penfever commited on
Commit
b713018
·
verified ·
1 Parent(s): 4d69456

Add parsed metric surface and redacted trainer log for run 1ae770

Browse files
.gitattributes CHANGED
@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  tokenizer.json filter=lfs diff=lfs merge=lfs -text
 
 
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
  tokenizer.json filter=lfs diff=lfs merge=lfs -text
37
+ training_logs/reward_plot.png filter=lfs diff=lfs merge=lfs -text
training_logs/1ae770_trainer.log ADDED
The diff for this file is too large to render. See raw diff
 
training_logs/20260730_100022_metrics_finelog.csv ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ async/discard_rate,async/discarded_count,async/effective_batch_groups,async/effective_batch_samples,async/staleness_max,async/staleness_mean,async/staleness_min,async/staleness_ratio,generate/avg_num_tokens,generate/avg_tokens_non_zero_rewards,generate/avg_tokens_zero_rewards,generate/max_num_tokens,generate/min_num_tokens,generate/std_num_tokens,generate/tis/aligned_tokens,generate/tis/alignment_fail_count,generate/tis/exact_match_fraction,generate/tis/lcs_fallback_alert,generate/tis/lcs_fallback_fraction,generate/tis/lcs_fallback_messages,generate/tis/unaligned_fraction,loss/avg_final_rewards,loss/avg_raw_advantages,loss/avg_raw_advantages_abs,policy/final_loss,policy/policy_entropy,policy/policy_loss,policy/policy_lr,policy/policy_update_steps,policy/ppo_clip_ratio,policy/raw_grad_norm,policy/rollout_train_prob_diff_mean,policy/rollout_train_prob_diff_std,reward/avg_pass_at_8,reward/avg_raw_reward,system/process_rss_gb,system/process_vms_gb,system/ram_available_gb,system/ram_percent,system/ram_total_gb,system/ram_used_gb,timing/cleanup_old_checkpoints,timing/compute_advantages_and_returns,timing/convert_to_training_input,timing/fwd_logprobs_values_reward,timing/policy_train,timing/run_training,timing/save_checkpoints,timing/step,timing/sync_weights,timing/train_critic_and_policy,timing/wait_for_generation_buffer,tis/batch_skipped_no_logprobs,tis/skipped_fraction,trainer/epoch,trainer/global_step,timing/save_hf_model
2
+ 0.0,0,64,512,1,0.03125,0,0.03125,2352.162109375,2012.2652173913043,2629.3829787234044,21648,271,2308.2586931329156,580442.0,0.0,1.0,0.0,0.0,0.0,0.0,0.44921875,-0.022331397980451584,0.10573633015155792,-0.0001103550021070987,0.916983806702774,-0.0001103550021070987,7.999999979801942e-06,1,0.0,0.07379782199859619,1.0393970012664795,0.7227853536605835,0.515625,0.44921875,17.535274505615234,120.92271041870117,1872.6025466918945,7.0,2014.4287147521973,141.82616806030273,5.63318657875061,0.0873723030090332,3.595928192138672,69.1614830493927,183.64072036743164,253.0149073600769,479.92912912368774,877.4440121650696,82.811190366745,183.74106550216675,538.0217173099518,0.0,0.0,0,11,
3
+ 0.0,0,64,512,1,1.0,1,1.0,2537.443359375,2154.151515151515,2670.5868421052633,23204,460,2391.8542421188295,621973.0,0.0,1.0,0.0,0.0,0.0,0.0,0.2578125,-0.004582593217492104,0.09677589684724808,-1.4968361938372254e-05,0.9270317238988355,-1.4968361938372254e-05,7.999999979801942e-06,1,0.0,0.07234907150268555,1.3672617673873901,189.98556518554688,0.453125,0.2578125,18.890262603759766,125.78672409057617,1869.3302154541016,7.2,2014.4287147521973,145.0984992980957,0.006937503814697266,0.08305644989013672,3.0499768257141113,46.31576132774353,148.18072295188904,194.67333555221558,463.34481477737427,462.7308204174042,264.97486186027527,148.25221467018127,0.000518798828125,0.0,0.0,0,12,
4
+ 0.0,0,64,512,2,1.828125,1,1.0,3358.01953125,3274.4132653061224,3409.8765822784812,25674,247,3377.537795809031,850578.0,0.0,1.0,0.0,0.0,0.0,0.0,0.3828125,-0.021296631544828415,0.1237059086561203,0.00020468563889153302,0.9662019034731202,0.00020468563889153302,7.999999979801942e-06,1,0.0,0.07480604201555252,32.1727294921875,26229.494140625,0.5,0.3828125,20.376602172851562,130.76318359375,1866.640037536621,7.3,2014.4287147521973,147.78867721557617,0.006958484649658203,0.08489203453063965,3.6398375034332275,48.58984732627869,159.764310836792,208.5389130115509,461.6352093219757,293.3591275215149,81.14270424842834,159.84030389785767,0.0006954669952392578,0.0,0.0,0,13,
5
+ 0.0,0,64,512,2,1.65625,1,1.0,3771.751953125,3321.8522167487686,4067.3171521035597,24369,300,3173.1146787729913,955649.0,0.0,1.0,0.0,0.0,0.0,0.0,0.396484375,-0.02060784213244915,0.13694418966770172,4.18531708419323e-05,1.013929680455476,4.18531708419323e-05,7.999999979801942e-06,1,0.0,0.06541100889444351,11.94090461730957,8152.9208984375,0.53125,0.396484375,21.203872680664062,135.46282958984375,1865.4100379943848,7.4,2014.4287147521973,149.0186767578125,0.009932994842529297,0.07995367050170898,3.4873104095458984,46.4379518032074,152.79619932174683,199.44171118736267,462.82099890708923,312.19322419166565,109.21838808059692,152.8915433883667,0.0005068778991699219,0.0,0.0,0,14,
