diff --git a/.gitattributes b/.gitattributes index adae362e5b86fb5c62759ea7b5b6c0b6a6fae6bb..d59b41e0701b0c757ee6be188d618aa4d05c6f05 100644 --- a/.gitattributes +++ b/.gitattributes @@ -63,3 +63,10 @@ zain/Activation/wandb/run-20260814_212208-n6fjpg9o/run-n6fjpg9o.wandb filter=lfs zain/Activation/out/glu-waleed10-100L_trash_run/training_log.jsonl filter=lfs diff=lfs merge=lfs -text zain/Activation/wandb/run-20260814_212835-gqskne0x/run-gqskne0x.wandb filter=lfs diff=lfs merge=lfs -text zain/Activation/out/glu-silu-waleed10-100L_trash_run/training_log.jsonl filter=lfs diff=lfs merge=lfs -text +zain/Activation/wandb/run-20260814_213153-q9vcmje5/run-q9vcmje5.wandb filter=lfs diff=lfs merge=lfs -text +zain/Activation/wandb/run-20260814_214549-k24y90ws/run-k24y90ws.wandb filter=lfs diff=lfs merge=lfs -text +zain/Activation/wandb/run-20260814_215902-npap5k3g/run-npap5k3g.wandb filter=lfs diff=lfs merge=lfs -text +zain/Activation/wandb/run-20260814_221154-a3607q94/run-a3607q94.wandb filter=lfs diff=lfs merge=lfs -text +zain/Activation/wandb/run-20260814_222456-wcvy5vyw/run-wcvy5vyw.wandb filter=lfs diff=lfs merge=lfs -text +zain/Activation/wandb/run-20260814_223829-uuv4h936/run-uuv4h936.wandb filter=lfs diff=lfs merge=lfs -text +zain/Activation/wandb/run-20260814_225218-x87hwrtd/run-x87hwrtd.wandb filter=lfs diff=lfs merge=lfs -text diff --git a/zain/Activation/wandb/run-20260814_212835-gqskne0x/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_212835-gqskne0x/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..b3ef1934ca0195624f58a65526750a266f85b5f7 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_212835-gqskne0x/files/wandb-summary.json @@ -0,0 +1 @@ +{"train/tensor_act_model_layers_78_mlp/max":4,"train/tensor_act_model_layers_12_mlp_gate_proj/norm":5173.50341796875,"train/tensor_act_model_layers_40_mlp_down_proj/mean":14.75,"train/tensor_param_model_layers_68_mlp_up_proj_weight/max_abs":0.134765625,"train/tensor_act_model_layers_74_mlp_gate_proj/std":1.65625,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_18_mlp_gate_proj/max":4.71875,"train/tensor_param_model_layers_45_mlp_up_proj_weight/range":0.2548828125,"train/tensor_act_model_layers_83_residual_post_attn/min":330,"train/tensor_param_model_layers_47_mlp_up_proj_weight/mean":-0.002838134765625,"train/tensor_param_model_layers_88_mlp_up_proj_weight/min":-0.158203125,"train/tensor_param_model_layers_7_mlp_up_proj_weight/range":0.20263671875,"train/tensor_act_model_layers_50_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_3_residual_post_attn/max_abs":12.625,"train/tensor_param_model_layers_22_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_49_residual_post_mlp/max":201,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/min":-0.125,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_55_mlp_down_proj/mean":14.875,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_44_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_86_mlp/mean":4,"train/tensor_act_model_layers_59_residual_post_mlp/min":238,"train/tensor_param_model_layers_26_mlp_down_proj_weight/std":0.03173828125,"train/tensor_param_model_layers_32_mlp_down_proj_weight/norm":5.78125,"train/tensor_act_model_layers_42_residual_post_attn/norm":343916.65625,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_1_mlp_down_proj/max_abs":20.625,"train/tensor_act_model_layers_66_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/mean":0.0035247802734375,"train/tensor_act_model_layers_69_mlp_down_proj/min":14.1875,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_92_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_48_mlp_up_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_94_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_29_residual_post_mlp/min":118.5,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_82_mlp_up_proj_weight/norm":8.0625,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_98_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_15_mlp/std":0,"train/tensor_param_model_layers_87_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_70_residual_post_mlp/max":284,"train/tensor_act_model_layers_8_mlp/std":0,"train/tensor_act_model_layers_13_residual_pre_attn/mean":52,"train/tensor_act_model_layers_89_residual_post_attn/mean":356,"train/tensor_act_model_layers_34_mlp_gate_proj/min":-1.28125,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_97_mlp/norm":8192,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/std":0.03173828125,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_15_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_mlp_down_proj/range":4.1875,"train/tensor_act_model_layers_59_mlp/min":4,"train/tensor_act_model_layers_15_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_33_residual_post_mlp/norm":278380.8125,"train/tensor_act_model_layers_64_mlp/max_abs":4,"train/tensor_param_model_layers_67_mlp_up_proj_weight/max":0.13671875,"train/tensor_act_model_layers_94_residual_post_attn/max_abs":376,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_residual_post_mlp/min":222,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_99_mlp_gate_proj/std":1.734375,"train/tensor_act_model_layers_75_residual_post_mlp/norm":622557.3125,"train/tensor_param_model_layers_22_mlp_up_proj_weight/max_abs":0.11767578125,"train/tensor_act_model_layers_7_residual_post_attn/mean":28,"train/tensor_act_model_layers_84_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_87_mlp_up_proj_weight/std":0.0458984375,"train/tensor_act_model_layers_22_residual_post_attn/min":86.5,"train/tensor_act_model_layers_53_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_49_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_65_mlp/min":4,"train/tensor_act_model_layers_1_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_residual_post_attn/mean":252,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/max":0.1259765625,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_97_mlp_up_proj/norm":5186.61767578125,"train/tensor_param_model_layers_35_mlp_down_proj_weight/max_abs":0.12158203125,"train/tensor_act_model_layers_54_residual_pre_attn/max":217,"train/tensor_act_model_layers_28_mlp_gate_proj/min":-1.2578125,"train/tensor_param_model_layers_44_mlp_down_proj_weight/norm":6.125,"train/tensor_act_model_layers_22_residual_post_mlp/norm":188177.328125,"train/tensor_param_model_layers_94_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_77_mlp_up_proj/mean":-0.232421875,"train/tensor_act_model_layers_96_mlp_up_proj/max_abs":6.28125,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_79_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_75_residual_post_attn/min":298,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/norm":5.8125,"train/tensor_act_model_layers_45_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_78_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_9_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_33_residual_post_attn/min":130,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/mean":0.00439453125,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_12_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_90_mlp/norm":8192,"train/tensor_act_model_layers_31_mlp_gate_proj/max":4.59375,"train/tensor_act_model_layers_13_residual_pre_attn/min":51,"train/tensor_act_model_layers_32_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_83_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_35_residual_pre_attn/min":138,"train/tensor_act_model_layers_8_mlp_up_proj/max_abs":4.71875,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/std":0.034423828125,"train/tensor_act_model_layers_33_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_22_mlp_down_proj/norm":29815.462890625,"train/tensor_param_model_layers_43_mlp_down_proj_weight/norm":6.125,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_82_residual_post_attn/mean":328,"train/tensor_act_model_layers_75_mlp_gate_proj/max":4.3125,"train/tensor_act_model_layers_65_mlp/mean":4,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/max":0.1376953125,"train/tensor_act_model_layers_35_residual_pre_attn/std":0.31640625,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_43_mlp/mean":4,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_87_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_76_residual_post_mlp/range":2,"train/tensor_act_model_layers_8_residual_pre_attn/mean":32,"train/tensor_act_model_layers_6_residual_pre_attn/norm":49117.7578125,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_95_mlp_up_proj/max":6,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/min":-0.12255859375,"train/tensor_act_model_layers_22_mlp_up_proj/range":8.1875,"train/tensor_act_model_layers_51_mlp_gate_proj/mean":0.484375,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_86_mlp_down_proj_weight/max":0.14453125,"train/tensor_act_model_layers_92_residual_pre_attn/mean":368,"train/tensor_act_model_layers_73_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_22_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_13_residual_pre_attn/std":0.267578125,"train/tensor_param_model_layers_39_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_mlp/max_abs":4,"train/tensor_param_model_layers_38_mlp_down_proj_weight/min":-0.1279296875,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_82_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/std":0.03955078125,"train/tensor_act_model_layers_50_mlp_gate_proj/mean":0.6875,"train/tensor_act_model_layers_78_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/std":0.03564453125,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_87_residual_post_mlp/max_abs":352,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_62_mlp/range":0,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/max":0.1162109375,"train/tensor_act_model_layers_29_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_54_residual_pre_attn/range":3,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_54_mlp_down_proj/max_abs":16.75,"train/tensor_act_model_layers_22_mlp_gate_proj/max":4.46875,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_13_residual_post_attn/max":53,"train/tensor_param_model_layers_1_mlp_up_proj_weight/min":-0.11865234375,"train/tensor_act_model_layers_23_mlp_down_proj/std":0.42578125,"train/tensor_act_model_layers_92_residual_post_mlp/range":2,"train/tensor_act_model_layers_61_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_43_mlp_down_proj_weight/max":0.12255859375,"train/tensor_param_model_layers_10_mlp_up_proj_weight/range":0.2099609375,"train/tensor_act_model_layers_98_mlp_gate_proj/min":-1.59375,"train/tensor_act_model_layers_51_residual_post_attn/max_abs":205,"train/tensor_param_model_layers_34_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_88_residual_post_attn/max_abs":352,"train/tensor_act_model_layers_81_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_9_mlp_down_proj_weight/range":0.224609375,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_79_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_73_mlp_down_proj_weight/std":0.0380859375,"train/tensor_act_model_layers_34_mlp/min":4,"train/tensor_param_model_layers_55_mlp_up_proj_weight/norm":6.71875,"train/tensor_act_model_layers_75_mlp_up_proj/range":8.875,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_56_mlp_down_proj_weight/max_abs":0.1376953125,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_75_mlp_up_proj/std":1.8359375,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/std":0.03466796875,"train/tensor_act_model_layers_31_mlp_up_proj/range":8.4375,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_92_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/min":-0.12890625,"train/tensor_param_model_layers_71_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_57_mlp_up_proj/max_abs":5,"train/tensor_act_model_layers_97_mlp_down_proj/range":5.875,"train/tensor_param_model_layers_99_mlp_up_proj_weight/range":0.3291015625,"train/tensor_act_model_layers_65_mlp_gate_proj/std":1.609375,"train/tensor_act_model_layers_80_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_2_mlp_down_proj_weight/max":0.103515625,"train/tensor_act_model_layers_11_mlp_gate_proj/max_abs":3.859375,"train/tensor_act_model_layers_4_mlp_down_proj/range":1.8125,"train/tensor_act_model_layers_7_mlp/mean":4,"train/tensor_act_model_layers_1_residual_pre_attn/frac_near_user_limit":0,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_45_mlp_up_proj/norm":5460.86376953125,"train/tensor_act_model_layers_6_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_93_residual_pre_attn/min":370,"train/tensor_act_model_layers_44_residual_post_mlp/norm":368492.625,"train/tensor_act_model_layers_39_mlp_up_proj/std":1.8359375,"train/tensor_act_model_layers_51_residual_post_attn/mean":204,"train/tensor_param_model_layers_91_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_14_mlp_gate_proj/std":1.7578125,"train/tensor_act_model_layers_44_mlp_up_proj/norm":5352.05712890625,"train/tensor_act_model_layers_94_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_26_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_84_residual_post_mlp/range":2,"train/tensor_param_model_layers_52_mlp_up_proj_weight/range":0.2822265625,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_52_mlp/norm":8192,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_49_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_92_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_68_mlp_up_proj/mean":-0.2470703125,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/std":0.056396484375,"train/tensor_act_model_layers_49_mlp/std":0,"train/tensor_act_model_layers_29_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_44_residual_post_attn/min":174,"train/tensor_act_model_layers_68_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_25_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_44_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_52_mlp/min":4,"train/tensor_act_model_layers_93_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_65_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_8_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_47_mlp_gate_proj/std":1.6953125,"train/tensor_act_model_layers_62_mlp_down_proj/norm":30491.349609375,"train/tensor_param_model_layers_7_mlp_down_proj_weight/mean":0.000659942626953125,"train/tensor_param_model_layers_18_mlp_down_proj_weight/max":0.11279296875,"train/tensor_act_model_layers_54_mlp_gate_proj/norm":4975.09423828125,"train/tensor_act_model_layers_72_residual_pre_attn/max_abs":288,"train/tensor_act_model_layers_45_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_32_mlp/mean":4,"train/tensor_param_model_layers_69_mlp_down_proj_weight/mean":-6.532669067382812e-05,"train/tensor_act_model_layers_99_mlp/range":0,"train/tensor_act_model_layers_55_residual_post_mlp/norm":458540.8125,"train/tensor_act_model_layers_17_mlp_gate_proj/max_abs":4.625,"train/tensor_param_model_layers_56_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_5_mlp_up_proj/range":8.5,"train/tensor_act_model_layers_14_residual_post_mlp/std":0.3125,"train/tensor_act_model_layers_92_residual_post_attn/max":368,"train/tensor_param_model_layers_77_mlp_up_proj_weight/mean":-0.0018157958984375,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_50_mlp_down_proj/range":2.0625,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/range":0.30078125,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/range":0.2880859375,"train/tensor_act_model_layers_52_mlp_gate_proj/max_abs":4.75,"train/tensor_param_model_layers_52_mlp_down_proj_weight/std":0.0361328125,"train/tensor_act_model_layers_46_residual_pre_attn/norm":376684.59375,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_35_mlp_up_proj/max":3.984375,"train/tensor_act_model_layers_8_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/norm":8.3125,"train/tensor_act_model_layers_53_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_22_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_43_residual_post_mlp/mean":176,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_22_mlp_down_proj_weight/norm":5.625,"train/tensor_param_model_layers_93_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_residual_post_attn/min":218,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_13_residual_post_attn/std":0.267578125,"train/tensor_act_model_layers_17_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_43_residual_post_mlp/min":174,"train/tensor_param_model_layers_99_mlp_down_proj_weight/range":0.3486328125,"train/tensor_param_model_layers_74_mlp_down_proj_weight/mean":-0.0004787445068359375,"train/tensor_param_model_layers_84_mlp_down_proj_weight/min":-0.1455078125,"train/tensor_act_model_layers_92_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_22_mlp_up_proj/mean":-0.1669921875,"train/tensor_param_model_layers_64_mlp_up_proj_weight/std":0.0380859375,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_19_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_9_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_96_mlp_up_proj_weight/min":-0.1953125,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/mean":0.004058837890625,"train/tensor_act_model_layers_99_residual_post_attn/norm":810973.25,"train/tensor_act_model_layers_18_mlp_up_proj/range":8.296875,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_26_residual_post_attn/min":102.5,"train/tensor_act_model_layers_85_mlp_gate_proj/min":-1.3515625,"train/tensor_act_model_layers_62_mlp_up_proj/norm":5301.6376953125,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_24_residual_post_mlp/max":101,"train/tensor_act_model_layers_68_residual_post_mlp/min":274,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/std":0.030029296875,"train/tensor_act_model_layers_7_mlp_gate_proj/max":4.34375,"train/tensor_act_model_layers_75_mlp_gate_proj/norm":5064.3134765625,"train/tensor_act_model_layers_26_residual_post_mlp/norm":220945.109375,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_48_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/mean":0.0040283203125,"train/tensor_param_model_layers_5_mlp_down_proj_weight/max_abs":0.123046875,"train/tensor_act_model_layers_52_residual_pre_attn/norm":425836.53125,"train/tensor_param_model_layers_71_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/max":0.10986328125,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_37_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_mlp_up_proj/min":-4.9375,"train/tensor_act_model_layers_63_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_mlp_gate_proj/max_abs":5.09375,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/norm":7.625,"train/tensor_param_model_layers_49_mlp_down_proj_weight/norm":6.4375,"train/tensor_act_model_layers_30_residual_post_mlp/std":0.388671875,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_95_mlp_gate_proj/max_abs":6.15625,"train/tensor_act_model_layers_82_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_34_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/min":-0.123046875,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_94_mlp_up_proj/range":9.5625,"train/tensor_param_model_layers_86_mlp_up_proj_weight/std":0.04638671875,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_45_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_residual_post_attn/norm":114487.640625,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/frac_near_dtype_limit":0,"_runtime":389,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_90_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_64_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_43_mlp_up_proj_weight/max":0.12255859375,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/std":0.02978515625,"train/tensor_act_model_layers_70_mlp_up_proj/min":-4.3125,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_73_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_residual_post_mlp/min":270,"train/tensor_act_model_layers_74_mlp_up_proj/max":4.40625,"train/tensor_act_model_layers_77_mlp_up_proj/range":9.15625,"train/tensor_act_model_layers_34_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_0_residual_post_mlp/max":4.3125,"train/tensor_param_model_layers_87_mlp_up_proj_weight/mean":0.001953125,"train/tensor_act_model_layers_79_mlp/range":0,"train/tensor_param_model_layers_95_mlp_down_proj_weight/std":0.050537109375,"train/tensor_act_model_layers_79_residual_post_mlp/max_abs":320,"train/tensor_act_model_layers_94_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_61_mlp_up_proj_weight/min":-0.1318359375,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/frac_near_user_limit":0.00177001953125,"train/tensor_act_model_layers_97_