1 line
9.8 KiB
JSON
1 line
9.8 KiB
JSON
{"timing_s/agent_loop/tool_calls/min":0,"timing_s/agent_loop/tool_calls/mean":0,"actor/kl_loss":0,"val-aux/tooluse/reward/worst@2/mean":0.5376176470588235,"timing_s/adv":0.1153049636632204,"val-aux/tooluse/score/std@16":0.07216799159273436,"val-aux/tooluse/score/maj@2/std":0.07217664475658733,"val-aux/tooluse/incorrect_format/worst@8/std":0,"val-aux/tooluse/acc/worst@2/std":0.065873083983826,"timing_s/save_checkpoint":6.70261013135314,"val-aux/tooluse/incorrect_format/maj@16/std":0,"rollout_corr/rollout_is_min":0.6138291954994202,"critic/returns/max":0.875,"critic/score/max":1,"critic/score/mean":0.19140625,"val-aux/tooluse/reward/mean@16":0.5689338235294118,"response_length/mean":78.55859375,"critic/rewards/min":0,"val-aux/tooluse/acc/std@16":0.07216799159273436,"timing_s/reward":0.07088929787278175,"val-aux/tooluse/acc/maj@2/std":0.07217664475658733,"rollout_corr/rollout_is_seq_fraction_high":0,"timing_s/step":20.5681207254529,"val-core/tooluse/acc/mean@16":0.5689338235294118,"critic/score/min":0,"val-aux/tooluse/incorrect_format/best@8/mean":0,"timing_s/agent_loop/generate_sequences/max":2.4420138355344534,"training/rollout_probs_diff_max":0.12089437246322632,"val-aux/tooluse/reward/maj@4/std":0.06112092824371511,"_step":100,"timing_s/testing":19.794103000313044,"val-aux/tooluse/acc/best@4/mean":0.6216029411764705,"num_turns/min":2,"val-aux/tooluse/acc/worst@8/std":0.03169698340721522,"actor/pg_clipfrac_lower":0,"val-aux/tooluse/reward/maj@4/mean":0.571764705882353,"rollout_corr/rollout_is_mean":0.9998654127120972,"val-aux/tooluse/score/maj@8/mean":0.5750294117647058,"val-aux/tooluse/score/worst@16/mean":0.47960294117647057,"timing_per_token_ms/update_actor":0.058253035554515414,"perf/max_memory_allocated_gb":124.25863409042358,"val-aux/tooluse/score/worst@4/mean":0.5104117647058823,"timing_per_token_ms/adv":0.0006851685731115321,"global_seqlen/min":18523,"response_length/max":403,"val-core/tooluse/acc/maj@16/mean":0.5770294117647059,"val-aux/tooluse/score/best@16/std":0.011314008550944105,"val-aux/tooluse/acc/best@4/std":0.03612363935746805,"rollout_corr/rollout_is_seq_max_deviation":0.01654648780822754,"rollout_corr/rollout_is_ratio_fraction_low":0,"val-aux/tooluse/score/best@16/mean":0.6409558823529412,"val-aux/tooluse/score/maj@8/std":0.04699356851985658,"val-aux/tooluse/reward/worst@2/std":0.065873083983826,"perf/max_memory_reserved_gb":130.046875,"training/rollout_probs_diff_mean":0.002164518227800727,"timing_s/agent_loop/tool_calls/max":0,"val-aux/tooluse/acc/maj@4/std":0.06112092824371511,"training/rollout_actor_probs_pearson_corr":0.9992616772651672,"val-aux/tooluse/score/maj@4/std":0.06112092824371511,"val-aux/tooluse/incorrect_format/maj@2/std":0,"training/rollout_probs_diff_valid":1,"perf/time_per_step":20.5681207254529,"rollout_corr/chi2_token":0.0007952849846333265,"val-aux/tooluse/reward/worst@4/mean":0.5104117647058823,"timing_s/start_profile":0.00012426264584064484,"perf/mfu/actor_infer":0,"val-aux/tooluse/incorrect_format/mean@16":0,"timing_s/agent_loop/generate_sequences/min":0.25598333589732647,"val-aux/tooluse/score/best@8/std":0.021009683207312644,"val-aux/tooluse/score/best@8/mean":0.6337205882352942,"timing_per_token_ms/gen":0.40311589798174624,"rollout_corr/rollout_is_seq_fraction_low":0,"val-aux/tooluse/score/maj@4/mean":0.571764