6
+ 0.0,0,64,512,4,1.265625,0,0.90625,3671.82421875,3321.214659685864,3880.442367601246,21768,376,2500.7308635503846,887909.0,0.0,1.0,0.0,0.0,0.0,0.0,0.373046875,-0.027924705296754837,0.15362447500228882,-0.00017171968647744507,1.001769600668922,-0.00017171968647744507,7.999999979801942e-06,1,0.0,0.0640101507306099,5253.3203125,4885459.0,0.53125,0.373046875,21.476531982421875,140.16432571411133,1864.5351448059082,7.4,2014.4287147521973,149.89356994628906,0.006436586380004883,0.0720057487487793,2.9676759243011475,43.895183086395264,144.5767424106598,188.62816667556763,472.7359070777893,672.1971838474274,216.38741636276245,144.6401011943817,264.1604299545288,0.0,0.0,0,15,35.39909744262695
7
+ 0.0,0,64,512,2,0.734375,0,0.6875,3629.326171875,3374.0632411067195,3878.675675675676,25577,307,2985.3460391870876,875476.0,0.0,1.0,0.0,0.0,0.0,0.0,0.494140625,0.0013031653361395001,0.11755329370498657,-0.0003406114992685616,1.0002137462724932,-0.0003406114992685616,7.999999979801942e-06,1,0.0,0.058628324419260025,1.9664862155914307,573.5465698242188,0.625,0.494140625,21.878799438476562,140.45945739746094,1863.2931518554688,7.5,2014.4287147521973,151.13556289672852,0.0063478946685791016,0.08103275299072266,3.3881232738494873,44.876765966415405,147.02753973007202,192.07576894760132,469.7542881965637,985.7916593551636,82.15019392967224,147.09987020492554,708.1371586322784,0.0,0.0,0,16,
8
+ 0.0,0,64,512,4,0.953125,0,0.890625,3002.22265625,2752.0490797546013,3119.0659025787963,26854,651,2790.015845685145,719255.0,1.0,0.9999944386900335,0.0,0.0,0.0,5.561309966562623e-06,0.318359375,-0.0008369925781153142,0.09486871212720871,-0.00022500994964502752,0.9017434206325561,-0.00022500994964502752,7.999999979801942e-06,1,0.0,0.05405773967504501,1.348465919494629,201.1986083984375,0.46875,0.318359375,21.203689575195312,140.65093612670898,1863.7885856628418,7.5,2014.4287147521973,150.64012908935547,0.0067882537841796875,0.09145879745483398,3.419133186340332,44.97348093986511,147.18843388557434,192.34168219566345,465.55749225616455,428.29968667030334,81.63673996925354,147.25872898101807,150.85458874702454,0.0,0.0,0,17,
9
+ 0.0,0,64,512,2,0.71875,0,0.703125,2478.958984375,2188.1207729468597,2676.3475409836065,23859,319,2411.9658079799956,598075.0,0.0,1.0,0.0,0.0,0.0,0.0,0.404296875,-0.023721639066934586,0.12942376732826233,-8.077389793470502e-06,0.8676014154916629,-8.077389793470502e-06,7.999999979801942e-06,1,0.0,0.07506733387708664,1.1716746091842651,28.08149528503418,0.5625,0.404296875,21.832080841064453,140.57871627807617,1863.176570892334,7.5,2014.4287147521973,151.25214385986328,0.00798487663269043,0.07138228416442871,2.9453682899475098,43.19208216667175,138.68992280960083,182.03820514678955,461.42279839515686,693.5649092197418,81.85265922546387,138.75854396820068,426.6822907924652,0.0,0.0,0,18,
10
+ 0.0,0,64,512,4,0.8125,0,0.765625,2556.279296875,2093.8447488584475,2901.921501706485,20704,473,2699.305868189104,610144.0,0.0,1.0,0.0,0.0,0.0,0.0,0.427734375,-0.029633240774273872,0.14690423011779785,-6.726852734573185e-05,0.8370082812616602,-6.726852734573185e-05,7.999999979801942e-06,1,0.0,0.05800345540046692,1.3862186670303345,100.81414794921875,0.546875,0.427734375,22.089065551757812,149.64979553222656,1862.121425628662,7.6,2014.4287147521973,152.30728912353516,0.00617218017578125,0.06627368927001953,2.776560068130493,43.89715623855591,140.9724726676941,185.03911590576172,460.6599955558777,836.5690820217133,192.18002724647522,141.05319905281067,456.53414034843445,0.0,0.0,0,19,
11
+ 0.0,0,64,512,4,0.609375,0,0.546875,2340.130859375,1941.570446735395,2864.9321266968327,22150,378,2765.9895078407653,583260.0,0.0,1.0,0.0,0.0,0.0,0.0,0.568359375,1.252095080417348e-05,0.06844891607761383,-0.00018067998462356627,0.7792838950990699,-0.00018067998462356627,7.999999979801942e-06,1,0.0,0.043685052543878555,4.070535659790039,1505.0963134765625,0.625,0.568359375,21.034622192382812,149.92840194702148,1863.1474533081055,7.5,2014.4287147521973,151.2812614440918,0.006177186965942383,0.07135415077209473,2.839724063873291,43.34166693687439,138.39206886291504,181.88161945343018,455.8801598548889,1160.3470537662506,81.77262330055237,138.45159482955933,893.8027949333191,0.0,0.0,0,20,38.88026189804077
12