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_residual_pre_attn/max":105,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_13_mlp_gate_proj/max":4.625,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/mean":0.004730224609375,"train/tensor_act_model_layers_74_mlp/norm":8192,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_48_residual_pre_attn/range":3,"train/tensor_act_model_layers_5_mlp_down_proj/max":16.5,"train/tensor_act_model_layers_55_mlp_gate_proj/min":-1.3203125,"train/tensor_act_model_layers_98_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_84_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_77_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_58_mlp/max_abs":4,"train/tensor_act_model_layers_31_residual_post_mlp/min":126.5,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/std":0.033203125,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_70_mlp/norm":8192,"train/tensor_act_model_layers_22_mlp_down_proj/mean":14.5625,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_3_mlp_up_proj_weight/min":-0.10302734375,"train/tensor_act_model_layers_51_mlp_up_proj/std":1.8515625,"train/tensor_param_model_layers_86_mlp_down_proj_weight/min":-0.1455078125,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_68_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp_up_proj/range":9.28125,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_23_mlp_down_proj_weight/std":0.031005859375,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/max":0.11962890625,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_35_mlp/norm":8192,"train/tensor_act_model_layers_63_mlp_down_proj/max":16.125,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_93_mlp_down_proj_weight/mean":0.0005645751953125,"train/tensor_act_model_layers_2_mlp/mean":4,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_94_mlp_up_proj_weight/max":0.1484375,"train/tensor_act_model_layers_38_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_residual_post_mlp/max_abs":20.625,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/max_abs":0.1162109375,"train/tensor_param_model_layers_92_mlp_up_proj_weight/norm":8.9375,"train/tensor_act_model_layers_45_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_97_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/min":-0.130859375,"train/tensor_act_model_layers_82_mlp_down_proj/max":18.25,"train/tensor_act_model_layers_11_mlp_down_proj/max":15.875,"train/tensor_act_model_layers_34_residual_post_mlp/max_abs":141,"train/tensor_act_model_layers_51_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_72_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_15_residual_post_attn/range":2,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_36_mlp_gate_proj/range":6.0703125,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_68_mlp_up_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_88_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_96_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_65_mlp_up_proj_weight/max_abs":0.1357421875,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/range":0.275390625,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/max":0.1201171875,"train/tensor_act_model_layers_40_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_27_mlp_down_proj/max_abs":15.875,"train/tensor_act_model_layers_70_mlp_gate_proj/max_abs":4.3125,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_72_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_mlp_up_proj/min":-4.28125,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/range":0.40234375,"train/tensor_act_model_layers_70_mlp/min":4,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_43_mlp_down_proj/norm":30296.5546875,"train/tensor_param_model_layers_81_mlp_down_proj_weight/mean":-0.0003223419189453125,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/norm":7.28125,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_65_mlp_down_proj_weight/range":0.2529296875,"train/tensor_act_model_layers_52_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_59_mlp_down_proj_weight/std":0.036865234375,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_61_residual_pre_attn/mean":244,"train/tensor_act_model_layers_40_mlp_down_proj/norm":30199.955078125,"train/tensor_act_model_layers_21_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_residual_pre_attn/range":2,"train/tensor_act_model_layers_56_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_16_mlp_gate_proj/max":5.0625,"train/tensor_act_model_layers_55_mlp_down_proj/norm":30443.46484375,"train/tensor_act_model_layers_3_mlp_up_proj/mean":0.03857421875,"train/tensor_act_model_layers_5_residual_post_mlp/range":1.375,"train/tensor_act_model_layers_69_residual_pre_attn/mean":276,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_89_mlp/max_abs":4,"train/tensor_act_model_layers_40_residual_post_mlp/min":162,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_93_mlp_up_proj/max":4.75,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/mean":0.0040283203125,"train/tensor_act_model_layers_71_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_9_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_68_mlp_gate_proj/max_abs":4.625,"train/tensor_act_model_layers_43_mlp_up_proj/max":4.90625,"train/tensor_param_model_layers_24_mlp_up_proj_weight/range":0.22216796875,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_86_residual_post_mlp/range":2,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_12_residual_post_attn/range":2,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_93_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_46_residual_pre_attn/min":182,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/max_abs":0.13671875,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/mean":0.0038909912109375,"train/tensor_param_model_layers_25_mlp_up_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_78_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_79_mlp_gate_proj/std":1.6796875,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_46_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_32_residual_pre_attn/max_abs":129,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_57_residual_post_attn/range":3,"train/tensor_param_model_layers_83_mlp_down_proj_weight/range":0.2861328125,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_10_mlp_gate_proj/mean":0.431640625,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/mean":0.00433349609375,"train/tensor_act_model_layers_62_mlp_up_proj/range":9.6875,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/range":0.2392578125,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_20_mlp_down_proj/max":16.25,"train/tensor_act_model_layers_36_residual_post_attn/min":142,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_41_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_16_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_mlp_gate_proj/max":4.6875,"train/tensor_act_model_layers_34_mlp_up_proj/max_abs":4.59375,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_91_mlp_up_proj_weight/range":0.3115234375,"train/tensor_act_model_layers_62_mlp_down_proj/mean":14.875,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_52_mlp_up_proj_weight/max_abs":0.1484375,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_38_residual_post_mlp/mean":156,"train/tensor_param_model_layers_54_mlp_up_proj_weight/range":0.25390625,"train/tensor_act_model_layers_58_mlp/std":0,"train/tensor_param_model_layers_84_mlp_down_proj_weight/max":0.158203125,"train/tensor_param_model_layers_27_mlp_up_proj_weight/norm":5.75,"train/tensor_act_model_layers_19_residual_post_attn/min":74.5,"train/tensor_param_model_layers_94_mlp_down_proj_weight/max_abs":0.1767578125,"train/tensor_act_model_layers_73_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_93_residual_post_mlp/mean":376,"train/tensor_act_model_layers_59_residual_pre_attn/range":3,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/std":0.029052734375,"train/tensor_act_model_layers_19_residual_post_mlp/norm":163610.28125,"train/tensor_act_model_layers_19_mlp_up_proj/max_abs":5.0625,"train/tensor_act_model_layers_37_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_18_mlp_down_proj_weight/min":-0.11767578125,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp/max_abs":4,"train/tensor_act_model_layers_52_mlp_gate_proj/std":1.6484375,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_10_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_21_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_43_mlp/std":0,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_32_mlp_up_proj/max_abs":4.4375,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/range":0.23095703125,"train/tensor_act_model_layers_49_residual_post_attn/norm":401260.59375,"train/tensor_act_model_layers_55_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_84_mlp_up_proj_weight/mean":0.000522613525390625,"train/tensor_param_model_layers_58_mlp_down_proj_weight/mean":0.0006103515625,"train/tensor_act_model_layers_36_residual_post_mlp/norm":302956.75,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_64_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/max_abs":0.169921875,"train/tensor_act_model_layers_89_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_69_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/norm":10.5,"train/tensor_act_model_layers_99_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_48_mlp_down_proj_weight/max_abs":0.125,"train/tensor_param_model_layers_30_mlp_up_proj_weight/std":0.032470703125,"train/tensor_act_model_layers_70_mlp_up_proj/mean":0.05322265625,"train/tensor_act_model_layers_95_residual_post_attn/min":378,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/range":0.3427734375,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_64_mlp_down_proj_weight/std":0.037353515625,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_25_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_53_residual_pre_attn/max":213,"train/tensor_act_model_layers_49_mlp_gate_proj/std":1.6640625,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/norm":5.5625,"train/tensor_param_model_layers_88_mlp_down_proj_weight/norm":8.3125,"train/tensor_param_model_layers_90_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_25_residual_post_mlp/std":0.359375,"train/tensor_act_model_layers_16_mlp_up_proj/range":9.25,"train/tensor_act_model_layers_66_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_82_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_68_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/min":-0.11474609375,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_60_mlp_up_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_16_residual_post_attn/min":62.75,"train/tensor_param_model_layers_35_mlp_down_proj_weight/norm":5.875,"train/tensor_act_model_layers_94_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_residual_post_attn/std":0.416015625,"train/tensor_param_model_layers_75_mlp_up_proj_weight/mean":0.00148773193359375,"train/tensor_param_model_layers_89_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_49_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_16_residual_post_mlp/norm":139066.421875,"train/tensor_act_model_layers_93_mlp_gate_proj/min":-1.6875,"train/tensor_act_model_layers_59_mlp_down_proj/min":14.1875,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_51_mlp/min":4,"train/tensor_param_model_layers_55_mlp_up_proj_weight/range":0.25390625,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/std":0.039306640625,"train/tensor_act_model_layers_89_mlp_gate_proj/std":1.625,"train/tensor_act_model_layers_11_mlp_up_proj/max":4.03125,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/range":0.24072265625,"train/tensor_act_model_layers_69_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_95_residual_post_mlp/norm":786397.25,"train/tensor_param_model_layers_15_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_7_residual_post_mlp/max":32.75,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_42_residual_post_attn/max_abs":169,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_99_residual_post_attn/min":394,"train/tensor_act_model_layers_64_mlp_down_proj/range":2.4375,"train/tensor_act_model_layers_76_mlp_up_proj/norm":5307.18505859375,"train/tensor_param_model_layers_25_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/range":0.279296875,"train/tensor_act_model_layers_54_residual_post_attn/range":3,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/range":0.3056640625,"train/tensor_param_model_layers_31_mlp_up_proj_weight/mean":0.00121307373046875,"train/tensor_param_model_layers_51_mlp_up_proj_weight/max":0.126953125,"train/tensor_act_model_layers_0_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_28_mlp/min":4,"train/tensor_act_model_layers_39_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_17_mlp_up_proj_weight/min":-0.11279296875,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/max":0.125,"train/tensor_param_model_layers_4_mlp_up_proj_weight/max":0.10546875,"train/tensor_act_model_layers_22_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_33_mlp_gate_proj/norm":5263.6650390625,"train/tensor_param_model_layers_51_mlp_up_proj_weight/mean":0.00030517578125,"train/tensor_act_model_layers_54_mlp_up_proj/max_abs":4.46875,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_45_mlp_gate_proj/max":4.84375,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/mean":0.0054931640625,"train/tensor_param_model_layers_68_mlp_up_proj_weight/norm":6.96875,"train/tensor_param_model_layers_96_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_38_mlp_down_proj_weight/max":0.1181640625,"train/tensor_param_model_layers_61_mlp_down_proj_weight/range":0.2646484375,"train/tensor_act_model_layers_93_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_7_mlp_down_proj/std":0.3828125,"train/tensor_act_model_layers_48_residual_pre_attn/max":193,"train/tensor_param_model_layers_43_mlp_down_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_43_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_92_mlp_down_proj/range":3.8125,"train/tensor_act_model_layers_99_residual_post_mlp/norm":819165.25,"train/tensor_act_model_layers_17_residual_post_attn/norm":139066.421875,"train/tensor_act_model_layers_10_mlp_up_proj/max":3.96875,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_81_mlp_gate_proj/min":-1.34375,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_65_mlp_down_proj_weight/std":0.03759765625,"train/tensor_act_model_layers_83_mlp_gate_proj/mean":0.7109375,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/norm":6.28125,"train/tensor_act_model_layers_61_residual_post_attn/max":245,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_13_mlp_up_proj/range":8.5625,"train/tensor_param_model_layers_3_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_residual_post_attn/max":276,"train/tensor_act_model_layers_39_mlp_gate_proj/norm":5001.115234375,"train/tensor_act_model_layers_12_mlp_down_proj/min":14,"train/tensor_act_model_layers_2_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_63_residual_pre_attn/norm":515868.75,"train/tensor_act_model_layers_37_residual_post_attn/norm":302956.75,"train/tensor_act_model_layers_26_mlp_down_proj/norm":29944.880859375,"train/tensor_act_model_layers_15_residual_post_mlp/range":2,"train/tensor_act_model_layers_64_mlp/mean":4,"train/tensor_act_model_layers_93_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_65_residual_post_mlp/mean":264,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/std":0.038818359375,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_77_mlp_gate_proj/max":4.78125,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_77_residual_pre_attn/range":2,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_mlp_down_proj/range":4.1875,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_34_mlp_down_proj_weight/max_abs":0.12890625,"train/tensor_act_model_layers_10_mlp_down_proj/max":16,"train/tensor_param_model_layers_65_mlp_down_proj_weight/mean":-0.00077056884765625,"train/tensor_param_model_layers_19_mlp_up_proj_weight/range":0.2568359375,"train/tensor_act_model_layers_94_mlp_gate_proj/max":5.34375,"train/tensor_act_model_layers_68_residual_post_attn/norm":557021.3125,"train/tensor_act_model_layers_40_mlp_up_proj/min":-4.25,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_10_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_53_mlp_down_proj/range":2.125,"train/tensor_act_model_layers_22_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_45_mlp/mean":4,"train/tensor_act_model_layers_60_mlp_up_proj/min":-4.21875,"train/tensor_act_model_layers_80_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_74_mlp_down_proj_weight/max_abs":0.1328125,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_25_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_residual_post_mlp/norm":688093.3125,"train/tensor_act_model_layers_81_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_57_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_mlp_down_proj/mean":14.8125,"train/tensor_param_model_layers_63_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_56_mlp_down_proj/std":0.435546875,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_59_mlp_up_proj_weight/max":0.126953125,"train/tensor_act_model_layers_25_mlp_gate_proj/std":1.703125,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_99_residual_pre_attn/max":396,"train/tensor_act_model_layers_90_mlp_down_proj/norm":31081.642578125,"train/tensor_act_model_layers_81_residual_pre_attn/max_abs":324,"train/tensor_act_model_layers_49_mlp_gate_proj/min":-1.2734375,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/max":0.1220703125,"train/tensor_act_model_layers_53_residual_post_mlp/std":0.408203125,"train/tensor_act_model_layers_82_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_91_residual_pre_attn/max_abs":364,"train/tensor_act_model_layers_64_mlp/min":4,"train/tensor_param_model_layers_45_mlp_up_proj_weight/std":0.034912109375,"train/tensor_act_model_layers_31_residual_post_attn/norm":253697.171875,"train/tensor_param_model_layers_52_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_25_mlp_up_proj/std":1.90625,"train/tensor_act_model_layers_25_residual_pre_attn/max":101,"train/tensor_param_model_layers_58_mlp_up_proj_weight/range":0.263671875,"train/tensor_act_model_layers_40_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/std":0.031982421875,"train/tensor_param_model_layers_23_mlp_down_proj_weight/max":0.11865234375,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_46_mlp/max_abs":4,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_71_mlp_up_proj_weight/std":0.0390625,"train/tensor_act_model_layers_5_residual_pre_attn/range":1.375,"train/tensor_act_model_layers_9_residual_pre_attn/std":0.2197265625,"train/tensor_act_model_layers_55_mlp/min":4,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_57_mlp_gate_proj/norm":5095.15283203125,"train/tensor_param_model_layers_95_mlp_up_proj_weight/range":0.3212890625,"train/tensor_act_model_layers_43_mlp/range":0,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_31_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_93_residual_post_mlp/range":2,"train/tensor_act_model_layers_70_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_27_mlp_down_proj_weight/max_abs":0.1259765625,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_52_mlp_gate_proj/min":-1.3125,"train/tensor_act_model_layers_33_residual_pre_attn/max_abs":133,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_19_mlp_up_proj/std":1.8671875,"train/tensor_act_model_layers_41_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_84_mlp/max_abs":4,"train/tensor_act_model_layers_13_mlp_up_proj/max_abs":4.5625,"train/tensor_act_model_layers_5_mlp_gate_proj/max_abs":4.53125,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_66_mlp_up_proj_weight/max":0.1494140625,"train/tensor_act_model_layers_48_residual_post_mlp/max_abs":197,"train/tensor_act_model_layers_36_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_95_mlp_down_proj_weight/max":0.189453125,"train/tensor_param_model_layers_36_mlp_down_proj_weight/norm":5.96875,"train/tensor_param_model_layers_63_mlp_up_proj_weight/max_abs":0.12890625,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_43_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_60_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_11_residual_post_mlp/mean":48,"train/tensor_act_model_layers_62_residual_post_attn/max":249,"train/tensor_act_model_layers_62_residual_post_mlp/max_abs":253,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_40_