4053.961258180198,

   "rhat": 0.9998181165054028,

   "ess_per_sec": 430.4784970965024

   "mean": 6.96264756837587,

   "std": 0.2057236023485613,

   "ess": 3658.6312369897782,

   "rhat": 0.999749978053342,

   "ess_per_sec": 415.89197646326426

  }

 },

 "fixed_params": {

paper_benchmarks/toi5398_rm_pyorbit/runs/ccf/PyAETNAouts/toi5398_ccf_run.log

+339 −340

File changed.

Preview size limit exceeded, changes collapsed.