+ 0.0,0,64,512,7,0.5,0,0.40625,2234.611328125,1952.3653846153845,2427.7269736842104,20143,394,2265.943199919693,566081.0,0.0,1.0,0.0,0.0,0.0,0.0,0.40625,-0.019281430169939995,0.11397971957921982,-0.00015442364383488894,0.7739015080878744,-0.00015442364383488894,7.999999979801942e-06,1,0.0,0.06565333157777786,159.1624298095703,88023.6875,0.5,0.40625,21.767887115478516,149.9852638244629,1863.4467658996582,7.5,2014.4287147521973,150.98194885253906,0.006381511688232422,0.07059240341186523,2.5255768299102783,41.81991696357727,135.55581903457642,177.515766620636,459.9064302444458,1965.6425108909607,81.717937707901,135.60997200012207,1703.8462085723877,0.0,0.0,0,21,
training_logs/20260730_100022_metrics_table.csv ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ log_file,async/discard_rate,async/discarded_count,async/effective_batch_groups,async/effective_batch_samples,async/staleness_max,async/staleness_mean,async/staleness_min,async/staleness_ratio,generate/avg_num_tokens,generate/avg_tokens_non_zero_rewards,generate/avg_tokens_zero_rewards,generate/max_num_tokens,generate/min_num_tokens,generate/std_num_tokens,generate/tis/aligned_tokens,generate/tis/alignment_fail_count,generate/tis/exact_match_fraction,generate/tis/lcs_fallback_alert,generate/tis/lcs_fallback_fraction,generate/tis/lcs_fallback_messages,generate/tis/unaligned_fraction,loss/avg_final_rewards,loss/avg_raw_advantages,loss/avg_raw_advantages_abs,policy/final_loss,policy/policy_entropy,policy/policy_loss,policy/policy_lr,policy/policy_update_steps,policy/ppo_clip_ratio,policy/raw_grad_norm,policy/rollout_train_prob_diff_mean,policy/rollout_train_prob_diff_std,reward/avg_pass_at_8,reward/avg_raw_reward,system/process_rss_gb,system/process_vms_gb,system/ram_available_gb,system/ram_percent,system/ram_total_gb,system/ram_used_gb,timing/cleanup_old_checkpoints,timing/compute_advantages_and_returns,timing/convert_to_training_input,timing/fwd_logprobs_values_reward,timing/policy_train,timing/run_training,timing/save_checkpoints,timing/step,timing/sync_weights,timing/train_critic_and_policy,timing/wait_for_generation_buffer,tis/batch_skipped_no_logprobs,tis/skipped_fraction,trainer/epoch,trainer/global_step,timing/save_hf_model,global_step
2
+ finelog,0.0,0,64,512,1,0.03125,0,0.03125,2352.162109375,2012.2652173913043,2629.3829787234044,21648,271,2308.2586931329156,580442.0,0.0,1.0,0.0,0.0,0.0,0.0,0.44921875,-0.022331397980451584,0.10573633015155792,-0.0001103550021070987,0.916983806702774,-0.0001103550021070987,7.999999979801942e-06,1,0.0,0.07379782199859619,1.0393970012664795,0.7227853536605835,0.515625,0.44921875,17.535274505615234,120.92271041870117,1872.6025466918945,7.0,2014.4287147521973,141.82616806030273,5.63318657875061,0.0873723030090332,3.595928192138672,69.1614830493927,183.64072036743164,253.0149073600769,479.92912912368774,877.4440121650696,82.811190366745,183.74106550216675,538.0217173099518,0.0,0.0,0,11,,11
3
+ finelog,0.0,0,64,512,1,1.0,1,1.0,2537.443359375,2154.151515151515,2670.5868421052633,23204,460,2391.8542421188295,621973.0,0.0,1.0,0.0,0.0,0.0,0.0,0.2578125,-0.004582593217492104,0.09677589684724808,-1.4968361938372254e-05,0.9270317238988355,-1.4968361938372254e-05,7.999999979801942e-06,1,0.0,0.07234907150268555,1.3672617673873901,189.98556518554688,0.453125,0.2578125,18.890262603759766,125.78672409057617,1869.3302154541016,7.2,2014.4287147521973,145.0984992980957,0.006937503814697266,0.08305644989013672,3.0499768257141113,46.31576132774353,148.18072295188904,194.67333555221558,463.34481477737427,462.7308204174042,264.97486186027527,148.25221467018127,0.000518798828125,0.0,0.0,0,12,,12
4
+ finelog,0.0,0,64,512,2,1.828125,1,1.0,3358.01953125,3274.4132653061224,3409.8765822784812,25674,247,3377.537795809031,850578.0,0.0,1.0,0.0,0.0,0.0,0.0,0.3828125,-0.021296631544828415,0.1237059086561203,0.00020468563889153302,0.9662019034731202,0.00020468563889153302,7.999999979801942e-06,1,0.0,0.07480604201555252,32.1727294921875,26229.494140625,0.5,0.3828125,20.376602172851562,130.76318359375,1866.640037536621,7.3,2014.4287147521973,147.78867721557617,0.006958484649658203,0.08489203453063965,3.6398375034332275,48.58984732627869,159.764310836792,208.5389130115509,461.6352093219757,293.3591275215149,81.14270424842834,159.84030389785767,0.0006954669952392578,0.0,0.0,0,13,,13
5
+ finelog,0.0,0,64,512,2,1.65625,1,1.0,3771.751953125,3321.8522167487686,4067.3171521035597,24369,300,3173.1146787729913,955649.0,0.0,1.0,0.0,0.0,0.0,0.0,0.396484375,-0.02060784213244915,0.13694418966770172,4.18531708419323e-05,1.013929680455476,4.18531708419323e-05,7.999999979801942e-06,1,0.0,0.06541100889444351,11.94090461730957,8152.9208984375,0.53125,0.396484375,21.203872680664062,135.46282958984375,1865.4100379943848,7.4,2014.4287147521973,149.0186767578125,0.009932994842529297,0.07995367050170898,3.4873104095458984,46.4379518032074,152.79619932174683,199.44171118736267,462.82099890708923,312.19322419166565,109.21838808059692,152.8915433883667,0.0005068778991699219,0.0,0.0,0,14,,14