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_30_mlp_gate_proj/norm":5141.57421875,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_83_residual_post_attn/mean":332,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/min":-0.11474609375,"train/tensor_act_model_layers_84_mlp/norm":8192,"train/tensor_act_model_layers_41_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_68_mlp_down_proj_weight/range":0.263671875,"train/tensor_act_model_layers_89_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_29_residual_post_attn/mean":116,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_mlp_down_proj/std":0.41015625,"train/tensor_param_model_layers_26_mlp_down_proj_weight/min":-0.1181640625,"train/tensor_act_model_layers_36_residual_post_mlp/range":3,"train/tensor_act_model_layers_52_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_44_mlp_up_proj_weight/max":0.11865234375,"train/tensor_act_model_layers_92_residual_post_mlp/min":370,"train/tensor_act_model_layers_59_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_63_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_29_residual_post_attn/max_abs":117,"train/tensor_act_model_layers_27_residual_post_attn/min":106.5,"train/tensor_act_model_layers_45_residual_post_mlp/norm":376684.59375,"train/tensor_act_model_layers_0_mlp_gate_proj/range":14.15625,"train/tensor_act_model_layers_57_residual_post_mlp/max":233,"train/tensor_act_model_layers_24_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_72_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_residual_post_attn/min":178,"train/tensor_act_model_layers_22_mlp/norm":8192,"train/tensor_act_model_layers_0_residual_post_attn/mean":0.00015735626220703125,"train/tensor_act_model_layers_16_residual_post_attn/norm":130773.3515625,"train/tensor_act_model_layers_53_mlp_down_proj/max":16.375,"train/tensor_act_model_layers_7_mlp_gate_proj/range":5.6796875,"train/tensor_act_model_layers_14_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_59_mlp_up_proj/std":1.78125,"train/tensor_act_model_layers_98_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_30_mlp/std":0,"train/tensor_act_model_layers_61_residual_pre_attn/range":3,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_27_residual_post_mlp/norm":229137.046875,"train/tensor_act_model_layers_52_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_35_mlp_gate_proj/range":5.78125,"train/tensor_act_model_layers_21_mlp_down_proj/range":2.375,"train/tensor_param_model_layers_53_mlp_up_proj_weight/max":0.1337890625,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_85_mlp_up_proj/std":1.7734375,"train/tensor_act_model_layers_17_mlp_down_proj/max":16.375,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/min":-0.119140625,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_96_mlp_up_proj_weight/norm":9.75,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/range":0.263671875,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_72_mlp_up_proj_weight/std":0.038818359375,"train/tensor_act_model_layers_58_mlp_down_proj/range":2.9375,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_82_residual_post_attn/range":2,"train/tensor_act_model_layers_3_mlp/norm":8192,"train/tensor_act_model_layers_40_residual_post_mlp/mean":164,"train/tensor_param_model_layers_18_mlp_down_proj_weight/range":0.23046875,"train/tensor_param_model_layers_34_mlp_up_proj_weight/mean":0.0003147125244140625,"train/tensor_param_model_layers_56_mlp_up_proj_weight/std":0.036865234375,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_91_mlp/range":0,"train/tensor_param_model_layers_20_mlp_up_proj_weight/std":0.03125,"train/tensor_act_model_layers_48_residual_post_mlp/std":0.31640625,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/std":0.21875,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_62_residual_post_attn/std":0.416015625,"train/tensor_act_model_layers_74_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_51_mlp_down_proj/max":16.5,"train/tensor_act_model_layers_81_mlp_gate_proj/std":1.6015625,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_54_mlp_down_proj_weight/norm":6.59375,"train/tensor_act_model_layers_12_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_mlp_gate_proj/max_abs":4.875,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_3_residual_pre_attn/max_abs":12.4375,"train/tensor_param_model_layers_59_mlp_down_proj_weight/max":0.1396484375,"train/tensor_param_model_layers_30_mlp_up_proj_weight/norm":5.875,"train/tensor_param_model_layers_14_mlp_up_proj_weight/mean":9.107589721679688e-05,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_84_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_mlp_gate_proj/max":4.40625,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_74_residual_pre_attn/min":294,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_51_mlp_down_proj_weight/norm":6.46875,"train/tensor_act_model_layers_80_residual_post_mlp/min":322,"train/tensor_act_model_layers_18_residual_pre_attn/range":2,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_19_mlp_up_proj_weight/min":-0.140625,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_residual_post_mlp/max":36.75,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_78_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_post_attn/range":3,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/norm":5.84375,"train/tensor_param_model_layers_33_mlp_up_proj_weight/norm":5.84375,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/max":0.1328125,"train/tensor_act_model_layers_66_mlp_up_proj/norm":5293.830078125,"train/tensor_act_model_layers_34_mlp_gate_proj/norm":5153.59716796875,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/std":0.032470703125,"train/tensor_param_model_layers_47_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_24_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_residual_post_mlp/max":268,"train/tensor_act_model_layers_53_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_33_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_81_residual_post_mlp/range":2,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/min":-1856,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_72_residual_post_mlp/min":290,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_mlp_up_proj/range":9.0625,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_62_residual_pre_attn/mean":248,"train/tensor_act_model_layers_74_residual_pre_attn/norm":606173.3125,"train/tensor_act_model_layers_57_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_86_mlp/norm":8192,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/norm":7.0625,"train/tensor_act_model_layers_86_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_71_mlp_gate_proj/range":6.09375,"train/tensor_act_model_layers_19_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_68_mlp_up_proj/std":1.765625,"train/tensor_act_model_layers_36_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_mlp_gate_proj/std":1.6640625,"train/tensor_act_model_layers_48_mlp_up_proj/mean":0.050048828125,"train/tensor_param_model_layers_64_mlp_down_proj_weight/max":0.125,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_50_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_61_mlp_up_proj/min":-4.625,"train/tensor_act_model_layers_83_mlp_down_proj/max":19.5,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_66_residual_post_attn/range":2,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_88_mlp_up_proj_weight/norm":8.5,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_11_residual_pre_attn/max_abs":44.75,"train/tensor_act_model_layers_10_residual_post_mlp/min":43,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_33_residual_pre_attn/min":130,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_65_mlp_gate_proj/max_abs":4.96875,"train/tensor_act_model_layers_83_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_31_mlp_down_proj/range":1.9375,"train/tensor_act_model_layers_46_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_54_mlp/std":0,"train/tensor_act_model_layers_50_mlp_up_proj/min":-4.09375,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_54_mlp_up_proj_weight/max_abs":0.126953125,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_26_residual_post_mlp/std":0.359375,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_28_residual_post_attn/min":110.5,"train/tensor_act_model_layers_28_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_9_mlp_up_proj_weight/max_abs":0.12109375,"train/tensor_act_model_layers_90_residual_post_mlp/mean":364,"train/tensor_act_model_layers_27_mlp_down_proj/min":14.1875,"train/tensor_param_model_layers_0_mlp_down_proj_weight/range":0.25,"train/tensor_act_model_layers_77_residual_pre_attn/max":308,"train/tensor_act_model_layers_39_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_mlp_gate_proj/max_abs":4.6875,"train/tensor_param_model_layers_54_mlp_up_proj_weight/norm":6.65625,"train/tensor_param_model_layers_52_mlp_down_proj_weight/min":-0.1328125,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_4_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_96_mlp_up_proj/std":1.796875,"train/tensor_act_model_layers_98_residual_pre_attn/mean":392,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/norm":5.75,"train/tensor_act_model_layers_21_residual_post_mlp/max":89,"train/tensor_act_model_layers_30_residual_pre_attn/max":121,"train/tensor_act_model_layers_54_mlp_up_proj/mean":-0.10302734375,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_mlp_up_proj/max_abs":4.625,"train/tensor_param_model_layers_82_mlp_up_proj_weight/range":0.283203125,"train/tensor_act_model_layers_28_residual_post_mlp/norm":237305.21875,"train/tensor_act_model_layers_88_mlp_up_proj/max":4.53125,"train/tensor_act_model_layers_9_residual_pre_attn/max_abs":36.75,"train/tensor_act_model_layers_71_mlp_up_proj/norm":5239.36669921875,"train/tensor_param_model_layers_18_mlp_down_proj_weight/norm":5.59375,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_21_residual_post_attn/norm":171793.453125,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_80_residual_post_attn/max":320,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_39_residual_post_mlp/mean":160,"train/tensor_act_model_layers_4_mlp_up_proj/norm":5395.5888671875,"train/tensor_act_model_layers_38_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_44_mlp_up_proj/max_abs":5.0625,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_80_mlp_up_proj_weight/norm":7.6875,"train/tensor_act_model_layers_49_mlp_up_proj/norm":5415.50390625,"train/tensor_param_model_layers_84_mlp_down_proj_weight/std":0.044677734375,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_91_mlp_up_proj/max_abs":5.125,"train/tensor_param_model_layers_53_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp/std":0,"train/tensor_act_model_layers_79_mlp_gate_proj/min":-1.3046875,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/mean":1.9375,"train/tensor_param_model_layers_9_mlp_up_proj_weight/min":-0.12109375,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/norm":5.71875,"train/tensor_act_model_layers_2_mlp_down_proj/mean":15,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/max":0.1279296875,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_76_residual_pre_attn/norm":622557.3125,"train/tensor_act_model_layers_46_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_52_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_76_mlp_up_proj_weight/max":0.1279296875,"train/tensor_act_model_layers_59_mlp_up_proj/mean":-0.1416015625,"train/tensor_act_model_layers_14_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_20_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_44_residual_pre_attn/range":3,"train/tensor_act_model_layers_37_mlp_up_proj/min":-3.875,"train/tensor_act_model_layers_47_residual_post_attn/range":3,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/min":-0.12060546875,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/min":-0.181640625,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_59_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_29_residual_post_attn/range":2.5,"train/tensor_param_model_layers_65_mlp_up_proj_weight/min":-0.1328125,"train/tensor_act_model_layers_42_mlp_up_proj/max":4.4375,"train/tensor_param_model_layers_65_mlp_down_proj_weight/max":0.125,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_11_residual_pre_attn/range":1.75,"train/tensor_act_model_layers_75_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_residual_pre_attn/range":2,"train/tensor_act_model_layers_96_mlp_down_proj/norm":31293.40234375,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_98_residual_post_mlp/max_abs":396,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_46_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_27_mlp_up_proj_weight/min":-0.12109375,"train/tensor_act_model_layers_57_mlp_up_proj/min":-4.5,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_74_mlp_gate_proj/max_abs":4.6875,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_28_mlp_down_proj/norm":30080.130859375,"train/tensor_act_model_layers_97_mlp_up_proj/std":1.7890625,"train/tensor_act_model_layers_67_mlp_gate_proj/mean":0.51953125,"train/tensor_act_model_layers_36_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_87_residual_post_attn/mean":348,"train/tensor_act_model_layers_78_mlp_up_proj/max":4.5,"train/tensor_param_model_layers_38_mlp_down_proj_weight/std":0.033447265625,"train/tensor_act_model_layers_27_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_6_mlp_gate_proj/range":6.0625,"train/tensor_act_model_layers_36_mlp_down_proj/mean":14.6875,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_9_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/min":-0.134765625,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_95_residual_pre_attn/max":380,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/min":-0.130859375,"train/tensor_act_model_layers_39_mlp_up_proj/max_abs":4.1875,"train/tensor_act_model_layers_68_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_29_residual_pre_attn/norm":237305.21875,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_80_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_5_mlp_up_proj_weight/max_abs":0.1201171875,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/frac_near_dtype_limit":0,"total_flos":1.1291993309184e+16,"train/tensor_act_model_layers_60_residual_post_mlp/norm":499484.78125,"train/tensor_act_model_layers_71_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_89_mlp_up_proj/std":1.796875,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_73_mlp_down_proj_weight/min":-0.1318359375,"train/tensor_act_model_layers_61_mlp/std":0,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_60_residual_post_attn/min":238,"train/tensor_param_model_layers_23_mlp_up_proj_weight/mean":0.0014801025390625,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/min":-0.1259765625,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_79_mlp_down_proj/mean":14.9375,"train/tensor_act_model_layers_51_residual_post_attn/range":3,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_16_mlp_gate_proj/mean":0.322265625,"train/tensor_act_model_layers_39_residual_post_mlp/range":3,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/range":0.263671875,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/max_abs":0.2158203125,"train/tensor_act_model_layers_36_residual_pre_attn/min":142,"train/tensor_act_model_layers_0_mlp_down_proj/max_abs":99.5,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/std":0.034912109375,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_74_mlp/min":4,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_35_mlp/max":4,"train/tensor_act_model_layers_87_residual_post_attn/max_abs":348,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/max_abs":0.1357421875,"train/tensor_param_model_layers_12_mlp_down_proj_weight/max_abs":0.1201171875,"train/tensor_act_model_layers_31_residual_post_attn/std":0.396484375,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_79_mlp_down_proj_weight/range":0.2802734375,"train/tensor_act_model_layers_24_mlp_down_proj/max_abs":16.5,"train/tensor_act_model_layers_68_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_20_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_20_mlp/mean":4,"train/tensor_act_model_layers_92_residual_post_attn/range":2,"train/tensor_act_model_layers_58_mlp_up_proj/range":9.5625,"train/tensor_param_model_layers_33_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_84_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_89_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/max":0.126953125,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_9_mlp/norm":8192,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/mean":0.0048828125,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_89_residual_pre_attn/min":354,"train/tensor_param_model_layers_46_mlp_up_proj_weight/std":0.03466796875,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_46_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_pre_attn/max_abs":276,"train/tensor_act_model_layers_14_mlp_gate_proj/min":-1.28125,"train/tensor_param_model_layers_26_mlp_down_proj_weight/max":0.11669921875,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_28_mlp_down_proj/range":2.0625,"train/tensor_act_model_layers_88_residual_post_attn/norm":720861.25,"train/tensor_param_model_layers_72_mlp_down_proj_weight/mean":-0.0024566650390625,"train/tensor_act_model_layers_77_mlp/mean":4,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/mean":0.003265380859375,"train/tensor_act_model_layers_80_residual_post_mlp/range":2,"train/tensor_act_model_layers_49_residual_pre_attn/max_abs":197,"train/tensor_act_model_layers_2_residual_post_attn/mean":8,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_19_residual_pre_attn/max_abs":77,"train/tensor_act_model_layers_79_mlp_up_proj/max":4.1875,"train/tensor_act_model_layers_14_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_33_mlp_gate_proj/range":6.3359375,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_98_residual_post_attn/norm":802781.25,"train/tensor_param_model_layers_41_mlp_down_proj_weight/norm":6.0625,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_56_mlp_gate_proj/norm":5088.244140625,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/max_abs":0.1318359375,"train/tensor_act_model_layers_41_mlp/norm":8192,"train/tensor_act_model_layers_41_residual_post_attn/min":162,"train/tensor_act_model_layers_88_residual_post_attn/min":350,"train/tensor_param_model_layers_37_mlp_up_proj_weight/max_abs":0.12890625,"train/tensor_act_model_layers_56_residual_post_attn/std":0.408203125,"train/tensor_act_model_layers_88_mlp_up_proj/mean":-0.0693359375,"train/tensor_act_model_layers_7_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/max_abs":0.14453125,"train/tensor_act_model_layers_7_residual_post_attn/std":0.19921875,"train/tensor_act_model_layers_33_mlp_up_proj/std":1.90625,"train/tensor_act_model_layers_65_mlp_gate_proj/min":-1.2890625,"train/tensor_act_model_layers_20_residual_post_attn/max":81,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/range":0.2705078125,"train/tensor_act_model_layers_26_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_30_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_pre_attn/mean":256,"train/tensor_param_model_layers_59_mlp_down_proj_weight/min":-0.1318359375,"train/tensor_act_model_layers_84_residual_post_attn/max":336,"train/tensor_act_model_layers_87_mlp_gate_proj/max":5.09375,"train/tensor_act_model_layers_65_residual_post_attn/mean":260,"train/tensor_act_model_layers_75_residual_post_mlp/max":304,"train/tensor_param_model_layers_93_mlp_up_proj_weight/max":0.162109375,"train/tensor_act_model_layers_88_residual_post_mlp/norm":729053.25,"train/tensor_act_model_layers_46_mlp_down_proj/range":2.3125,"train/tensor_act_model_layers_41_residual_post_mlp/norm":343916.65625,"train/tensor_act_model_layers_5_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_97_mlp_up_proj_weight/norm":9.6875,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_54_mlp_up_proj_weight/max":0.126953125,"train/tensor_act_model_layers_33_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_98_mlp_gate_proj/std":1.6953125,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_70_mlp_gate_proj/mean":0.55078125,"train/tensor_act_model_layers_56_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_20_mlp_gate_proj/norm":5185.51904296875,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_98_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_10_mlp_up_proj_weight/max_abs":0.1123046875,"train/tensor_act_model_layers_66_mlp_up_proj/range":8.375,"train/tensor_act_model_layers_5_residual_pre_attn/norm":40939.65234375,"train/tensor_param_model_layers_32_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_94_residual_pre_attn/norm":770013.25,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/min":-0.125,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_residual_post_attn/min":67,"train/tensor_param_model_layers_86_mlp_down_proj_weight/std":0.04541015625,"train/tensor_param_model_layers_23_mlp_down_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_37_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_33_mlp/norm":8192,"train/tensor_act_model_layers_17_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_12_mlp_up_proj/min":-4.90625,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/range":0.23681640625,"train/tensor_param_model_layers_31_mlp_up_proj_weight/range":0.2578125,"train/tensor_act_model_layers_86_mlp/max_abs":4,"train/tensor_act_model_layers_31_residual_post_attn/max":125,"train/tensor_act_model_layers_37_residual_post_mlp/mean":152,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_90_mlp_down_proj/mean":15.1875,"train/tensor_act_model_layers_17_residual_post_mlp/norm":147258.328125,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_69_mlp_up_proj_weight/norm":6.9375,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/std":0.04052734375,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_56_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/max":0.126953125,"train/tensor_act_model_layers_85_mlp_gate_proj/std":1.6171875,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_63_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_47_mlp_gate_proj/norm":5241.56689453125,"train/tensor_act_model_layers_52_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_84_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_91_mlp_down_proj_weight/max_abs":0.142578125,"train/tensor_act_model_layers_13_mlp/norm":8192,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/std":0.032958984375,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/max_abs":0.1181640625,"train/tensor_param_model_layers_95_mlp_down_proj_weight/min":-0.1708984375,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_53_mlp/max":4,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_76_mlp_gate_proj/max_abs":5.46875,"train/tensor_param_model_layers_34_mlp_up_proj_weight/std":0.03271484375,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_0_mlp/range":0.296875,"train/tensor_act_model_layers_85_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_47_mlp_gate_proj/min":-1.265625,"train/tensor_act_model_layers_2_residual_post_mlp/max":12.4375,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_45_residual_pre_attn/max":181,"train/tensor_act_model_layers_14_residual_pre_attn/max":57,"train/tensor_act_model_layers_67_mlp_up_proj/mean":-0.123046875,"train/tensor_act_model_layers_14_mlp_up_proj/std":1.921875,"train/tensor_act_model_layers_39_mlp_up_proj/norm":5399.8388671875,"train/tensor_act_model_layers_71_mlp_up_proj/std":1.8125,"train/tensor_act_model_layers_19_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_69_mlp_up_proj_weight/std":0.038330078125,"train/tensor_act_model_layers_62_residual_post_mlp/range":3,"train/tensor_act_model_layers_29_mlp_up_proj/range":9.046875,"train/tensor_act_model_layers_64_mlp/norm":8192,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_56_residual_post_attn/max":225,"train/tensor_param_model_layers_13_mlp_down_proj_weight/mean":0.0011749267578125,"train/tensor_act_model_layers_21_mlp/min":4,"train/epoch":0.23592854735422986,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_57_mlp_up_proj_weight/mean":-0.000331878662109375,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_57_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_20_residual_post_attn/std":0.359375,"train/tensor_act_model_layers_82_mlp_gate_proj/mean":0.60546875,"train/tensor_act_model_layers_11_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_65_mlp_down_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_7_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_84_residual_pre_attn/mean":336,"train/tensor_act_model_layers_7_mlp_gate_proj/min":-1.3359375,"train/tensor_act_model_layers_78_mlp_down_proj/max":17,"train/tensor_act_model_layers_15_mlp/min":4,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/max":0.1953125,"train/tensor_act_model_layers_49_residual_post_attn/range":3,"train/tensor_act_model_layers_64_residual_post_attn/max":256,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_5_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_29_mlp_gate_proj/max":5.0625,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_35_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_56_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_12_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_70_residual_post_mlp/max_abs":284,"train/tensor_act_model_layers_67_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_34_mlp_gate_proj/std":1.6875,"train/tensor_act_model_layers_49_mlp_up_proj/min":-4.40625,"train/tensor_act_model_layers_58_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/min":-0.130859375,"train/tensor_act_model_layers_7_residual_pre_attn/max_abs":28.625,"train/tensor_act_model_layers_68_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_20_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_mlp_up_proj/norm":5331.75634765625,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp_up_proj/mean":0.0390625,"train/tensor_act_model_layers_72_residual_post_mlp/mean":292,"train/tensor_act_model_layers_4_residual_pre_attn/mean":16,"train/tensor_param_model_layers_85_mlp_down_proj_weight/range":0.2939453125,"train/tensor_param_model_layers_94_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_73_residual_post_mlp/mean":296,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_3_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_23_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_20_mlp_up_proj/norm":5484.4677734375,"train/tensor_act_model_layers_44_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_7_mlp_gate_proj/mean":0.376953125,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/range":0.3193359375,"train/tensor_act_model_layers_86_residual_pre_attn/max_abs":344,"train/tensor_act_model_layers_53_mlp/norm":8192,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_30_mlp/max_abs":4,"train/tensor_act_model_layers_84_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_51_mlp_up_proj/max_abs":4.90625,"train/tensor_param_model_layers_98_mlp_down_proj_weight/max_abs":0.1962890625,"train/tensor_act_model_layers_2_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_48_mlp_up_proj/max":4.8125,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_40_mlp_down_proj/range":2.25,"train/tensor_act_model_layers_44_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_28_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_94_mlp_up_proj/std":1.75,"train/tensor_param_model_layers_73_mlp_down_proj_weight/max":0.1357421875,"train/tensor_act_model_layers_95_mlp_gate_proj/range":7.8671875,"train/tensor_act_model_layers_74_mlp_gate_proj/min":-1.28125,"train/tensor_act_model_layers_98_mlp_up_proj/max_abs":5.28125,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_41_residual_pre_attn/min":162,"train/tensor_act_model_layers_25_residual_post_attn/norm":204561.203125,"train/tensor_act_model_layers_62_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/max":0.126953125,"train/tensor_act_model_layers_96_mlp_down_proj/mean":15.25,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_40_residual_post_mlp/range":3,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/range":0.3505859375,"train/tensor_act_model_layers_40_mlp/max_abs":4,"train/tensor_act_model_layers_36_mlp_down_proj/norm":30077.74609375,"train/tensor_param_model_layers_6_mlp_down_proj_weight/max_abs":0.10693359375,"train/tensor_act_model_layers_26_mlp_down_proj/max":16.125,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/max":0.134765625,"train/tensor_act_model_layers_99_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_41_residual_pre_attn/std":0.31640625,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/max":0.1318359375,"train/tensor_act_model_layers_5_mlp/max_abs":4,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_68_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_79_mlp_down_proj/norm":30620.7109375,"train/tensor_act_model_layers_78_mlp_down_proj/max_abs":17,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_mlp_down_proj/range":4.625,"train/tensor_param_model_layers_43_mlp_down_proj_weight/std":0.03369140625,"train/tensor_param_model_layers_0_mlp_down_proj_weight/min":-0.125,"train/tensor_act_model_layers_84_residual_pre_attn/min":334,"train/tensor_act_model_layers_78_residual_pre_attn/norm":638941.25,"train/tensor_param_model_layers_72_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_25_mlp_gate_proj/min":-1.3046875,"train/tensor_act_model_layers_66_mlp_gate_proj/max":4.4375,"train/tensor_act_model_layers_93_residual_post_attn/range":2,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_99_mlp_gate_proj/range":7.3125,"train/tensor_act_model_layers_79_residual_pre_attn/norm":647133.3125,"train/tensor_param_model_layers_17_mlp_down_proj_weight/norm":5.59375,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_47_mlp_down_proj/max_abs":16,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_1_mlp_up_proj/norm":5277.21630859375,"train/tensor_act_model_layers_94_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_residual_post_attn/min":7.46875,"train/tensor_act_model_layers_42_mlp_gate_proj/mean":0.55859375,"train/tensor_act_model_layers_20_residual_pre_attn/mean":80,"train/tensor_param_model_layers_64_mlp_down_proj_weight/min":-0.1328125,"train/tensor_act_model_layers_12_residual_post_attn/min":47,"train/tensor_act_model_layers_93_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/max":0.11376953125,"train/tensor_act_model_layers_22_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_35_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_15_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_10_mlp_down_proj/range":1.875,"train/tensor_act_model_layers_87_mlp_up_proj/range":9.5625,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/std":0.0299072265625,"train/tensor_act_model_layers_45_mlp_up_proj/min":-4.75,"train/tensor_act_model_layers_56_residual_pre_attn/mean":224,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_26_residual_post_mlp/range":2.5,"train/tensor_param_model_layers_99_mlp_up_proj_weight/mean":-0.00083160400390625,"train/tensor_act_model_layers_56_mlp_up_proj/min":-4.125,"train/tensor_act_model_layers_82_mlp/max":4,"train/tensor_act_model_layers_89_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_64_residual_post_attn/std":0.37109375,"train/tensor_act_model_layers_75_mlp_gate_proj/range":5.6640625,"train/tensor_param_model_layers_54_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_mlp_gate_proj/norm":4953.912109375,"train/tensor_act_model_layers_21_mlp/mean":4,"train/tensor_act_model_layers_32_mlp/max_abs":4,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/min":-0.138671875,"train/tensor_act_model_layers_41_residual_post_attn/mean":164,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_12_residual_post_mlp/norm":106329.78125,"train/tensor_act_model_layers_73_residual_post_mlp/range":2,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_47_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_91_residual_pre_attn/norm":745437.25,"train/tensor_act_model_layers_76_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_41_mlp_up_proj/max":3.90625,"train/tensor_act_model_layers_72_mlp_gate_proj/range":6.1484375,"train/tensor_param_model_layers_17_mlp_down_proj_weight/max_abs":0.11669921875,"train/tensor_act_model_layers_95_residual_pre_attn/range":2,"train/tensor_act_model_layers_64_residual_post_attn/norm":524001.8125,"train/tensor_act_model_layers_34_residual_pre_attn/norm":278380.8125,"train/tensor_act_model_layers_47_mlp/range":0,"train/tensor_act_model_layers_79_residual_post_mlp/max":320,"train/tensor_param_model_layers_29_mlp_down_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_28_mlp_gate_proj/std":1.7265625,"train/tensor_act_model_layers_54_residual_post_mlp/std":0.408203125,"train/tensor_param_model_layers_47_mlp_down_proj_weight/range":0.24560546875,"train/tensor_act_model_layers_34_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_31_residual_pre_attn/mean":124,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/range":0.2568359375,"train/tensor_act_model_layers_46_mlp_down_proj/norm":30148.970703125,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_81_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_85_residual_post_attn/norm":696285.25,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_63_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_30_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_66_residual_post_attn/mean":264,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/range":0.3291015625,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_96_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_68_mlp_up_proj/max_abs":4.78125,"train/tensor_act_model_layers_98_mlp_up_proj/min":-5.28125,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_91_mlp_up_proj/norm":5328.90576171875,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_22_mlp/min":4,"train/tensor_act_model_layers_58_mlp_gate_proj/max":5.21875,"train/tensor_param_model_layers_18_mlp_down_proj_weight/std":0.0308837890625,"train/tensor_act_model_layers_63_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_99_mlp_gate_proj/min":-1.46875,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_52_mlp_down_proj_weight/max":0.1279296875,"train/tensor_act_model_layers_12_mlp/max_abs":4,"train/tensor_act_model_layers_91_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_mlp_up_proj/norm":5398.25341796875,"train/tensor_param_model_layers_28_mlp_up_proj_weight/min":-0.1259765625,"train/tensor_param_model_layers_17_mlp_up_proj_weight/mean":-0.0003509521484375,"train/tensor_act_model_layers_62_mlp_gate_proj/range":5.9453125,"train/tensor_param_model_layers_11_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp/max_abs":4,"train/tensor_act_model_layers_24_mlp/mean":4,"train/tensor_param_model_layers_10_mlp_down_proj_weight/norm":5.34375,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_93_mlp_gate_proj/std":1.671875,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_98_residual_post_attn/max":392,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_27_mlp_up_proj/max_abs":4.8125,"train/tensor_act_model_layers_89_mlp_gate_proj/min":-1.3515625,"train/tensor_act_model_layers_94_mlp_up_proj/max_abs":5.25,"train/tensor_act_model_layers_57_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_47_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_1_residual_pre_attn/min":3.578125,"train/tensor_act_model_layers_86_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_act_model_layers_20_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_64_mlp_gate_proj/mean":0.65234375,"train/tensor_act_model_layers_80_residual_pre_attn/mean":320,"train/tensor_param_model_layers_94_mlp_up_proj_weight/mean":0.00086212158203125,"train/tensor_param_model_layers_21_mlp_down_proj_weight/range":0.234375,"train/tensor_act_model_layers_17_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_26_mlp/max":4,"train/tensor_param_model_layers_77_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_39_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_88_residual_pre_attn/norm":720861.25,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_41_residual_pre_attn/norm":335724.6875,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/min":-0.125,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/min":-0.130859375,"train/tensor_act_model_layers_62_residual_post_mlp/min":250,"train/tensor_act_model_layers_2_residual_post_attn/norm":16363.89453125,"train/tensor_act_model_layers_1_mlp_up_proj/range":8.9375,"train/tensor_act_model_layers_30_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_32_mlp_gate_proj/norm":5281.70751953125,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_89_mlp_down_proj_weight/max":0.1455078125,"train/tensor_act_model_layers_40_mlp/min":4,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_73_mlp_up_proj_weight/max_abs":0.1357421875,"train/tensor_act_model_layers_56_mlp/norm":8192,"train/tensor_act_model_layers_9_mlp_up_proj/range":8.8125,"train/tensor_act_model_layers_61_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_85_mlp_up_proj_weight/std":0.04541015625,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/range":0.3046875,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/min":-0.17578125,"train/tensor_act_model_layers_25_residual_post_attn/mean":100,"train/tensor_param_model_layers_86_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_6_mlp_down_proj_weight/range":0.20849609375,"train/tensor_act_model_layers_48_mlp_gate_proj/mean":0.62890625,"train/tensor_param_model_layers_9_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_62_mlp/max":4,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_89_mlp_down_proj/max_abs":16.625,"train/tensor_act_model_layers_3_residual_pre_attn/range":0.9375,"train/tensor_param_model_layers_40_mlp_down_proj_weight/norm":6.09375,"train/tensor_act_model_layers_63_mlp_gate_proj/mean":0.5546875,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_53_residual_post_attn/max":213,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/range":0.2880859375,"train/tensor_act_model_layers_28_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_43_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_mlp/std":0,"train/tensor_param_model_layers_40_mlp_up_proj_weight/min":-0.123046875,"train/tensor_act_model_layers_47_mlp/mean":4,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/range":0.25,"train/tensor_act_model_layers_0_residual_pre_attn/max":0.28515625,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_84_mlp_up_proj/range":9.1875,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_param_model_layers_36_mlp_down_proj_weight/range":0.23193359375,"train/tensor_act_model_layers_54_mlp_up_proj/min":-4.46875,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_6_mlp/max":4,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_54_mlp_gate_proj/std":1.6328125,"train/tensor_act_model_layers_34_mlp/range":0,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_18_residual_post_mlp/mean":76,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/norm":7.25,"train/tensor_act_model_layers_61_mlp_gate_proj/min":-1.265625,"train/tensor_param_model_layers_83_mlp_up_proj_weight/max":0.140625,"train/tensor_act_model_layers_57_residual_pre_attn/mean":228,"train/tensor_act_model_layers_4_residual_post_attn/min":15.3125,"train/tensor_act_model_layers_85_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_45_mlp_gate_proj/norm":5121.10205078125,"train/tensor_act_model_layers_4_mlp_gate_proj/mean":0.51953125,"train/tensor_act_model_layers_52_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_42_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_85_residual_post_attn/mean":340,"train/tensor_act_model_layers_24_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_66_mlp_down_proj_weight/mean":-1.9311904907226562e-05,"train/tensor_act_model_layers_66_residual_post_attn/norm":540637.3125,"train/tensor_act_model_layers_55_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_38_mlp_up_proj/std":1.8671875,"train/tensor_act_model_layers_57_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_post_attn/mean":256,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_59_residual_post_mlp/mean":240,"train/tensor_act_model_layers_20_mlp_gate_proj/max_abs":4.96875,"train/tensor_act_model_layers_39_mlp/max_abs":4,"train/tensor_act_model_layers_71_mlp_gate_proj/norm":4978.2919921875,"train/tensor_param_model_layers_49_mlp_up_proj_weight/std":0.03564453125,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/mean":0.0052490234375,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_76_mlp_down_proj/norm":30904.021484375,"train/tensor_act_model_layers_27_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_41_residual_post_attn/norm":335724.6875,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_53_residual_post_attn/std":0.408203125,"train/tensor_act_model_layers_47_mlp_up_proj/norm":5395.5556640625,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_77_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