705882353,"val-aux/tooluse/acc/best@2/std":0.05681896001528811,"val-aux/tooluse/score/best@4/std":0.03612363935746805,"val-aux/tooluse/reward/maj@8/mean":0.5750294117647058,"global_seqlen/mean":21035.875,"prompt_length/mean":578.8125,"val-aux/tooluse/reward/maj@8/std":0.04699356851985658,"val-aux/tooluse/incorrect_format/std@16":0,"val-aux/tooluse/reward/best@2/mean":0.599735294117647,"val-aux/tooluse/reward/worst@8/mean":0.4913382352941176,"val-aux/tooluse/incorrect_format/worst@16/mean":0,"val-aux/tooluse/acc/worst@16/std":0.018056061970488427,"critic/advantages/mean":0.0050159115344285965,"val-aux/tooluse/reward/maj@2/std":0.07217664475658733,"val-aux/tooluse/incorrect_format/best@16/std":0,"val-aux/tooluse/incorrect_format/maj@8/std":0,"val-aux/tooluse/reward/best@16/std":0.011314008550944105,"timing_s/agent_loop/slowest/prompt_length":533,"timing_s/update_actor":9.803228594362736,"rollout_corr/training_ppl":1.1511951768770814,"val-aux/tooluse/incorrect_format/maj@8/mean":0,"response_length/min":22,"prompt_length/clip_ratio":0,"rollout_corr/rollout_is_seq_min":0.9906290769577026,"val-aux/num_turns/mean":2,"val-aux/num_turns/min":2,"perf/throughput":1022.7417118360385,"actor/pg_loss":-0.00013048574328422546,"rollout_corr/training_log_ppl":0.1364561979426071,"rollout_corr/k3_kl":0.00043782229743827283,"val-aux/tooluse/acc/best@8/mean":0.6337205882352942,"timing_s/stop_profile":7.58729875087738e-05,"timing_s/gen":8.107063824310899,"num_turns/mean":2,"training/epoch":0,"val-aux/tooluse/acc/best@2/mean":0.599735294117647,"val-aux/tooluse/reward/worst@4/std":0.050276642601264465,"critic/returns/min":-0.75,"val-aux/tooluse/incorrect_format/best@8/std":0,"critic/rewards/mean":0.19140625,"timing_s/agent_loop/slowest/generate_sequences":2.4420138355344534,"_timestamp":1.7830071043161972e+09,"val-aux/tooluse/incorrect_format/maj@4/std":0,"timing_s/old_log_prob":2.385152492672205,"val-aux/tooluse/score/worst@8/mean":0.4913382352941176,"val-aux/tooluse/reward/best@8/mean":0.6337205882352942,"val-aux/tooluse/incorrect_format/worst@8/mean":0,"val-aux/tooluse/score/best@4/mean":0.6216029411764705,"val-aux/tooluse/incorrect_format/maj@2/mean":0,"critic/rewards/max":1,"critic/advantages/max":0.875,"timing_s/agent_loop/slowest/tool_calls":0,"val-aux/tooluse/reward/worst@8/std":0.03169698340721522,"_wandb":{"runtime":3003},"rollout_corr/rollout_is_ratio_fraction_high":4.97240325785242e-05,"val-aux/tooluse/acc/maj@4/mean":0.571764705882353,"rollout_corr/rollout_log_ppl":0.13595498194990796,"rollout_corr/ppl_ratio":1.0005107254255563,"response_length_non_aborted/max":403,"training/rollout_probs_diff_std":0.007261213846504688,"rollout_corr/log_ppl_diff_min":-0.018534213304519653,"val-aux/tooluse/score/mean@16":0.5689338235294118,"val-aux/tooluse/acc/worst@2/mean":0.5376176470588235,"val-aux/tooluse/reward/std@16":0.07216799159273436,"val-aux/tooluse/score/best@2/mean":0.599735294117647,"val-aux/tooluse/incorrect_format/best@16/mean":0,"perf/total_num_tokens":168287,"val-aux/tooluse/score/worst@4/std":0.050276642601264465,"val-aux/tooluse/acc/maj@8/std":0.04699356851985658,"prompt_length/max":1063,"val-aux/tooluse/score/worst@8/std":0.03169698340721522,"val-aux/tooluse/acc/worst@4/mean":0.5104117647058823,"prompt_length/min":176,"actor/ppo_kl":9.126732021513817e-05,"val-aux/tooluse/incorrect_format/maj@16/mean":0,