6
+ finelog,0.0,0,64,512,4,1.265625,0,0.90625,3671.82421875,3321.214659685864,3880.442367601246,21768,376,2500.7308635503846,887909.0,0.0,1.0,0.0,0.0,0.0,0.0,0.373046875,-0.027924705296754837,0.15362447500228882,-0.00017171968647744507,1.001769600668922,-0.00017171968647744507,7.999999979801942e-06,1,0.0,0.0640101507306099,5253.3203125,4885459.0,0.53125,0.373046875,21.476531982421875,140.16432571411133,1864.5351448059082,7.4,2014.4287147521973,149.89356994628906,0.006436586380004883,0.0720057487487793,2.9676759243011475,43.895183086395264,144.5767424106598,188.62816667556763,472.7359070777893,672.1971838474274,216.38741636276245,144.6401011943817,264.1604299545288,0.0,0.0,0,15,35.39909744262695,15
7
+ finelog,0.0,0,64,512,2,0.734375,0,0.6875,3629.326171875,3374.0632411067195,3878.675675675676,25577,307,2985.3460391870876,875476.0,0.0,1.0,0.0,0.0,0.0,0.0,0.494140625,0.0013031653361395001,0.11755329370498657,-0.0003406114992685616,1.0002137462724932,-0.0003406114992685616,7.999999979801942e-06,1,0.0,0.058628324419260025,1.9664862155914307,573.5465698242188,0.625,0.494140625,21.878799438476562,140.45945739746094,1863.2931518554688,7.5,2014.4287147521973,151.13556289672852,0.0063478946685791016,0.08103275299072266,3.3881232738494873,44.876765966415405,147.02753973007202,192.07576894760132,469.7542881965637,985.7916593551636,82.15019392967224,147.09987020492554,708.1371586322784,0.0,0.0,0,16,,16
8
+ finelog,0.0,0,64,512,4,0.953125,0,0.890625,3002.22265625,2752.0490797546013,3119.0659025787963,26854,651,2790.015845685145,719255.0,1.0,0.9999944386900335,0.0,0.0,0.0,5.561309966562623e-06,0.318359375,-0.0008369925781153142,0.09486871212720871,-0.00022500994964502752,0.9017434206325561,-0.00022500994964502752,7.999999979801942e-06,1,0.0,0.05405773967504501,1.348465919494629,201.1986083984375,0.46875,0.318359375,21.203689575195312,140.65093612670898,1863.7885856628418,7.5,2014.4287147521973,150.64012908935547,0.0067882537841796875,0.09145879745483398,3.419133186340332,44.97348093986511,147.18843388557434,192.34168219566345,465.55749225616455,428.29968667030334,81.63673996925354,147.25872898101807,150.85458874702454,0.0,0.0,0,17,,17
9
+ finelog,0.0,0,64,512,2,0.71875,0,0.703125,2478.958984375,2188.1207729468597,2676.3475409836065,23859,319,2411.9658079799956,598075.0,0.0,1.0,0.0,0.0,0.0,0.0,0.404296875,-0.023721639066934586,0.12942376732826233,-8.077389793470502e-06,0.8676014154916629,-8.077389793470502e-06,7.999999979801942e-06,1,0.0,0.07506733387708664,1.1716746091842651,28.08149528503418,0.5625,0.404296875,21.832080841064453,140.57871627807617,1863.176570892334,7.5,2014.4287147521973,151.25214385986328,0.00798487663269043,0.07138228416442871,2.9453682899475098,43.19208216667175,138.68992280960083,182.03820514678955,461.42279839515686,693.5649092197418,81.85265922546387,138.75854396820068,426.6822907924652,0.0,0.0,0,18,,18
10
+ finelog,0.0,0,64,512,4,0.8125,0,0.765625,2556.279296875,2093.8447488584475,2901.921501706485,20704,473,2699.305868189104,610144.0,0.0,1.0,0.0,0.0,0.0,0.0,0.427734375,-0.029633240774273872,0.14690423011779785,-6.726852734573185e-05,0.8370082812616602,-6.726852734573185e-05,7.999999979801942e-06,1,0.0,0.05800345540046692,1.3862186670303345,100.81414794921875,0.546875,0.427734375,22.089065551757812,149.64979553222656,1862.121425628662,7.6,2014.4287147521973,152.30728912353516,0.00617218017578125,0.06627368927001953,2.776560068130493,43.89715623855591,140.9724726676941,185.03911590576172,460.6599955558777,836.5690820217133,192.18002724647522,141.05319905281067,456.53414034843445,0.0,0.0,0,19,,19
11
+ finelog,0.0,0,64,512,4,0.609375,0,0.546875,2340.130859375,1941.570446735395,2864.9321266968327,22150,378,2765.9895078407653,583260.0,0.0,1.0,0.0,0.0,0.0,0.0,0.568359375,1.252095080417348e-05,0.06844891607761383,-0.00018067998462356627,0.7792838950990699,-0.00018067998462356627,7.999999979801942e-06,1,0.0,0.043685052543878555,4.070535659790039,1505.0963134765625,0.625,0.568359375,21.034622192382812,149.92840194702148,1863.1474533081055,7.5,2014.4287147521973,151.2812614440918,0.006177186965942383,0.07135415077209473,2.839724063873291,43.34166693687439,138.39206886291504,181.88161945343018,455.8801598548889,1160.3470537662506,81.77262330055237,138.45159482955933,893.8027949333191,0.0,0.0,0,20,38.88026189804077,20
12
+ finelog,0.0,0,64,512,7,0.5,0,0.40625,2234.611328125,1952.3653846153845,2427.7269736842104,20143,394,2265.943199919693,566081.0,0.0,1.0,0.0,0.0,0.0,0.0,0.40625,-0.019281430169939995,0.11397971957921982,-0.00015442364383488894,0.7739015080878744,-0.00015442364383488894,7.999999979801942e-06,1,0.0,0.06565333157777786,159.1624298095703,88023.6875,0.5,0.40625,21.767887115478516,149.9852638244629,1863.4467658996582,7.5,2014.4287147521973,150.98194885253906,0.006381511688232422,0.07059240341186523,2.5255768299102783,41.81991696357727,135.55581903457642,177.515766620636,459.9064302444458,1965.6425108909607,81.717937707901,135.60997200012207,1703.8462085723877,0.0,0.0,0,21,,21