_35_residual_post_mlp/range":3,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_87_mlp/mean":4,"train/tensor_act_model_layers_94_residual_post_mlp/norm":778205.25,"train/tensor_act_model_layers_2_residual_pre_attn/mean":8,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/min":-0.11279296875,"train/tensor_act_model_layers_90_mlp_up_proj/range":9.15625,"train/tensor_act_model_layers_58_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_70_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_residual_post_mlp/max_abs":113,"train/tensor_act_model_layers_23_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_79_mlp_down_proj/range":2.25,"train/tensor_act_model_layers_26_mlp/mean":4,"train/tensor_act_model_layers_81_mlp/min":4,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_32_residual_pre_attn/range":2.5,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_0_mlp_gate_proj/max_abs":7.46875,"train/tensor_act_model_layers_67_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_94_residual_post_attn/min":374,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_70_mlp_down_proj_weight/norm":6.90625,"train/tensor_param_model_layers_89_mlp_up_proj_weight/range":0.2978515625,"train/tensor_act_model_layers_33_mlp/max_abs":4,"train/tensor_act_model_layers_17_residual_pre_attn/norm":139066.421875,"train/tensor_param_model_layers_0_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_84_mlp_up_proj/max_abs":4.75,"train/tensor_act_model_layers_81_mlp_up_proj/max_abs":5,"train/tensor_act_model_layers_44_residual_post_mlp/std":0.31640625,"train/tensor_param_model_layers_68_mlp_down_proj_weight/mean":0.0004444122314453125,"train/tensor_act_model_layers_65_residual_post_attn/norm":532445.3125,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_31_mlp_up_proj_weight/std":0.0322265625,"train/tensor_act_model_layers_0_residual_post_attn/max":0.3046875,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_53_residual_pre_attn/mean":212,"train/tensor_act_model_layers_75_residual_pre_attn/max_abs":300,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/range":0.2578125,"train/tensor_param_model_layers_81_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_78_residual_post_attn/min":310,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/mean":0.004180908203125,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_15_mlp_up_proj_weight/norm":5.5,"train/tensor_act_model_layers_50_mlp/range":0,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_6_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_59_mlp_down_proj/range":3.0625,"train/tensor_act_model_layers_66_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_63_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_3_mlp_up_proj_weight/norm":5.5,"train/tensor_param_model_layers_53_mlp_down_proj_weight/mean":-0.0004673004150390625,"train/tensor_act_model_layers_79_residual_post_attn/min":314,"train/tensor_act_model_layers_83_mlp_up_proj/max_abs":4.21875,"train/tensor_act_model_layers_84_mlp_gate_proj/max_abs":5,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/min":-0.189453125,"train/tensor_param_model_layers_27_mlp_up_proj_weight/range":0.2392578125,"train/tensor_act_model_layers_79_residual_post_mlp/norm":655325.25,"train/tensor_act_model_layers_50_mlp/min":4,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_65_mlp_up_proj/max":4.28125,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_25_mlp_up_proj/mean":-0.07421875,"train/tensor_act_model_layers_90_mlp_gate_proj/max":4.75,"train/tensor_act_model_layers_33_residual_post_mlp/range":3,"train/tensor_param_model_layers_46_mlp_down_proj_weight/range":0.2529296875,"train/tensor_act_model_layers_4_mlp_down_proj/max":15.9375,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_19_mlp_up_proj/mean":-0.047119140625,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_7_residual_pre_attn/max":28.625,"train/tensor_act_model_layers_78_residual_post_mlp/norm":647133.3125,"train/tensor_act_model_layers_98_residual_post_attn/mean":392,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_10_residual_post_mlp/mean":44,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/mean":0.00482177734375,"train/tensor_act_model_layers_88_residual_post_attn/mean":352,"train/tensor_act_model_layers_35_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_48_mlp_down_proj/max_abs":17,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_9_residual_post_attn/max":36.75,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_57_residual_post_attn/std":0.416015625,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_99_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_66_mlp_up_proj_weight/norm":6.84375,"train/tensor_act_model_layers_23_mlp_up_proj/mean":0.189453125,"train/tensor_act_model_layers_41_mlp/std":0,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_34_mlp_up_proj/norm":5468.9638671875,"train/tensor_act_model_layers_55_residual_pre_attn/range":3,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/min":-0.142578125,"train/tensor_param_model_layers_91_mlp_up_proj_weight/mean":0.0006103515625,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/mean":0.0037078857421875,"train/tensor_act_model_layers_4_mlp_up_proj/std":1.859375,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_92_mlp_up_proj_weight/max":0.171875,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/mean":-0.007354736328125,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_36_residual_pre_attn/max_abs":145,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_21_mlp_gate_proj/std":1.71875,"train/tensor_act_model_layers_44_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/mean":0.004302978515625,"train/tensor_act_model_layers_44_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_46_mlp_up_proj/norm":5375.4287109375,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_56_mlp/mean":4,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_89_mlp_up_proj/norm":5211.37353515625,"train/tensor_act_model_layers_61_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_8_mlp_gate_proj/std":1.6953125,"train/tensor_act_model_layers_7_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_11_mlp/mean":4,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_91_mlp_up_proj_weight/min":-0.14453125,"train/tensor_act_model_layers_26_mlp_down_proj/std":0.40234375,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_residual_post_mlp/max":189,"train/tensor_param_model_layers_99_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_22_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_84_mlp_gate_proj/max":5,"train/tensor_param_model_layers_7_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_86_mlp/min":4,"train/tensor_act_model_layers_49_mlp_down_proj/range":1.9375,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/max":0.11962890625,"train/tensor_param_model_layers_5_mlp_up_proj_weight/max":0.1103515625,"train/tensor_act_model_layers_65_mlp_gate_proj/mean":0.38671875,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/norm":6.3125,"train/tensor_act_model_layers_3_mlp_down_proj/range":2.25,"train/tensor_act_model_layers_37_residual_post_mlp/max_abs":153,"train/tensor_act_model_layers_62_mlp_down_proj/max_abs":16.875,"train/tensor_act_model_layers_34_residual_post_mlp/norm":286572.78125,"train/tensor_act_model_layers_55_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_39_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_34_residual_post_attn/min":134,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_39_mlp_up_proj_weight/norm":6.125,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_61_mlp_down_proj/norm":30503.017578125,"train/tensor_param_model_layers_41_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_53_residual_post_mlp/max_abs":217,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_19_mlp_down_proj_weight/max":0.10888671875,"train/tensor_act_model_layers_23_mlp_down_proj/max_abs":16,"train/tensor_param_model_layers_42_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_24_mlp_down_proj/range":2.375,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_48_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_90_mlp/std":0,"train/tensor_act_model_layers_40_residual_pre_attn/max_abs":161,"train/tensor_param_model_layers_8_mlp_up_proj_weight/max":0.11962890625,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/max_abs":0.1259765625,"train/tensor_param_model_layers_3_mlp_down_proj_weight/norm":5.40625,"train/tensor_act_model_layers_91_residual_post_attn/min":362,"train/tensor_act_model_layers_18_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_0_mlp_up_proj_weight/range":0.25,"train/tensor_act_model_layers_68_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_65_mlp_gate_proj/range":6.2578125,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_93_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_77_mlp_up_proj/std":1.828125,"train/tensor_act_model_layers_61_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_12_mlp/norm":8192,"train/tensor_act_model_layers_92_mlp_up_proj/mean":-0.01312255859375,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_87_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_60_mlp/std":0,"train/tensor_act_model_layers_12_residual_pre_attn/norm":98169.8359375,"train/tensor_act_model_layers_78_mlp_down_proj/std":0.55859375,"train/tensor_param_model_layers_72_mlp_up_proj_weight/mean":-0.002197265625,"train/tensor_act_model_layers_91_residual_post_mlp/norm":753629.25,"train/tensor_act_model_layers_57_residual_post_attn/max":229,"train/tensor_act_model_layers_88_residual_post_mlp/min":354,"train/tensor_act_model_layers_14_mlp/std":0,"train/tensor_act_model_layers_18_residual_pre_attn/max":73,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/min":-0.1845703125,"train/tensor_act_model_layers_12_residual_post_mlp/min":51,"train/tensor_act_model_layers_64_mlp_gate_proj/std":1.609375,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_59_mlp_down_proj/mean":14.8125,"train/tensor_act_model_layers_85_mlp/range":0,"train/tensor_act_model_layers_36_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_21_residual_post_attn/mean":84,"train/tensor_act_model_layers_85_mlp/norm":8192,"train/tensor_act_model_layers_53_residual_pre_attn/range":3,"train/tensor_act_model_layers_52_mlp_down_proj/max_abs":16.625,"train/tensor_act_model_layers_21_mlp_gate_proj/mean":0.6015625,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_9_residual_post_mlp/mean":40,"train/tensor_act_model_layers_76_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_76_residual_post_mlp/min":306,"train/tensor_act_model_layers_67_mlp_up_proj/range":8.90625,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_42_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_38_residual_pre_attn/range":3,"train/tensor_act_model_layers_93_residual_post_mlp/max":376,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_79_mlp_up_proj_weight/range":0.2744140625,"train/tensor_act_model_layers_91_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_61_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp/max":4,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_60_mlp/max_abs":4,"train/tensor_param_model_layers_61_mlp_up_proj_weight/max_abs":0.1318359375,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_21_residual_post_mlp/mean":88,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_42_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_44_mlp_gate_proj/norm":5280.20947265625,"train/tensor_act_model_layers_5_residual_post_mlp/min":23.25,"train/tensor_act_model_layers_86_mlp_gate_proj/max":4.6875,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_51_mlp_down_proj/std":0.4296875,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_69_mlp_up_proj/max":3.859375,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_71_mlp_up_proj_weight/mean":0.0001773834228515625,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_14_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_53_mlp_down_proj/std":0.46484375,"train/tensor_act_model_layers_93_mlp/norm":8192,"train/tensor_act_model_layers_48_mlp_gate_proj/min":-1.234375,"train/tensor_act_model_layers_32_mlp/range":0,"train/tensor_act_model_layers_76_mlp_down_proj/mean":15.0625,"train/tensor_act_model_layers_37_mlp/max":4,"train/tensor_act_model_layers_66_mlp_gate_proj/mean":0.451171875,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/max":0.1240234375,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_59_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_23_mlp_down_proj/norm":30111.484375,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_0_residual_post_mlp/mean":4,"train/tensor_act_model_layers_97_residual_pre_attn/max_abs":388,"train/tensor_act_model_layers_78_residual_pre_attn/max_abs":312,"train/tensor_act_model_layers_60_mlp_down_proj/range":2.25,"train/tensor_act_model_layers_17_mlp_down_proj/std":0.365234375,"train/tensor_act_model_layers_89_mlp/min":4,"train/tensor_act_model_layers_3_residual_post_attn/std":0.1572265625,"train/tensor_act_model_layers_8_mlp/max":4,"train/tensor_act_model_layers_72_mlp_up_proj/std":1.7421875,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_96_residual_post_attn/max":384,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_76_mlp_gate_proj/min":-1.296875,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_78_residual_pre_attn/range":2,"train/tensor_param_model_layers_79_mlp_down_proj_weight/mean":-0.0009765625,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_90_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_60_mlp_down_proj/max":16.5,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_24_mlp_down_proj_weight/range":0.24658203125,"train/tensor_param_model_layers_39_mlp_down_proj_weight/std":0.033447265625,"train/tensor_act_model_layers_56_residual_pre_attn/range":3,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_89_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_25_mlp_up_proj_weight/min":-0.125,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/norm":7.96875,"train/tensor_param_model_layers_30_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_residual_post_attn/max_abs":173,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_47_residual_pre_attn/mean":188,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/std":0.02978515625,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/max":0.1328125,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_6_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_34_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_1_mlp_up_proj/max_abs":4.625,"train/tensor_act_model_layers_39_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp_up_proj/range":7.734375,"train/tensor_act_model_layers_24_residual_pre_attn/min":94.5,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/max":0.126953125,"train/tensor_param_model_layers_97_mlp_down_proj_weight/max_abs":0.181640625,"train/tensor_act_model_layers_49_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_mlp_up_proj/std":1.7890625,"train/tensor_act_model_layers_1_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_38_mlp_up_proj_weight/mean":0.00057220458984375,"train/tensor_act_model_layers_8_mlp_up_proj/std":1.875,"train/tensor_act_model_layers_44_mlp_gate_proj/max_abs":5.03125,"train/tensor_act_model_layers_0_residual_post_attn/norm":187.2716522216797,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_82_mlp_down_proj/range":4.125,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_14_mlp/max":4,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/range":0.26171875,"train/tensor_act_model_layers_7_mlp_down_proj/min":14.0625,"train/tensor_param_model_layers_20_mlp_up_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_87_residual_post_attn/norm":712669.25,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/min":-0.1220703125,"train/tensor_param_model_layers_2_mlp_down_proj_weight/std":0.030517578125,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_72_mlp_gate_proj/max_abs":4.875,"train/tensor_act_model_layers_95_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_24_residual_pre_attn/std":0.359375,"train/tensor_act_model_layers_45_residual_pre_attn/norm":368492.625,"train/tensor_act_model_layers_55_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_43_mlp_gate_proj/max":4.65625,"train/tensor_act_model_layers_8_mlp/mean":4,"train/tensor_act_model_layers_65_residual_post_attn/range":2,"train/tensor_act_model_layers_97_mlp/max":4,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/min":-0.11669921875,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_67_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_mlp/mean":4,"train/tensor_param_model_layers_40_mlp_up_proj_weight/max_abs":0.12451171875,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_69_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_81_mlp_up_proj_weight/std":0.0439453125,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_98_mlp_up_proj_weight/min":-0.1982421875,"train/tensor_act_model_layers_14_residual_post_mlp/max_abs":61,"train/tensor_act_model_layers_30_residual_pre_attn/mean":120,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/mean":0.0032501220703125,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_73_mlp_down_proj_weight/range":0.267578125,"train/tensor_act_model_layers_48_mlp/max_abs":4,"train/tensor_act_model_layers_73_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_29_mlp_up_proj_weight/norm":5.8125,"train/tensor_param_model_layers_10_mlp_up_proj_weight/max":0.1123046875,"train/tensor_act_model_layers_6_mlp_up_proj/std":1.8828125,"train/tensor_act_model_layers_27_mlp_up_proj/std":1.859375,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/max":0.1748046875,"train/tensor_param_model_layers_41_mlp_up_proj_weight/std":0.03369140625,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_48_residual_pre_attn/mean":192,"train/tensor_param_model_layers_89_mlp_down_proj_weight/min":-0.138671875,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_6_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_48_mlp_up_proj_weight/mean":0.0003910064697265625,"train/tensor_param_model_layers_15_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_41_mlp_gate_proj/min":-1.3359375,"train/tensor_param_model_layers_66_mlp_down_proj_weight/std":0.03759765625,"train/tensor_param_model_layers_59_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_84_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_92_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_80_mlp_up_proj/mean":-0.001678466796875,"train/tensor_act_model_layers_50_mlp_gate_proj/max_abs":4.125,"train/tensor_act_model_layers_94_mlp/norm":8192,"train/tensor_act_model_layers_53_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_27_residual_pre_attn/norm":220945.109375,"train/tensor_param_model_layers_73_mlp_up_proj_weight/min":-0.1318359375,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/range":0.2685546875,"train/tensor_act_model_layers_87_mlp_up_proj/norm":5169.3525390625,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_96_residual_pre_attn/mean":384,"train/tensor_act_model_layers_48_mlp/max":4,"train/tensor_param_model_layers_85_mlp_up_proj_weight/mean":0.000766754150390625,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_3_mlp_up_proj_weight/mean":0.00029754638671875,"train/tensor_act_model_layers_30_residual_post_attn/std":0.388671875,"train/tensor_act_model_layers_80_mlp/norm":8192,"train/tensor_act_model_layers_12_mlp_up_proj/max_abs":4.90625,"train/tensor_act_model_layers_69_mlp/max_abs":4,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/range":0.224609375,"train/tensor_param_model_layers_47_mlp_up_proj_weight/max_abs":0.13671875,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_77_residual_post_attn/max_abs":308,"train/tensor_param_model_layers_50_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/mean":0.006011962890625,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/max_abs":0.130859375,"train/tensor_act_model_layers_60_mlp_gate_proj/norm":5148.857421875,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_39_mlp/max":4,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_1_mlp_down_proj_weight/norm":6.1875,"train/tensor_param_model_layers_17_mlp_up_proj_weight/max_abs":0.11572265625,"train/tensor_act_model_layers_29_mlp/mean":4,"train/tensor_act_model_layers_48_residual_post_attn/std":0.31640625,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_31_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_mlp/min":4,"train/tensor_act_model_layers_61_mlp_down_proj/mean":14.875,"train/tensor_act_model_layers_43_mlp_up