"val-aux/tooluse/score/maj@2/mean":0.5681911764705883,"val-aux/tooluse/incorrect_format/best@4/std":0,"timing_s/agent_loop/slowest/response_length":403,"response_length/clip_ratio":0,"val-aux/tooluse/incorrect_format/best@2/std":0,"rollout_corr/chi2_seq":0.05812919745221734,"rollout_corr/log_ppl_diff":0.0005012159926991444,"val-core/tooluse/acc/best@16/std":0.011314008550944105,"val-aux/tooluse/score/best@2/std":0.05681896001528811,"rollout_corr/rollout_is_seq_mean":0.9998652338981628,"val-aux/tooluse/reward/worst@16/std":0.018056061970488427,"val-aux/tooluse/reward/best@4/mean":0.6216029411764705,"val-aux/tooluse/acc/best@8/std":0.021009683207312644,"val-aux/tooluse/incorrect_format/worst@2/std":0,"num_turns/max":2,"global_seqlen/balanced_min":21035,"val-aux/tooluse/acc/worst@8/mean":0.4913382352941176,"val-aux/tooluse/acc/maj@8/mean":0.5750294117647058,"val-aux/tooluse/score/worst@2/std":0.065873083983826,"rollout_corr/rollout_is_max":2.2650070190429688,"actor/entropy":0.14357170462608337,"val-aux/tooluse/reward/maj@16/std":0.03730316048999515,"val-aux/tooluse/reward/best@2/std":0.05681896001528811,"val-aux/tooluse/reward/best@16/mean":0.6409558823529412,"val-aux/tooluse/score/maj@16/mean":0.5770294117647059,"response_length_non_aborted/min":22,"val-aux/tooluse/reward/worst@16/mean":0.47960294117647057,"critic/returns/mean":0.0050159115344285965,"critic/advantages/min":-0.75,"val-aux/tooluse/reward/best@4/std":0.03612363935746805,"val-aux/tooluse/acc/maj@2/mean":0.5681911764705883,"val-aux/tooluse/incorrect_format/worst@16/std":0,"rollout_corr/log_ppl_diff_max":0.017064914107322693,"val-aux/tooluse/reward/maj@2/mean":0.5681911764705883,"rollout_corr/rollout_is_std":0.02362494170665741,"global_seqlen/balanced_max":21036,"actor/pg_clipfrac":0.0002753048320300877,"rollout_corr/rollout_ppl":1.1506293234415352,"timing_s/agent_loop/generate_sequences/mean":0.7321565171878319,"val-aux/tooluse/incorrect_format/best@4/mean":0,"response_length_non_aborted/mean":78.55859375,"rollout_corr/log_ppl_abs_diff":0.002873804867704166,"val-aux/tooluse/incorrect_format/worst@2/mean":0,"response/aborted_ratio":0,"val-aux/tooluse/incorrect_format/worst@4/mean":0,"rollout_corr/kl":0.0005012150181968877,"val-aux/tooluse/incorrect_format/maj@4/mean":0,"val-aux/tooluse/reward/best@8/std":0.021009683207312644,"val-aux/tooluse/incorrect_format/worst@4/std":0,"val-aux/num_turns/max":2,"actor/grad_norm":0.23568445537239313,"_runtime":3003.977272667,"global_seqlen/minmax_diff":6763,"val-aux/tooluse/acc/worst@16/mean":0.47960294117647057,"val-aux/tooluse/reward/maj@16/mean":0.5770294117647059,"rollout_corr/rollout_is_seq_std":0.003272402100265026,"val-core/tooluse/acc/best@16/mean":0.6409558823529412,"val-aux/tooluse/score/worst@2/mean":0.5376176470588235,"rollout_corr/rollout_is_seq_max":1.0165464878082275,"val-aux/tooluse/score/worst@16/std":0.018056061970488427,"actor/lr":1e-06,"val-aux/tooluse/incorrect_format/best@2/mean":0,"val-core/tooluse/acc/maj@16/std":0.03730316048999515,"response_length_non_aborted/clip_ratio":0,"val-aux/tooluse/acc/worst@4/std":0.050276642601264465,"rollout_corr/rollout_is_eff_sample_size":0.9994420543741424,"training/global_step":100,"perf/mfu/actor":0.06166838756342609,"global_seqlen/max":25286,"val-aux/tooluse/score/maj@16/std":0.03730316048999515,"perf/cpu_memory_used_gb":105.7668685913086} |