training_logs/20260730_100022_vllm_metrics_finelog.csv ADDED
The diff for this file is too large to render. See raw diff
 
training_logs/20260730_100022_vllm_metrics_table.csv ADDED
The diff for this file is too large to render. See raw diff
 
training_logs/metrics.csv ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ log_file,async/discard_rate,async/discarded_count,async/effective_batch_groups,async/effective_batch_samples,async/staleness_max,async/staleness_mean,async/staleness_min,async/staleness_ratio,generate/avg_num_tokens,generate/avg_tokens_non_zero_rewards,generate/avg_tokens_zero_rewards,generate/max_num_tokens,generate/min_num_tokens,generate/std_num_tokens,generate/tis/aligned_tokens,generate/tis/alignment_fail_count,generate/tis/exact_match_fraction,generate/tis/lcs_fallback_alert,generate/tis/lcs_fallback_fraction,generate/tis/lcs_fallback_messages,generate/tis/unaligned_fraction,loss/avg_final_rewards,loss/avg_raw_advantages,loss/avg_raw_advantages_abs,policy/final_loss,policy/policy_entropy,policy/policy_loss,policy/policy_lr,policy/policy_update_steps,policy/ppo_clip_ratio,policy/raw_grad_norm,policy/rollout_train_prob_diff_mean,policy/rollout_train_prob_diff_std,reward/avg_pass_at_8,reward/avg_raw_reward,system/process_rss_gb,system/process_vms_gb,system/ram_available_gb,system/ram_percent,system/ram_total_gb,system/ram_used_gb,timing/cleanup_old_checkpoints,timing/compute_advantages_and_returns,timing/convert_to_training_input,timing/fwd_logprobs_values_reward,timing/policy_train,timing/run_training,timing/save_checkpoints,timing/step,timing/sync_weights,timing/train_critic_and_policy,timing/wait_for_generation_buffer,tis/batch_skipped_no_logprobs,tis/skipped_fraction,trainer/epoch,trainer/global_step,timing/save_hf_model,global_step
2
+ finelog,0.0,0,64,512,1,0.03125,0,0.03125,2352.162109375,2012.2652173913043,2629.3829787234044,21648,271,2308.2586931329156,580442.0,0.0,1.0,0.0,0.0,0.0,0.0,0.44921875,-0.022331397980451584,0.10573633015155792,-0.0001103550021070987,0.916983806702774,-0.0001103550021070987,7.999999979801942e-06,1,0.0,0.07379782199859619,1.0393970012664795,0.7227853536605835,0.515625,0.44921875,17.535274505615234,120.92271041870117,1872.6025466918945,7.0,2014.4287147521973,141.82616806030273,5.63318657875061,0.0873723030090332,3.595928192138672,69.1614830493927,183.64072036743164,253.0149073600769,479.92912912368774,877.4440121650696,82.811190366745,183.74106550216675,538.0217173099518,0.0,0.0,0,11,,11
3
+ finelog,0.0,0,64,512,1,1.0,1,1.0,2537.443359375,2154.151515151515,2670.5868421052633,23204,460,2391.8542421188295,621973.0,0.0,1.0,0.0,0.0,0.0,0.0,0.2578125,-0.004582593217492104,0.09677589684724808,-1.4968361938372254e-05,0.9270317238988355,-1.4968361938372254e-05,7.999999979801942e-06,1,0.0,0.07234907150268555,1.3672617673873901,189.98556518554688,0.453125,0.2578125,18.890262603759766,125.78672409057617,1869.3302154541016,7.2,2014.4287147521973,145.0984992980957,0.006937503814697266,0.08305644989013672,3.0499768257141113,46.31576132774353,148.18072295188904,194.67333555221558,463.34481477737427,462.7308204174042,264.97486186027527,148.25221467018127,0.000518798828125,0.0,0.0,0,12,,12
4
+ finelog,0.0,0,64,512,2,1.828125,1,1.0,3358.01953125,3274.4132653061224,3409.8765822784812,25674,247,3377.537795809031,850578.0,0.0,1.0,0.0,0.0,0.0,0.0,0.3828125,-0.021296631544828415,0.1237059086561203,0.00020468563889153302,0.9662019034731202,0.00020468563889153302,7.999999979801942e-06,1,0.0,0.07480604201555252,32.1727294921875,26229.494140625,0.5,0.3828125,20.376602172851562,130.76318359375,1866.640037536621,7.3,2014.4287147521973,147.78867721557617,0.006958484649658203,0.08489203453063965,3.6398375034332275,48.58984732627869,159.764310836792,208.5389130115509,461.6352093219757,293.3591275215149,81.14270424842834,159.84030389785767,0.0006954669952392578,0.0,0.0,0,13,,13
5
+ finelog,0.0,0,64,512,2,1.65625,1,1.0,3771.751953125,3321.8522167487686,4067.3171521035597,24369,300,3173.1146787729913,955649.0,0.0,1.0,0.0,0.0,0.0,0.0,0.396484375,-0.02060784213244915,0.13694418966770172,4.18531708419323e-05,1.013929680455476,4.18531708419323e-05,7.999999979801942e-06,1,0.0,0.06541100889444351,11.94090461730957,8152.9208984375,0.53125,0.396484375,21.203872680664062,135.46282958984375,1865.4100379943848,7.4,2014.4287147521973,149.0186767578125,0.009932994842529297,0.07995367050170898,3.4873104095458984,46.4379518032074,152.79619932174683,199.44171118736267,462.82099890708923,312.19322419166565,109.21838808059692,152.8915433883667,0.0005068778991699219,0.0,0.0,0,14,,14