_proj/range":8.9375,"train/tensor_act_model_layers_36_residual_post_attn/max_abs":145,"train/tensor_act_model_layers_28_residual_pre_attn/norm":229137.046875,"train/tensor_act_model_layers_37_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_70_residual_pre_attn/min":278,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_60_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_55_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_52_mlp_up_proj_weight/mean":-0.000812530517578125,"train/tensor_act_model_layers_86_mlp_up_proj/max_abs":4.96875,"train/tensor_act_model_layers_27_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_60_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_74_residual_post_mlp/norm":614365.25,"train/tensor_act_model_layers_28_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_62_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_36_mlp_up_proj_weight/max":0.126953125,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/max_abs":0.17578125,"train/tensor_act_model_layers_1_residual_post_attn/max":4.40625,"train/tensor_act_model_layers_47_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_18_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_49_residual_pre_attn/std":0.31640625,"train/tensor_param_model_layers_13_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_77_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_86_mlp_gate_proj/norm":5058.80322265625,"train/tensor_act_model_layers_77_mlp_gate_proj/norm":5022.517578125,"train/tensor_act_model_layers_31_residual_post_attn/mean":124,"train/tensor_act_model_layers_53_mlp_gate_proj/max_abs":5.0625,"train/tensor_param_model_layers_38_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_28_mlp_up_proj/std":1.9296875,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_49_residual_pre_attn/min":194,"train/tensor_act_model_layers_15_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_4_mlp_down_proj_weight/max_abs":0.11376953125,"train/tensor_param_model_layers_19_mlp_down_proj_weight/min":-0.12158203125,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/max_abs":0.119140625,"train/tensor_act_model_layers_59_mlp_down_proj/norm":30407.5234375,"train/tensor_act_model_layers_69_mlp_gate_proj/range":5.34375,"train/tensor_act_model_layers_5_mlp_gate_proj/mean":0.57421875,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_93_mlp_gate_proj/range":7.15625,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_9_mlp_down_proj/range":2.1875,"train/tensor_act_model_layers_53_mlp_up_proj/norm":5374.7548828125,"train/tensor_act_model_layers_1_residual_post_attn/range":0.921875,"train/tensor_act_model_layers_0_mlp_gate_proj/std":1.859375,"train/tensor_act_model_layers_49_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_30_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_74_residual_post_attn/min":294,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_50_mlp_up_proj_weight/mean":0.00013637542724609375,"train/tensor_param_model_layers_0_mlp_up_proj_weight/max":0.125,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_60_residual_pre_attn/norm":491292.78125,"train/tensor_act_model_layers_48_residual_post_attn/max":193,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/max_abs":0.134765625,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/max":0.212890625,"train/tensor_act_model_layers_35_residual_post_mlp/max":145,"train/tensor_act_model_layers_29_residual_post_mlp/norm":245497.171875,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_1_mlp_down_proj/max":20.625,"train/tensor_act_model_layers_15_residual_post_mlp/std":0.298828125,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_13_residual_post_mlp/max":57,"train/tensor_act_model_layers_81_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/max_abs":0.12890625,"train/tensor_param_model_layers_59_mlp_down_proj_weight/range":0.271484375,"train/tensor_act_model_layers_43_mlp_gate_proj/range":5.9140625,"train/tensor_act_model_layers_16_mlp/max":4,"train/tensor_act_model_layers_55_residual_post_mlp/max_abs":225,"train/tensor_act_model_layers_50_mlp/std":0,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_98_mlp/mean":4,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_20_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_70_mlp_down_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_9_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_11_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_91_residual_post_mlp/range":2,"train/tensor_act_model_layers_57_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_93_residual_post_attn/max_abs":372,"train/tensor_act_model_layers_44_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/max_abs":0.1328125,"train/tensor_param_model_layers_19_mlp_up_proj_weight/max_abs":0.140625,"train/tensor_act_model_layers_29_residual_post_attn/min":114.5,"train/tensor_act_model_layers_55_mlp_gate_proj/range":6.0078125,"train/tensor_act_model_layers_83_residual_post_attn/range":2,"train/tensor_act_model_layers_66_mlp_up_proj/max":4.21875,"train/tensor_act_model_layers_97_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_34_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_57_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_65_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_92_mlp_gate_proj/mean":0.68359375,"train/tensor_act_model_layers_88_mlp_gate_proj/min":-1.4453125,"train/tensor_act_model_layers_30_residual_post_attn/max":121,"train/tensor_act_model_layers_86_mlp_up_proj/range":9.625,"train/tensor_act_model_layers_23_mlp_up_proj/max":4.84375,"train/tensor_act_model_layers_24_mlp_up_proj/max":4.15625,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_14_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_32_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_76_residual_post_mlp/norm":630749.3125,"train/tensor_act_model_layers_91_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_46_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_16_residual_post_attn/mean":63.75,"train/tensor_act_model_layers_87_mlp/std":0,"train/tensor_act_model_layers_83_mlp_down_proj/std":0.77734375,"train/tensor_act_model_layers_91_mlp_gate_proj/min":-1.3359375,"train/tensor_act_model_layers_17_mlp/std":0,"train/tensor_param_model_layers_13_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_39_mlp_gate_proj/std":1.6328125,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_73_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_29_mlp/max_abs":4,"train/tensor_act_model_layers_82_mlp_up_proj/norm":5139.47900390625,"train/tensor_act_model_layers_42_residual_pre_attn/max":169,"train/tensor_param_model_layers_93_mlp_down_proj_weight/range":0.326171875,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_18_mlp/std":0,"train/tensor_act_model_layers_86_mlp_up_proj/norm":5252.0302734375,"train/tensor_act_model_layers_55_mlp/norm":8192,"train/tensor_act_model_layers_47_residual_post_attn/max":189,"train/tensor_act_model_layers_25_mlp_up_proj/max_abs":4.1875,"train/tensor_act_model_layers_76_residual_post_attn/min":302,"train/tensor_act_model_layers_75_residual_pre_attn/range":2,"train/tensor_act_model_layers_83_residual_post_attn/norm":679901.25,"train/tensor_act_model_layers_67_mlp_gate_proj/max":4.5625,"train/tensor_act_model_layers_48_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_35_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/max":0.1337890625,"train/tensor_act_model_layers_66_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_80_mlp_gate_proj/norm":4933.34423828125,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_15_mlp_up_proj_weight/std":0.0303955078125,"train/tensor_act_model_layers_63_residual_pre_attn/max_abs":253,"train/tensor_act_model_layers_29_residual_post_attn/norm":237305.21875,"train/tensor_act_model_layers_9_mlp_down_proj/max":16.25,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_34_mlp_up_proj_weight/range":0.236328125,"train/tensor_act_model_layers_26_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_1_residual_post_mlp/mean":8,"train/tensor_act_model_layers_38_mlp_gate_proj/range":5.9296875,"train/tensor_act_model_layers_50_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_9_mlp/std":0,"train/tensor_act_model_layers_86_residual_pre_attn/max":344,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_35_mlp_gate_proj/norm":5024.4677734375,"train/tensor_param_model_layers_56_mlp_up_proj_weight/norm":6.65625,"train/tensor_param_model_layers_42_mlp_down_proj_weight/max_abs":0.126953125,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_91_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_post_attn/max_abs":256,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/max":0.12255859375,"train/tensor_act_model_layers_12_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/std":0.03173828125,"train/tensor_act_model_layers_77_residual_pre_attn/mean":308,"train/tensor_act_model_layers_14_residual_pre_attn/mean":56,"train/tensor_act_model_layers_3_mlp/mean":4,"train/tensor_act_model_layers_85_mlp_down_proj/mean":15.125,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_89_residual_post_attn/max_abs":356,"train/tensor_act_model_layers_72_mlp/std":0,"train/tensor_act_model_layers_41_mlp_gate_proj/norm":5062.22607421875,"train/tensor_act_model_layers_90_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_mlp_gate_proj/range":5.5859375,"train/tensor_param_model_layers_26_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_26_mlp/norm":8192,"train/tensor_act_model_layers_5_mlp/min":4,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_52_mlp_up_proj/max":4.8125,"train/tensor_param_model_layers_89_mlp_up_proj_weight/min":-0.1416015625,"train/tensor_param_model_layers_5_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_84_residual_post_mlp/max":340,"train/tensor_param_model_layers_17_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_50_residual_post_mlp/min":202,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_28_mlp_up_proj_weight/norm":5.875,"train/tensor_act_model_layers_37_mlp_down_proj/range":2,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_87_mlp_down_proj_weight/norm":8.125,"train/tensor_param_model_layers_3_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_49_residual_post_attn/mean":196,"train/tensor_act_model_layers_37_mlp/mean":4,"train/tensor_act_model_layers_5_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_68_mlp_up_proj/range":8.578125,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_85_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/norm":7.78125,"train/tensor_param_model_layers_82_mlp_down_proj_weight/range":0.27734375,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_28_mlp_up_proj_weight/max_abs":0.1259765625,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/mean":0.003173828125,"train/tensor_act_model_layers_10_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_81_mlp_up_proj/min":-5,"train/tensor_act_model_layers_2_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_mlp_up_proj/norm":5391.111328125,"train/tensor_act_model_layers_75_residual_post_attn/max_abs":300,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_mlp/min":4,"train/tensor_act_model_layers_86_residual_post_mlp/mean":348,"train/tensor_act_model_layers_28_mlp_down_proj/max_abs":16.25,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_58_mlp_up_proj/min":-4.375,"train/tensor_act_model_layers_55_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_19_mlp_down_proj_weight/range":0.23046875,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_1_mlp_gate_proj/norm":4974.03515625,"train/tensor_act_model_layers_22_mlp_gate_proj/range":5.8203125,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_42_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_21_mlp/max":4,"train/tensor_act_model_layers_52_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_68_residual_post_attn/mean":272,"train/tensor_param_model_layers_64_mlp_down_proj_weight/norm":6.78125,"train/tensor_param_model_layers_38_mlp_down_proj_weight/norm":6.0625,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_30_mlp_gate_proj/range":5.6875,"train/tensor_param_model_layers_13_mlp_down_proj_weight/max_abs":0.1171875,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/std":0.05712890625,"train/tensor_act_model_layers_67_mlp_up_proj/std":1.8359375,"train/tensor_param_model_layers_28_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/range":0.2470703125,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_29_mlp_up_proj_weight/max_abs":0.11279296875,"train/tensor_param_model_layers_96_mlp_up_proj_weight/max_abs":0.1953125,"train/tensor_act_model_layers_11_mlp_down_proj/std":0.306640625,"train/tensor_act_model_layers_82_mlp_down_proj/std":0.7265625,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_19_residual_post_mlp/range":2.5,"train/tensor_param_model_layers_66_mlp_down_proj_weight/max":0.1279296875,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_41_mlp_up_proj/mean":-0.0228271484375,"train/tensor_act_model_layers_34_residual_post_mlp/min":138,"train/tensor_act_model_layers_80_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_64_mlp_up_proj_weight/range":0.2587890625,"train/tensor_act_model_layers_30_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_46_mlp_up_proj/range":8.5625,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_79_residual_pre_attn/min":314,"train/tensor_act_model_layers_79_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_68_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_2_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_50_mlp_down_proj/max_abs":16.25,"train/tensor_act_model_layers_37_residual_post_attn/mean":148,"train/tensor_act_model_layers_9_residual_post_mlp/norm":81849.6953125,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/norm":6.5,"train/tensor_act_model_layers_2_residual_pre_attn/max":8.375,"train/tensor_act_model_layers_91_mlp_gate_proj/std":1.6328125,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/range":2316,"train/tensor_act_model_layers_39_mlp_gate_proj/min":-1.3125,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_63_mlp_down_proj/std":0.40625,"train/tensor_act_model_layers_3_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_0_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_90_mlp_up_proj/mean":-0.08056640625,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_28_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_46_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_act_model_layers_86_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_17_residual_post_attn/mean":68,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/norm":0,"train/grad_norm":86528,"train/tensor_param_model_layers_61_mlp_down_proj_weight/norm":6.71875,"train/tensor_act_model_layers_64_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_3_residual_pre_attn/min":11.5,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_11_mlp_down_proj_weight/mean":-0.00159454345703125,"train/tensor_act_model_layers_97_residual_pre_attn/max":388,"train/tensor_act_model_layers_88_mlp_gate_proj/std":1.640625,"train/tensor_param_model_layers_26_mlp_down_proj_weight/max_abs":0.1181640625,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/norm":8.125,"train/tensor_act_model_layers_66_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_33_mlp/mean":4,"train/tensor_act_model_layers_12_mlp_gate_proj/max_abs":4.78125,"train/tensor_act_model_layers_0_residual_post_mlp/norm":8190.14501953125,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_92_mlp/max":4,"train/tensor_act_model_layers_15_residual_pre_attn/max":61,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/std":0.034423828125,"train/tensor_param_model_layers_97_mlp_down_proj_weight/norm":9.375,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/std":0.039306640625,"train/tensor_act_model_layers_95_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_post_mlp/min":226,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_20_mlp/min":4,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_78_mlp_up_proj_weight/max":0.1318359375,"train/tensor_param_model_layers_41_mlp_up_proj_weight/max_abs":0.12890625,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/range":0.22509765625,"train/tensor_act_model_layers_52_residual_post_mlp/max_abs":213,"train/tensor_act_model_layers_29_mlp/max":4,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_47_mlp/norm":8192,"train/tensor_param_model_layers_26_mlp_up_proj_weight/range":0.24267578125,"train/tensor_act_model_layers_56_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_mlp/std":0,"train/tensor_act_model_layers_76_mlp_gate_proj/std":1.625,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_10_mlp_down_proj/std":0.419921875,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_72_mlp/mean":4,"train/tensor_act_model_layers_6_mlp_up_proj/max":3.515625,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_38_mlp_up_proj_weight/max":0.1240234375,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_38_mlp_up_proj_weight/range":0.2548828125,"train/tensor_act_model_layers_75_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_13_mlp_up_proj_weight/range":0.22900390625,"train/tensor_act_model_layers_47_residual_post_mlp/min":190,"train/tensor_act_model_layers_48_mlp_gate_proj/std":1.640625,"train/tensor_param_model_layers_67_mlp_up_proj_weight/norm":6.9375,"train/tensor_act_model_layers_75_residual_pre_attn/min":298,"train/tensor_act_model_layers_95_residual_pre_attn/norm":778205.25,"train/tensor_act_model_layers_88_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/max_abs":0.240234375,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_64_mlp_gate_proj/max_abs":4.375,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_3_mlp_gate_proj/mean":0.482421875,"train/tensor_act_model_layers_51_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_37_mlp_gate_proj/norm":5114.3203125,"train/tensor_act_model_layers_16_residual_pre_attn/norm":130804.1796875,"train/tensor_act_model_layers_11_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_77_mlp_gate_proj/range":6.0625,"train/tensor_act_model_layers_18_mlp_up_proj/mean":-0.2734375,"train/tensor_act_model_layers_44_mlp/std":0,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_4_mlp_down_proj_weight/mean":0.00188446044921875,"train/tensor_act_model_layers_14_residual_post_attn/max":57,"train/tensor_act_model_layers_77_residual_post_attn/max":308,"train/tensor_act_model_layers_95_residual_post_attn/range":2,"train/tensor_act_model_layers_50_mlp_up_proj/norm":5427.044921875,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_83_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_mlp_gate_proj/max_abs":4.4375,"train/tensor_act_model_layers_28_residual_post_mlp/min":114.5,"train/tensor_act_model_layers_47_mlp_gate_proj/range":5.890625,"train/tensor_act_model_layers_46_mlp_gate_proj/max":4.59375,"train/tensor_param_model_layers_50_mlp_up_proj_weight/max":0.1259765625,"train/tensor_param_model_layers_91_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_residual_post_attn/mean":56,"train/tensor_act_model_layers_36_residual_pre_attn/norm":294764.78125,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_25_residual_post_attn/max_abs":101,"train/tensor_act_model_layers_81_residual_post_mlp/min":326,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/max_abs":0.12255859375,"train/tensor_act_model_layers_15_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_16_mlp_up_proj/max_abs":5.0625,"train/tensor_act_model_layers_10_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_4_mlp_up_proj_weight/range":0.21435546875,"train/tensor_act_model_layers_15_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_mlp_up_proj/max":4.71875,"train/tensor_act_model_layers_82_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_22_mlp_up_proj/min":-4.0625,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/norm":7.375,"train/tensor_act_model_layers_19_residual_pre_attn/norm":155434.3125,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_29_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_54_mlp/min":4,"train/tensor_act_model_layers_89_residual_post_mlp/max":360,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_91_mlp_up_proj/min":-4.5,"train/tensor_act_model_layers_21_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/mean":0.005096435546875,"train/tensor_param_model_layers_16_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_80_mlp_up_proj_weight/max":0.1513671875,"train/tensor_act_model_layers_4_mlp_up_proj/min":-3.796875,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_13_