6
+ finelog,0.0,0,64,512,4,1.265625,0,0.90625,3671.82421875,3321.214659685864,3880.442367601246,21768,376,2500.7308635503846,887909.0,0.0,1.0,0.0,0.0,0.0,0.0,0.373046875,-0.027924705296754837,0.15362447500228882,-0.00017171968647744507,1.001769600668922,-0.00017171968647744507,7.999999979801942e-06,1,0.0,0.0640101507306099,5253.3203125,4885459.0,0.53125,0.373046875,21.476531982421875,140.16432571411133,1864.5351448059082,7.4,2014.4287147521973,149.89356994628906,0.006436586380004883,0.0720057487487793,2.9676759243011475,43.895183086395264,144.5767424106598,188.62816667556763,472.7359070777893,672.1971838474274,216.38741636276245,144.6401011943817,264.1604299545288,0.0,0.0,0,15,35.39909744262695,15
7
+ finelog,0.0,0,64,512,2,0.734375,0,0.6875,3629.326171875,3374.0632411067195,3878.675675675676,25577,307,2985.3460391870876,875476.0,0.0,1.0,0.0,0.0,0.0,0.0,0.494140625,0.0013031653361395001,0.11755329370498657,-0.0003406114992685616,1.0002137462724932,-0.0003406114992685616,7.999999979801942e-06,1,0.0,0.058628324419260025,1.9664862155914307,573.5465698242188,0.625,0.494140625,21.878799438476562,140.45945739746094,1863.2931518554688,7.5,2014.4287147521973,151.13556289672852,0.0063478946685791016,0.08103275299072266,3.3881232738494873,44.876765966415405,147.02753973007202,192.07576894760132,469.7542881965637,985.7916593551636,82.15019392967224,147.09987020492554,708.1371586322784,0.0,0.0,0,16,,16
8
+ finelog,0.0,0,64,512,4,0.953125,0,0.890625,3002.22265625,2752.0490797546013,3119.0659025787963,26854,651,2790.015845685145,719255.0,1.0,0.9999944386900335,0.0,0.0,0.0,5.561309966562623e-06,0.318359375,-0.0008369925781153142,0.09486871212720871,-0.00022500994964502752,0.9017434206325561,-0.00022500994964502752,7.999999979801942e-06,1,0.0,0.05405773967504501,1.348465919494629,201.1986083984375,0.46875,0.318359375,21.203689575195312,140.65093612670898,1863.7885856628418,7.5,2014.4287147521973,150.64012908935547,0.0067882537841796875,0.09145879745483398,3.419133186340332,44.97348093986511,147.18843388557434,192.34168219566345,465.55749225616455,428.29968667030334,81.63673996925354,147.25872898101807,150.85458874702454,0.0,0.0,0,17,,17
9
+ finelog,0.0,0,64,512,2,0.71875,0,0.703125,2478.958984375,2188.1207729468597,2676.3475409836065,23859,319,2411.9658079799956,598075.0,0.0,1.0,0.0,0.0,0.0,0.0,0.404296875,-0.023721639066934586,0.12942376732826233,-8.077389793470502e-06,0.8676014154916629,-8.077389793470502e-06,7.999999979801942e-06,1,0.0,0.07506733387708664,1.1716746091842651,28.08149528503418,0.5625,0.404296875,21.832080841064453,140.57871627807617,1863.176570892334,7.5,2014.4287147521973,151.25214385986328,0.00798487663269043,0.07138228416442871,2.9453682899475098,43.19208216667175,138.68992280960083,182.03820514678955,461.42279839515686,693.5649092197418,81.85265922546387,138.75854396820068,426.6822907924652,0.0,0.0,0,18,,18
10
+ finelog,0.0,0,64,512,4,0.8125,0,0.765625,2556.279296875,2093.8447488584475,2901.921501706485,20704,473,2699.305868189104,610144.0,0.0,1.0,0.0,0.0,0.0,0.0,0.427734375,-0.029633240774273872,0.14690423011779785,-6.726852734573185e-05,0.8370082812616602,-6.726852734573185e-05,7.999999979801942e-06,1,0.0,0.05800345540046692,1.3862186670303345,100.81414794921875,0.546875,0.427734375,22.089065551757812,149.64979553222656,1862.121425628662,7.6,2014.4287147521973,152.30728912353516,0.00617218017578125,0.06627368927001953,2.776560068130493,43.89715623855591,140.9724726676941,185.03911590576172,460.6599955558777,836.5690820217133,192.18002724647522,141.05319905281067,456.53414034843445,0.0,0.0,0,19,,19
11
+ finelog,0.0,0,64,512,4,0.609375,0,0.546875,2340.130859375,1941.570446735395,2864.9321266968327,22150,378,2765.9895078407653,583260.0,0.0,1.0,0.0,0.0,0.0,0.0,0.568359375,1.252095080417348e-05,0.06844891607761383,-0.00018067998462356627,0.7792838950990699,-0.00018067998462356627,7.999999979801942e-06,1,0.0,0.043685052543878555,4.070535659790039,1505.0963134765625,0.625,0.568359375,21.034622192382812,149.92840194702148,1863.1474533081055,7.5,2014.4287147521973,151.2812614440918,0.006177186965942383,0.07135415077209473,2.839724063873291,43.34166693687439,138.39206886291504,181.88161945343018,455.8801598548889,1160.3470537662506,81.77262330055237,138.45159482955933,893.8027949333191,0.0,0.0,0,20,38.88026189804077,20
12