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_residual_post_mlp/mean":208,"train/tensor_act_model_layers_25_mlp/min":4,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_1_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_40_residual_post_attn/max_abs":161,"train/tensor_act_model_layers_81_residual_post_mlp/mean":328,"train/tensor_act_model_layers_41_mlp_gate_proj/std":1.6640625,"train/tensor_act_model_layers_19_residual_post_mlp/min":78.5,"train/tensor_act_model_layers_8_residual_post_attn/range":1.625,"train/tensor_act_model_layers_74_mlp_up_proj/max_abs":4.71875,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_86_mlp_up_proj_weight/mean":0.000766754150390625,"train/tensor_act_model_layers_90_mlp/range":0,"train/tensor_act_model_layers_68_mlp/norm":8192,"train/tensor_act_model_layers_63_residual_post_mlp/norm":524001.8125,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/std":0.0400390625,"train/tensor_act_model_layers_59_mlp/max_abs":4,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/range":0.255859375,"train/tensor_act_model_layers_51_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_29_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_31_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_98_mlp_down_proj_weight/range":0.3447265625,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/norm":6.03125,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_68_mlp_gate_proj/max":4.625,"train/tensor_act_model_layers_69_residual_pre_attn/max":276,"train/tensor_act_model_layers_63_residual_post_attn/min":250,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_39_residual_pre_attn/max_abs":157,"train/tensor_act_model_layers_26_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_8_mlp_down_proj_weight/norm":5.28125,"train/tensor_act_model_layers_39_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_7_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_25_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_61_mlp_down_proj/range":2.625,"train/tensor_act_model_layers_38_mlp_up_proj/mean":0.0732421875,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_98_mlp_down_proj/max_abs":20,"train/tensor_act_model_layers_53_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_31_mlp_gate_proj/mean":0.435546875,"train/tensor_act_model_layers_41_mlp_gate_proj/max_abs":4.53125,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_20_mlp_down_proj/std":0.41015625,"train/tensor_act_model_layers_80_mlp_gate_proj/max":4.5,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/std":0.0546875,"train/tensor_act_model_layers_99_residual_pre_attn/norm":810973.25,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_95_mlp_gate_proj/norm":5095.26611328125,"train/tensor_act_model_layers_77_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_76_mlp/range":0,"train/tensor_act_model_layers_65_mlp/std":0,"train/tensor_param_model_layers_5_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_14_residual_post_attn/std":0.3125,"train/tensor_param_model_layers_60_mlp_down_proj_weight/max_abs":0.1318359375,"train/tensor_act_model_layers_28_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_14_mlp_up_proj_weight/max":0.125,"train/tensor_param_model_layers_77_mlp_up_proj_weight/min":-0.15234375,"train/tensor_param_model_layers_33_mlp_up_proj_weight/max":0.125,"train/tensor_act_model_layers_40_mlp/max":4,"train/tensor_param_model_layers_42_mlp_up_proj_weight/std":0.0341796875,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/min":-0.125,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_18_mlp_up_proj_weight/max":0.10986328125,"train/tensor_act_model_layers_43_residual_post_mlp/max_abs":177,"train/tensor_param_model_layers_72_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_mlp_up_proj/min":-4.4375,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_mlp_down_proj/min":13.9375,"train/tensor_act_model_layers_42_mlp/max":4,"train/tensor_act_model_layers_21_residual_pre_attn/std":0.359375,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_31_mlp_up_proj/std":1.8671875,"train/tensor_act_model_layers_21_mlp_gate_proj/min":-1.2890625,"train/tensor_act_model_layers_47_residual_pre_attn/max_abs":189,"train/tensor_act_model_layers_42_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_84_residual_post_attn/range":2,"train/tensor_act_model_layers_69_mlp/norm":8192,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_96_mlp_down_proj/min":14,"train/tensor_param_model_layers_67_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_68_mlp/max":4,"train/tensor_act_model_layers_87_residual_post_attn/range":2,"train/tensor_act_model_layers_73_mlp_down_proj/range":3.3125,"train/tensor_act_model_layers_33_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/range":0.2607421875,"train/tensor_param_model_layers_8_mlp_down_proj_weight/min":-0.111328125,"train/tensor_act_model_layers_46_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_72_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_89_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_89_mlp_down_proj_weight/std":0.04638671875,"train/tensor_act_model_layers_67_residual_pre_attn/mean":268,"train/tensor_act_model_layers_12_mlp/range":0,"train/tensor_act_model_layers_78_mlp_gate_proj/mean":0.4296875,"train/tensor_param_model_layers_66_mlp_down_proj_weight/min":-0.12890625,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/mean":0.0037841796875,"train/tensor_act_model_layers_29_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_89_mlp_gate_proj/range":5.8828125,"train/tensor_act_model_layers_51_residual_post_attn/norm":417644.53125,"train/tensor_act_model_layers_11_residual_post_attn/max_abs":45,"train/tensor_act_model_layers_23_mlp/range":0,"train/tensor_act_model_layers_21_residual_pre_attn/min":82.5,"train/tensor_act_model_layers_38_mlp/range":0,"train/tensor_act_model_layers_48_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/norm":11.5,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_77_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_86_mlp_up_proj/std":1.8125,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_80_mlp_up_proj/max":4.375,"train/tensor_act_model_layers_52_mlp/std":0,"train/tensor_param_model_layers_6_mlp_up_proj_weight/max":0.1064453125,"train/tensor_act_model_layers_83_residual_pre_attn/norm":679901.25,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/max_abs":0.1416015625,"train/tensor_act_model_layers_30_mlp/mean":4,"train/tensor_param_model_layers_24_mlp_up_proj_weight/mean":0.000457763671875,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_80_mlp_down_proj/range":3.125,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/range":0.259765625,"train/tensor_param_model_layers_80_mlp_down_proj_weight/range":0.28125,"train/tensor_param_model_layers_32_mlp_up_proj_weight/max":0.123046875,"train/tensor_param_model_layers_75_mlp_up_proj_weight/max":0.1318359375,"train/tensor_act_model_layers_66_residual_pre_attn/min":262,"train/tensor_act_model_layers_57_residual_post_attn/max_abs":229,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_27_mlp_down_proj_weight/range":0.2509765625,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_34_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_71_mlp_down_proj_weight/norm":7,"train/tensor_act_model_layers_66_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_36_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_92_mlp_up_proj/range":10,"train/tensor_param_model_layers_12_mlp_down_proj_weight/mean":6.580352783203125e-05,"train/tensor_act_model_layers_23_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_residual_post_mlp/max_abs":157,"train/tensor_param_model_layers_24_mlp_up_proj_weight/std":0.03173828125,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/std":0.032958984375,"train/tensor_act_model_layers_8_mlp_up_proj/min":-4.5,"train/tensor_act_model_layers_2_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_33_residual_post_mlp/max":137,"train/tensor_act_model_layers_7_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_23_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_40_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_49_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_87_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_41_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp_up_proj/max":3.9375,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_62_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_99_mlp/min":4,"train/tensor_act_model_layers_18_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/range":0.2421875,"train/tensor_act_model_layers_64_mlp_gate_proj/min":-1.25,"train/tensor_act_model_layers_14_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_18_residual_post_mlp/std":0.3515625,"train/tensor_act_model_layers_7_residual_post_attn/norm":57287.57421875,"train/tensor_act_model_layers_16_mlp/max_abs":4,"train/tensor_param_model_layers_37_mlp_down_proj_weight/norm":6.03125,"train/tensor_act_model_layers_89_mlp/std":0,"train/tensor_act_model_layers_26_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/range":0.2509765625,"train/tensor_param_model_layers_4_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_82_mlp_up_proj/std":1.7734375,"train/tensor_act_model_layers_99_residual_post_attn/mean":396,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/mean":0.003814697265625,"train/tensor_act_model_layers_50_residual_post_mlp/max_abs":205,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_62_mlp/norm":8192,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_36_residual_post_mlp/max":149,"train/tensor_act_model_layers_70_residual_post_attn/max":280,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_85_mlp_down_proj/norm":31068.140625,"train/tensor_act_model_layers_1_residual_post_mlp/norm":16366.814453125,"train/tensor_param_model_layers_15_mlp_down_proj_weight/max_abs":0.1162109375,"train/tensor_param_model_layers_39_mlp_up_proj_weight/max":0.1259765625,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/max_abs":0.138671875,"train/tensor_act_model_layers_8_residual_pre_attn/max":32.75,"train/tensor_act_model_layers_97_mlp_gate_proj/max":5.09375,"train/tensor_param_model_layers_38_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_32_mlp_up_proj/std":1.90625,"train/tensor_act_model_layers_32_residual_pre_attn/mean":128,"train/tensor_param_model_layers_39_mlp_down_proj_weight/min":-0.1279296875,"train/tensor_param_model_layers_64_mlp_up_proj_weight/mean":-0.00135040283203125,"train/tensor_param_model_layers_11_mlp_up_proj_weight/mean":-0.00042724609375,"train/tensor_act_model_layers_10_mlp_down_proj/max_abs":16,"train/tensor_act_model_layers_22_mlp/max_abs":4,"train/tensor_act_model_layers_54_residual_post_attn/max_abs":217,"train/tensor_act_model_layers_9_residual_post_attn/max_abs":36.75,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_92_mlp_down_proj/max_abs":17.875,"train/tensor_act_model_layers_23_mlp_down_proj/range":1.875,"train/tensor_act_model_layers_36_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_65_residual_post_mlp/max_abs":264,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_35_mlp_up_proj/range":8.921875,"train/tensor_param_model_layers_3_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_3_residual_post_mlp/range":1.25,"train/tensor_param_model_layers_67_mlp_down_proj_weight/mean":-0.00201416015625,"train/tensor_act_model_layers_32_mlp_up_proj/range":8.5,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_13_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_75_residual_pre_attn/norm":614365.25,"train/tensor_act_model_layers_0_mlp_gate_proj/mean":0.08203125,"train/tensor_act_model_layers_33_residual_pre_attn/norm":270188.8125,"train/tensor_act_model_layers_44_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_31_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_63_mlp_down_proj_weight/max_abs":0.142578125,"train/tensor_act_model_layers_45_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_mlp_down_proj/range":1.5625,"train/tensor_act_model_layers_42_mlp_down_proj/range":2,"train/tensor_act_model_layers_22_residual_pre_attn/norm":179985.390625,"train/tensor_act_model_layers_56_mlp_down_proj/max":16.375,"train/tensor_act_model_layers_47_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_62_mlp_down_proj_weight/range":0.251953125,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_69_mlp/max":4,"train/tensor_act_model_layers_89_residual_post_attn/norm":729053.25,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/mean":0.0050048828125,"train/tensor_act_model_layers_6_mlp/std":0,"train/tensor_act_model_layers_44_residual_pre_attn/norm":360300.625,"train/tensor_act_model_layers_46_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_86_mlp_gate_proj/max_abs":4.6875,"train/tensor_act_model_layers_20_mlp_gate_proj/range":6.2421875,"train/tensor_act_model_layers_9_mlp_up_proj/norm":5442.03271484375,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_54_mlp_gate_proj/min":-1.359375,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_54_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/norm":5.8125,"train/tensor_act_model_layers_32_mlp_gate_proj/max":4.59375,"train/tensor_act_model_layers_65_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_35_mlp_gate_proj/max":4.46875,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_residual_pre_attn/std":0.1591796875,"train/tensor_act_model_layers_5_mlp_up_proj/max_abs":4.25,"train/tensor_act_model_layers_87_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_residual_post_attn/mean":168,"train/tensor_act_model_layers_12_mlp_down_proj/max_abs":16,"train/tensor_param_model_layers_4_mlp_down_proj_weight/min":-0.11376953125,"train/tensor_act_model_layers_67_residual_post_attn/max":268,"train/tensor_act_model_layers_39_residual_pre_attn/max":157,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_8_residual_post_attn/min":31.125,"train/tensor_act_model_layers_38_residual_post_mlp/min":154,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_mlp_up_proj/range":9.96875,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_61_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_94_residual_pre_attn/range":2,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_11_mlp_gate_proj/mean":0.451171875,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_34_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_78_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_29_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_65_mlp_up_proj/mean":0.0281982421875,"train/tensor_act_model_layers_25_mlp_down_proj/norm":29996.939453125,"train/tensor_param_model_layers_39_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_30_mlp/norm":8192,"train/tensor_act_model_layers_2_residual_post_attn/max_abs":8.4375,"train/tensor_param_model_layers_73_mlp_down_proj_weight/mean":-0.00023365020751953125,"train/tensor_act_model_layers_86_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_76_mlp_up_proj_weight/max_abs":0.1318359375,"train/tensor_param_model_layers_85_mlp_down_proj_weight/max_abs":0.1474609375,"train/tensor_param_model_layers_26_mlp_up_proj_weight/max":0.1298828125,"train/tensor_act_model_layers_34_mlp_up_proj/std":1.890625,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_87_mlp_up_proj/max":4.75,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_18_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/mean":4,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/norm":6.875,"train/tensor_act_model_layers_32_residual_post_attn/std":0.34375,"train/tensor_act_model_layers_56_mlp_up_proj/max_abs":4.34375,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_83_mlp_up_proj/min":-4.21875,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/max_abs":0.12255859375,"train/tensor_act_model_layers_18_mlp_up_proj/max_abs":4.875,"train/tensor_act_model_layers_52_mlp_down_proj/range":2.375,"train/tensor_act_model_layers_65_mlp/range":0,"train/tensor_act_model_layers_52_residual_post_attn/min":206,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_post_attn/range":2,"train/tensor_act_model_layers_92_mlp_gate_proj/max":5.3125,"train/tensor_act_model_layers_12_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_36_mlp/range":0,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_44_residual_pre_attn/max":177,"train/tensor_param_model_layers_81_mlp_up_proj_weight/norm":7.9375,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_21_mlp_up_proj_weight/range":0.2353515625,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_37_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_19_mlp_down_proj_weight/std":0.0306396484375,"train/tensor_act_model_layers_19_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_15_mlp_up_proj/max":4.15625,"train/tensor_act_model_layers_82_residual_post_attn/norm":671709.25,"train/tensor_act_model_layers_41_residual_post_mlp/max_abs":169,"train/tensor_act_model_layers_57_residual_post_mlp/range":3,"train/tensor_param_model_layers_16_mlp_up_proj_weight/max_abs":0.111328125,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_74_residual_post_mlp/max":300,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_52_residual_post_mlp/norm":434028.53125,"train/tensor_act_model_layers_22_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_residual_pre_attn/norm":212753.15625,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_47_mlp_up_proj/max":3.859375,"train/tensor_act_model_layers_97_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_48_residual_post_attn/range":3,"train/tensor_act_model_layers_46_residual_post_mlp/norm":384876.59375,"train/tensor_act_model_layers_82_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/norm":7.03125,"train/tensor_act_model_layers_70_residual_pre_attn/mean":280,"train/tensor_act_model_layers_57_residual_pre_attn/min":226,"train/tensor_param_model_layers_70_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_mlp_up_proj/std":1.828125,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/max_abs":0.1689453125,"train/tensor_act_model_layers_34_mlp_gate_proj/max_abs":4.71875,"train/tensor_param_model_layers_31_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_residual_pre_attn/norm":450348.8125,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_32_residual_post_attn/max":129,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_40_mlp_up_proj/norm":5433.17529296875,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/range":0.25390625,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/mean":0.004364013671875,"train/tensor_param_model_layers_52_mlp_down_proj_weight/range":0.2607421875,"train/tensor_act_model_layers_44_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_90_mlp_up_proj_weight/range":0.3076171875,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/norm":9.4375,"train/tensor_act_model_layers_61_mlp_gate_proj/std":1.6015625,"train/tensor_act_model_layers_17_mlp_gate_proj/range":5.90625,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_67_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_63_mlp_up_proj_weight/std":0.03759765625,"train/tensor_act_model_layers_92_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_90_mlp_down_proj_weight/range":0.2978515625,"train/tensor_param_model_layers_12_mlp_up_proj_weight/std":0.0303955078125,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_residual_post_attn/norm":515868.75,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_41_mlp_up_proj_weight/norm":6.09375,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_57_residual_pre_attn/std":0.408203125,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_23_mlp_down_proj/max":16,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_87_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_residual_post_attn/min":182,"train/tensor_param_model_layers_28_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_11_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_80_mlp/range":0,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/norm":6.03125,"train/tensor_act_model_layers_41_residual_post_mlp/max":169,"train/tensor_param_model_layers_87_mlp_up_proj_weight/range":0.3115234375,"train/tensor_act_model_layers_80_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_84_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_54_mlp/max":4,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_39_mlp_down_proj_weight/max":0.1259765625,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_80_mlp_up_proj/range":8.78125,"train/tensor_act_model_layers_25_mlp/std":0,"train/tensor_act_model_layers_69_mlp_gate_proj/max":4.0625,"train/tensor_act_model_layers_29_mlp_down_proj/norm":30109.40625,