+ finelog,0.0,0,64,512,7,0.5,0,0.40625,2234.611328125,1952.3653846153845,2427.7269736842104,20143,394,2265.943199919693,566081.0,0.0,1.0,0.0,0.0,0.0,0.0,0.40625,-0.019281430169939995,0.11397971957921982,-0.00015442364383488894,0.7739015080878744,-0.00015442364383488894,7.999999979801942e-06,1,0.0,0.06565333157777786,159.1624298095703,88023.6875,0.5,0.40625,21.767887115478516,149.9852638244629,1863.4467658996582,7.5,2014.4287147521973,150.98194885253906,0.006381511688232422,0.07059240341186523,2.5255768299102783,41.81991696357727,135.55581903457642,177.515766620636,459.9064302444458,1965.6425108909607,81.717937707901,135.60997200012207,1703.8462085723877,0.0,0.0,0,21,,21
training_logs/report.md ADDED
@@ -0,0 +1,157 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SkyRL Standard-GRPO Training Metrics Analysis
2
+
3
+ Generated from 1 log file(s) (`--format standard`)
4
+
5
+ ## Overview
6
+
7
+ | Log File | Max Step | Train Steps | Final Reward | Max Reward | Final Entropy |
8
+ |----------|----------|-------------|--------------|------------|---------------|
9
+ | finelog | 21 | 11 | 0.4062 | 0.5684 | 0.7739 |
10
+
11
+ ## Async Metrics
12
+
13
+ | | Mean | Std | Min | Max | Count |
14
+ |:------------------------------|-----------:|---------:|----------:|----------:|--------:|
15
+ | async/discard_rate | 0 | 0 | 0 | 0 | 11 |
16
+ | async/discarded_count | 0 | 0 | 0 | 0 | 11 |
17
+ | async/effective_batch_groups | 64 | 0 | 64 | 64 | 11 |
18
+ | async/effective_batch_samples | 512 | 0 | 512 | 512 | 11 |
19
+ | async/staleness_max | 3 | 1.78885 | 1 | 7 | 11 |
20
+ | async/staleness_mean | 0.919034 | 0.514099 | 0.03125 | 1.82812 | 11 |
21
+ | async/staleness_min | 0.272727 | 0.467099 | 0 | 1 | 11 |
22
+ | async/staleness_ratio | 0.721591 | 0.300863 | 0.03125 | 1 | 11 |
23
+
24
+ ## Generate Metrics
25
+
26
+ | | Mean | Std | Min | Max | Count |
27
+ |:-------------------------------------|-----------------:|----------------:|--------------:|-----------------:|--------:|
28
+ | generate/avg_num_tokens | 2902.98 | 599.241 | 2234.61 | 3771.75 | 11 |
29
+ | generate/avg_tokens_non_zero_rewards | 2580.54 | 627.47 | 1941.57 | 3374.06 | 11 |
30
+ | generate/avg_tokens_zero_rewards | 3138.75 | 579.896 | 2427.73 | 4067.32 | 11 |
31
+ | generate/max_num_tokens | 23268.2 | 2191.1 | 20143 | 26854 | 11 |
32
+ | generate/min_num_tokens | 379.636 | 115.639 | 247 | 651 | 11 |
33
+ | generate/std_num_tokens | 2697.28 | 365.539 | 2265.94 | 3377.54 | 11 |
34
+ | generate/tis/aligned_tokens | 713531 | 149297 | 566081 | 955649 | 11 |
35
+ | generate/tis/alignment_fail_count | 0.0909091 | 0.301511 | 0 | 1 | 11 |
36
+ | generate/tis/exact_match_fraction | 0.999999 | 1.6768e-06 | 0.999994 | 1 | 11 |
37
+ | generate/tis/lcs_fallback_alert | 0 | 0 | 0 | 0 | 11 |
38
+ | generate/tis/lcs_fallback_fraction | 0 | 0 | 0 | 0 | 11 |
39
+ | generate/tis/lcs_fallback_messages | 0 | 0 | 0 | 0 | 11 |
40
+ | generate/tis/unaligned_fraction | 5.05574e-07 | 1.6768e-06 | 0 | 5.56131e-06 | 11 |
41
+
42
+ ## Loss Metrics
43
+
44
+ | | Mean | Std | Min | Max | Count |
45
+ |:----------------------------|-----------:|----------:|-----------:|-----------:|--------:|
46
+ | loss/avg_final_rewards | 0.407138 | 0.0823485 | 0.257812 | 0.568359 | 11 |
47
+ | loss/avg_raw_advantages | -0.0153546 | 0.0118293 | -0.0296332 | 0.00130317 | 11 |
48
+ | loss/avg_raw_advantages_abs | 0.117088 | 0.0249498 | 0.0684489 | 0.153624 | 11 |
49
+
50
+ ## Policy Metrics
51
+
52
+ | | Mean | Std | Min | Max | Count |
53
+ |:------------------------------------|----------------:|---------------:|-------------:|---------------:|--------:|
54
+ | policy/final_loss | -9.3325e-05 | 0.000147465 | -0.000340611 | 0.000204686 | 11 |
55
+ | policy/policy_entropy | 0.907788 | 0.0857953 | 0.773902 | 1.01393 | 11 |
56
+ | policy/policy_loss | -9.3325e-05 | 0.000147465 | -0.000340611 | 0.000204686 | 11 |
57
+ | policy/policy_lr | 8e-06 | 0 | 8e-06 | 8e-06 | 11 |
58
+ | policy/policy_update_steps | 1 | 0 | 1 | 1 | 11 |
59
+ | policy/ppo_clip_ratio | 0 | 0 | 0 | 0 | 11 |
60
+ | policy/raw_grad_norm | 0.0641336 | 0.00992899 | 0.0436851 | 0.0750673 | 11 |
61
+ | policy/rollout_train_prob_diff_mean | 497.177 | 1578.13 | 1.0394 | 5253.32 | 11 |
62
+ | policy/rollout_train_prob_diff_std | 455497 | 1.46949e+06 | 0.722785 | 4.88546e+06 | 11 |
63
+
64
+ ## Reward Metrics
65
+
66
+ | | Mean | Std | Min | Max | Count |
67
+ |:----------------------|---------:|----------:|---------:|---------:|--------:|
68
+ | reward/avg_pass_at_8 | 0.53267 | 0.055663 | 0.453125 | 0.625 | 11 |
69
+ | reward/avg_raw_reward | 0.407138 | 0.0823485 | 0.257812 | 0.568359 | 11 |
70
+
71
+ ## System Metrics
72
+
73
+ | | Mean | Std | Min | Max | Count |
74
+ |:------------------------|----------:|---------:|----------:|----------:|--------:|
75
+ | system/process_rss_gb | 20.8444 | 1.41878 | 17.5353 | 22.0891 | 11 |
76
+ | system/process_vms_gb | 138.577 | 9.70567 | 120.923 | 149.985 | 11 |
77
+ | system/ram_available_gb | 1865.23 | 3.1764 | 1862.12 | 1872.6 | 11 |
78
+ | system/ram_percent | 7.4 | 0.173205 | 7 | 7.6 | 11 |
79
+ | system/ram_total_gb | 2014.43 | 0 | 2014.43 | 2014.43 | 11 |
80
+ | system/ram_used_gb | 149.202 | 3.1764 | 141.826 | 152.307 | 11 |
81
+
82
+ ## Timing Metrics
83
+
84
+ | | Mean | Std | Min | Max | Count |
85
+ |:--------------------------------------|------------:|-------------:|--------------:|-------------:|--------:|
86
+ | timing/cleanup_old_checkpoints | 0.518482 | 1.69636 | 0.00617218 | 5.63319 | 11 |
87
+ | timing/compute_advantages_and_returns | 0.0781249 | 0.00819244 | 0.0662737 | 0.0914588 | 11 |
88
+ | timing/convert_to_training_input | 3.14866 | 0.373154 | 2.52558 | 3.63984 | 11 |
89
+ | timing/fwd_logprobs_values_reward | 46.9547 | 7.59618 | 41.8199 | 69.1615 | 11 |
90
+ | timing/policy_train | 148.799 | 13.4953 | 135.556 | 183.641 | 11 |
91
+ | timing/run_training | 195.926 | 20.8822 | 177.516 | 253.015 | 11 |
92
+ | timing/save_checkpoints | 464.877 | 6.8193 | 455.88 | 479.929 | 11 |
93
+ | timing/step | 789.831 | 479.948 | 293.359 | 1965.64 | 11 |
94
+ | timing/sync_weights | 123.259 | 67.5995 | 81.1427 | 264.975 | 11 |
95
+ | timing/train_critic_and_policy | 148.872 | 13.5064 | 135.61 | 183.741 | 11 |
96
+ | timing/wait_for_generation_buffer | 467.458 | 507.172 | 0.000506878 | 1703.85 | 11 |
97
+ | timing/save_hf_model | 37.1397 | 2.46155 | 35.3991 | 38.8803 | 2 |
98
+
99
+ ## Tis Metrics
100
+
101
+ | | Mean | Std | Min | Max | Count |
102
+ |:------------------------------|-------:|------:|------:|------:|--------:|
103
+ | tis/batch_skipped_no_logprobs | 0 | 0 | 0 | 0 | 11 |
104
+ | tis/skipped_fraction | 0 | 0 | 0 | 0 | 11 |
105
+
106
+ ## Trainer Metrics
107
+
108
+ | | Mean | Std | Min | Max | Count |
109
+ |:--------------------|-------:|--------:|------:|------:|--------:|
110
+ | trainer/epoch | 0 | 0 | 0 | 0 | 11 |
111
+ | trainer/global_step | 16 | 3.31662 | 11 | 21 | 11 |
112
+
113
+ ## Training Progression (collapse signals)
114
+
115
+ ### finelog
116
+
117
+ | Step | Epoch | Reward | avg_pass_at_8 | Entropy | GradNorm | PPOClip | PolicyLoss | logRatio_mean | Step Time (s) |
118
+ |------|-------|--------|--------|---------|----------|---------|------------|---------------|---------------|
119
+ | 11 | 0 | 0.4492 | 0.5156 | 0.9170 | 0.0738 | 0.00000 | -0.00011 | nan | 877.4 |
120
+ | 12 | 0 | 0.2578 | 0.4531 | 0.9270 | 0.0723 | 0.00000 | -0.00001 | nan | 462.7 |
121
+ | 13 | 0 | 0.3828 | 0.5000 | 0.9662 | 0.0748 | 0.00000 | 0.00020 | nan | 293.4 |
122
+ | 14 | 0 | 0.3965 | 0.5312 | 1.0139 | 0.0654 | 0.00000 | 0.00004 | nan | 312.2 |
123
+ | 15 | 0 | 0.3730 | 0.5312 | 1.0018 | 0.0640 | 0.00000 | -0.00017 | nan | 672.2 |
124
+ | 16 | 0 | 0.4941 | 0.6250 | 1.0002 | 0.0586 | 0.00000 | -0.00034 | nan | 985.8 |
125
+ | 17 | 0 | 0.3184 | 0.4688 | 0.9017 | 0.0541 | 0.00000 | -0.00023 | nan | 428.3 |
126
+ | 18 | 0 | 0.4043 | 0.5625 | 0.8676 | 0.0751 | 0.00000 | -0.00001 | nan | 693.6 |
127
+ | 19 | 0 | 0.4277 | 0.5469 | 0.8370 | 0.0580 | 0.00000 | -0.00007 | nan | 836.6 |
128
+ | 20 | 0 | 0.5684 | 0.6250 | 0.7793 | 0.0437 | 0.00000 | -0.00018 | nan | 1160.3 |
129
+ | 21 | 0 | 0.4062 | 0.5000 | 0.7739 | 0.0657 | 0.00000 | -0.00015 | nan | 1965.6 |
130
+
131
+ ## vLLM Inference Engine Analysis (per-engine)
132
+
133
+ | Log | Avg Running/Engine | Avg Gen Throughput/Engine | Avg KV Cache % | Avg Prefix Hit % |
134
+ |-----|-------------------|--------------------------|----------------|------------------|
135
+ | finelog | 2.0 | 232.1 tok/s | 0.7% | 90.2% |
136
+
137
+ ## Best Checkpoint (trailing-5 EMA of reward/avg_raw_reward)
138
+
139
+ No step chosen: No saved-aligned checkpoint eligible (after cap + exports intersection). available_exports=[], cap_step=None, save_every=20
140
+
141
+ - available exports: `[]`
142
+ - cap (latest_ckpt_global_step.txt): `None`
143
+
144
+ | Step | Reward | EMA | Eligible |
145
+ |------|--------|-----|----------|
146
+ | 11 | 0.4492 | 0.4492 | |
147
+ | 12 | 0.2578 | 0.3854 | |
148
+ | 13 | 0.3828 | 0.3845 | |
149
+ | 14 | 0.3965 | 0.3885 | |
150
+ | 15 | 0.3730 | 0.3834 | |
151
+ | 16 | 0.4941 | 0.4203 | |
152
+ | 17 | 0.3184 | 0.3863 | |
153
+ | 18 | 0.4043 | 0.3923 | |
154
+ | 19 | 0.4277 | 0.4041 | |
155
+ | 20 | 0.5684 | 0.4589 | |
156
+ | 21 | 0.4062 | 0.4413 | |
157
+
training_logs/reward_plot.png ADDED

Git LFS Details

  • SHA256: 98d88bd579f64703945e2999e3db73a2b679b31c13ef9764bec88e2cb23fa492
  • Pointer size: 131 Bytes
  • Size of remote file: 158 kB
training_logs/vllm_metrics.csv ADDED
The diff for this file is too large to render. See raw diff