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_79_mlp_gate_proj/range":5.8984375,"train/tensor_act_model_layers_1_residual_pre_attn/max":4.3125,"train/tensor_act_model_layers_7_residual_post_attn/range":1.625,"train/tensor_param_model_layers_2_mlp_down_proj_weight/range":0.21435546875,"train/tensor_act_model_layers_64_residual_pre_attn/max":256,"train/tensor_act_model_layers_83_mlp/std":0,"train/tensor_act_model_layers_93_residual_post_attn/mean":372,"train/tensor_act_model_layers_22_mlp_up_proj/max":4.125,"train/tensor_act_model_layers_80_residual_post_mlp/max":324,"train/tensor_act_model_layers_77_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_mlp_down_proj/mean":15.25,"train/tensor_act_model_layers_45_mlp_gate_proj/mean":0.546875,"train/tensor_param_model_layers_92_mlp_down_proj_weight/norm":8.75,"train/tensor_act_model_layers_66_mlp/std":0,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_10_mlp/norm":8192,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/max_abs":0.1845703125,"train/tensor_act_model_layers_31_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_65_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_83_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_62_mlp_down_proj_weight/mean":3.790855407714844e-05,"train/tensor_act_model_layers_27_residual_post_attn/max":109,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/max":0.123046875,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_19_mlp_up_proj/range":9.09375,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_83_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_83_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_18_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_2_mlp_gate_proj/max":4.09375,"train/tensor_act_model_layers_92_residual_post_mlp/max":372,"train/tensor_act_model_layers_0_residual_post_mlp/range":0.734375,"train/tensor_param_model_layers_91_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_28_mlp/mean":4,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/norm":5.84375,"train/tensor_act_model_layers_81_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_29_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_57_residual_post_mlp/min":230,"train/tensor_act_model_layers_60_residual_pre_attn/mean":240,"train/tensor_param_model_layers_41_mlp_up_proj_weight/mean":-0.0001773834228515625,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/std":0.0498046875,"train/tensor_act_model_layers_93_mlp/range":0,"train/tensor_act_model_layers_88_mlp/std":0,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_45_residual_post_attn/max":181,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/range":0.291015625,"train/tensor_act_model_layers_17_mlp/min":4,"train/tensor_param_model_layers_4_mlp_up_proj_weight/norm":5.40625,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_73_mlp_gate_proj/max":4.40625,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_45_mlp_down_proj_weight/mean":-0.0004596710205078125,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_58_mlp/min":4,"train/tensor_act_model_layers_32_mlp_gate_proj/mean":0.640625,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_51_mlp_gate_proj/max_abs":5.0625,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/norm":8.75,"train/tensor_act_model_layers_82_mlp_gate_proj/min":-1.296875,"train/tensor_act_model_layers_11_mlp_gate_proj/min":-1.2890625,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_72_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_62_residual_post_mlp/norm":515868.75,"train/tensor_act_model_layers_35_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_41_mlp_down_proj_weight/min":-0.11962890625,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_33_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_91_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/norm":5.6875,"train/tensor_act_model_layers_35_mlp_gate_proj/mean":0.56640625,"train/tensor_act_model_layers_82_mlp_down_proj/max_abs":18.25,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_98_mlp_down_proj_weight/max":0.1962890625,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_75_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_0_residual_pre_attn/frac_near_user_limit":0,"train/tensor_param_model_layers_97_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_residual_pre_attn/max_abs":109,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_91_residual_post_attn/norm":745437.25,"train/tensor_act_model_layers_55_mlp_down_proj/max":16.375,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_63_mlp_gate_proj/min":-1.2734375,"train/tensor_act_model_layers_31_mlp_down_proj/min":14.1875,"train/tensor_param_model_layers_1_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_23_mlp/min":4,"train/tensor_act_model_layers_39_mlp_down_proj/std":0.40234375,"train/tensor_act_model_layers_25_mlp/norm":8192,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_88_mlp_down_proj/mean":15.0625,"train/tensor_param_model_layers_55_mlp_down_proj_weight/mean":0.0004634857177734375,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_83_mlp_gate_proj/std":1.640625,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_62_residual_post_attn/min":246,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_5_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_47_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_91_mlp_up_proj/max":5.125,"train/tensor_act_model_layers_57_mlp_up_proj/norm":5197.75244140625,"train/tensor_act_model_layers_45_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/mean":0.0023956298828125,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_45_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_91_residual_post_attn/range":2,"train/tensor_act_model_layers_63_mlp_gate_proj/norm":5057.82275390625,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_38_residual_pre_attn/min":150,"train/tensor_act_model_layers_96_residual_pre_attn/norm":786397.25,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_98_residual_post_attn/min":390,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/range":0.24560546875,"train/tensor_param_model_layers_14_mlp_down_proj_weight/std":0.0299072265625,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/min":-0.1259765625,"train/tensor_param_model_layers_36_mlp_down_proj_weight/min":-0.1181640625,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_98_residual_post_mlp/norm":810973.25,"train/tensor_act_model_layers_38_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_48_mlp_gate_proj/norm":5098.4501953125,"train/tensor_act_model_layers_11_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_28_residual_pre_attn/mean":112,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/norm":62208,"train/tensor_param_model_layers_80_mlp_down_proj_weight/norm":7.625,"train/tensor_act_model_layers_27_residual_post_attn/std":0.359375,"train/tensor_act_model_layers_14_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_44_mlp_up_proj/range":9.46875,"train/tensor_act_model_layers_36_residual_post_attn/mean":144,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_65_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_57_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_34_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_77_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_83_residual_post_mlp/range":2,"train/tensor_act_model_layers_38_mlp_down_proj/range":1.8125,"train/tensor_act_model_layers_31_mlp/max":4,"train/tensor_act_model_layers_81_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_29_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_82_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_27_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_mlp/norm":8192,"train/tensor_act_model_layers_30_residual_pre_attn/norm":245497.171875,"train/tensor_act_model_layers_9_residual_post_mlp/min":39,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_57_mlp/mean":4,"train/tensor_act_model_layers_90_residual_pre_attn/frac_near_dtype_limit":0,"eval/samples_per_second":451.977,"train/tensor_act_model_layers_10_mlp_gate_proj/min":-1.359375,"train/tensor_param_model_layers_51_mlp_down_proj_weight/std":0.03564453125,"train/tensor_act_model_layers_14_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_44_mlp_up_proj/std":1.8515625,"train/tensor_act_model_layers_44_mlp_gate_proj/max":5.03125,"train/tensor_param_model_layers_60_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_23_mlp_down_proj_weight/mean":0.00130462646484375,"train/tensor_act_model_layers_99_mlp/mean":4,"train/tensor_act_model_layers_36_residual_post_mlp/max_abs":149,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/mean":0.004791259765625,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_69_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_69_mlp_up_proj_weight/min":-0.126953125,"train/tensor_act_model_layers_81_mlp_up_proj/max":4.34375,"train/tensor_act_model_layers_45_residual_pre_attn/mean":180,"train/tensor_act_model_layers_59_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_28_residual_post_attn/std":0.388671875,"train/tensor_act_model_layers_14_residual_post_attn/range":2,"train/tensor_act_model_layers_65_mlp_down_proj/range":2.625,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_40_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_27_residual_pre_attn/mean":108,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/min":-0.1337890625,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_87_mlp_up_proj_weight/max_abs":0.1591796875,"train/tensor_act_model_layers_58_mlp_up_proj/max":5.1875,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/std":0.03076171875,"train/tensor_act_model_layers_53_mlp_gate_proj/std":1.6875,"train/tensor_act_model_layers_19_mlp_down_proj/norm":30095.490234375,"train/tensor_param_model_layers_34_mlp_up_proj_weight/min":-0.1142578125,"train/tensor_act_model_layers_65_mlp_gate_proj/norm":4798.4853515625,"train/tensor_act_model_layers_97_residual_post_attn/max_abs":388,"train/tensor_act_model_layers_54_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_10_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_6_mlp_gate_proj/mean":0.54296875,"train/tensor_act_model_layers_16_mlp_up_proj/norm":5572.62158203125,"train/tensor_act_model_layers_17_mlp_gate_proj/std":1.6796875,"train/tensor_act_model_layers_60_residual_post_attn/max":241,"train/tensor_act_model_layers_48_residual_pre_attn/norm":393068.59375,"train/tensor_param_model_layers_26_mlp_down_proj_weight/norm":5.71875,"train/tensor_act_model_layers_24_mlp_gate_proj/range":5.7734375,"train/tensor_act_model_layers_40_mlp/mean":4,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_70_mlp_up_proj_weight/range":0.2705078125,"train/tensor_param_model_layers_46_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_31_mlp_down_proj_weight/mean":0.0002155303955078125,"train/tensor_act_model_layers_95_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_21_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_99_mlp_down_proj_weight/std":0.05224609375,"train/tensor_param_model_layers_11_mlp_up_proj_weight/norm":5.4375,"train/tensor_act_model_layers_63_mlp_up_proj/max_abs":4.9375,"train/tensor_param_model_layers_37_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_81_mlp_down_proj/max":17.5,"train/tensor_param_model_layers_64_mlp_up_proj_weight/norm":6.875,"train/tensor_param_model_layers_43_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp_gate_proj/std":1.6640625,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/min":-0.1640625,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_77_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_47_residual_pre_attn/norm":384876.59375,"train/tensor_act_model_layers_71_mlp/mean":4,"train/tensor_act_model_layers_15_residual_post_mlp/mean":63.75,"_step":3464,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_73_mlp_up_proj_weight/range":0.267578125,"train/tensor_act_model_layers_1_mlp/range":0.015625,"train/tensor_act_model_layers_0_residual_pre_attn/max_abs":0.30078125,"train/tensor_act_model_layers_22_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_49_mlp_gate_proj/range":5.7109375,"train/tensor_act_model_layers_52_mlp_up_proj/min":-4.6875,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_75_residual_post_mlp/max_abs":304,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_73_mlp_up_proj/range":9.03125,"train/tensor_act_model_layers_10_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_81_residual_post_attn/mean":324,"train/tensor_act_model_layers_8_residual_post_attn/std":0.2080078125,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_28_residual_post_attn/max":113,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_46_residual_pre_attn/range":3,"train/tensor_act_model_layers_93_mlp_down_proj/max":17.625,"train/tensor_act_model_layers_59_mlp_gate_proj/max_abs":4.78125,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_60_residual_post_attn/std":0.416015625,"train/tensor_act_model_layers_98_mlp_up_proj/norm":5224.447265625,"train/tensor_act_model_layers_9_mlp_down_proj/max_abs":16.25,"train/tensor_act_model_layers_39_residual_post_attn/mean":156,"train/tensor_act_model_layers_92_mlp/min":4,"train/tensor_param_model_layers_76_mlp_down_proj_weight/norm":7.25,"train/tensor_param_model_layers_5_mlp_down_proj_weight/min":-0.123046875,"train/tensor_act_model_layers_80_residual_post_mlp/norm":663517.3125,"train/tensor_act_model_layers_63_residual_post_mlp/std":0.37109375,"train/tensor_param_model_layers_58_mlp_down_proj_weight/range":0.259765625,"train/tensor_act_model_layers_74_mlp_down_proj/mean":15,"train/tensor_act_model_layers_68_mlp_down_proj/mean":15.0625,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_40_residual_post_mlp/max_abs":165,"train/tensor_act_model_layers_21_residual_post_mlp/min":86.5,"train/tensor_act_model_layers_80_residual_post_attn/min":318,"train/tensor_act_model_layers_29_mlp_up_proj/max_abs":5.0625,"train/tensor_act_model_layers_8_mlp_down_proj/max":15.875,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_97_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_58_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_residual_post_attn/mean":108,"train/tensor_param_model_layers_83_mlp_down_proj_weight/std":0.04443359375,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/min":-0.1357421875,"train/tensor_param_model_layers_83_mlp_down_proj_weight/max_abs":0.1484375,"train/tensor_act_model_layers_19_mlp_gate_proj/max":4.53125,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_87_residual_post_mlp/range":2,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_75_mlp_down_proj_weight/norm":7.15625,"train/tensor_act_model_layers_62_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_6_mlp_gate_proj/max":4.71875,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_24_mlp_down_proj_weight/mean":-0.000904083251953125,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_91_mlp_gate_proj/max_abs":5.0625,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/max":0.166015625,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_24_mlp_up_proj_weight/norm":5.75,"train/tensor_param_model_layers_32_mlp_up_proj_weight/max_abs":0.123046875,"train/tensor_act_model_layers_28_mlp_gate_proj/max_abs":4.46875,"train/tensor_act_model_layers_81_mlp_down_proj/max_abs":17.5,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_residual_pre_attn/range":2,"train/tensor_act_model_layers_66_residual_post_mlp/min":266,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_95_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_53_residual_pre_attn/max_abs":213,"train/tensor_param_model_layers_15_mlp_up_proj_weight/max_abs":0.1259765625,"train/tensor_param_model_layers_46_mlp_up_proj_weight/min":-0.1240234375,"train/tensor_act_model_layers_72_residual_post_mlp/max":292,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/max":0.13671875,"train/tensor_act_model_layers_69_mlp_up_proj/range":8.046875,"train/tensor_act_model_layers_91_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_93_mlp_up_proj_weight/max_abs":0.173828125,"train/tensor_param_model_layers_42_mlp_down_proj_weight/max":0.126953125,"train/tensor_act_model_layers_59_mlp_gate_proj/range":6.0625,"train/tensor_act_model_layers_4_mlp_gate_proj/std":1.6484375,"train/tensor_act_model_layers_45_residual_pre_attn/range":3,"train/tensor_param_model_layers_20_mlp_down_proj_weight/std":0.0311279296875,"train/tensor_act_model_layers_39_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp/range":0,"train/tensor_act_model_layers_94_mlp_down_proj/norm":31011.5234375,"train/tensor_act_model_layers_37_residual_pre_attn/max_abs":149,"train/tensor_act_model_layers_24_residual_pre_attn/range":2.5,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/norm":9.25,"train/tensor_act_model_layers_20_mlp_gate_proj/min":-1.2734375,"train/tensor_act_model_layers_58_residual_pre_attn/max":233,"train/tensor_act_model_layers_95_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_84_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_23_mlp_up_proj_weight/max":0.12353515625,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_10_residual_post_mlp/max":44.75,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_36_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_53_mlp_down_proj/max_abs":16.375,"train/tensor_act_model_layers_21_mlp_down_proj/std":0.4609375,"train/tensor_act_model_layers_2_mlp_down_proj/std":0.59765625,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_70_mlp_gate_proj/max":4.3125,"train/tensor_act_model_layers_44_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_65_residual_post_mlp/range":2,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_86_mlp_up_proj_weight/norm":8.4375,"train/tensor_act_model_layers_97_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_98_mlp_down_proj_weight/min":-0.1484375,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_19_residual_pre_attn/min":74.5,"train/tensor_act_model_layers_8_mlp_up_proj/range":9.21875,"train/tensor_act_model_layers_16_mlp_up_proj/std":1.921875,"train/tensor_act_model_layers_98_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_70_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_post_mlp/norm":466732.78125,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_49_residual_post_mlp/mean":200,"train/tensor_act_model_layers_1_mlp_down_proj/norm":33034.4140625,"train/tensor_param_model_layers_42_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_5_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_59_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_56_residual_post_attn/min":222,"train/tensor_act_model_layers_71_mlp_down_proj/max_abs":17,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_97_residual_post_mlp/max_abs":392,"train/tensor_act_model_layers_99_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_43_mlp_down_proj_weight/range":0.25048828125,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_11_mlp_up_proj/norm":5551.71044921875,"train/tensor_act_model_layers_84_mlp/std":0,"train/tensor_act_model_layers_33_residual_post_attn/norm":270188.8125,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_41_mlp_down_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_6_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_12_residual_pre_attn/mean":48,"train/tensor_act_model_layers_87_mlp_up_proj/min":-4.8125,"train/tensor_act_model_layers_88_mlp_down_proj/norm":30826.08984375,"train/tensor_act_model_layers_88_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_98_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_41_residual_post_attn/std":0.31640625,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_39_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_78_residual_post_mlp/max_abs":316,"train/tensor_act_model_layers_19_residual_pre_attn/std":0.3515625,"train/tensor_act_model_layers_28_mlp_down_proj/max":16.25,"train/tensor_act_model_layers_17_residual_post_attn/max":69,"train/tensor_act_model_layers_80_residual_post_attn/mean":320,"train/tensor_param_model_layers_27_mlp_up_proj_weight/std":0.03173828125,"train/tensor_act_model_layers_35_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_31_residual_post_mlp/norm":261786.84375,"train/tensor_act_model_layers_80_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_13_mlp_down_proj/min":14.0625,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_3_mlp_down_proj/std":0.46875,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_65_mlp_up_proj_weight/std":0.0380859375,"train/tensor_act_model_layers_12_mlp_up_proj/range":9.78125,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_42_mlp_up_proj_weight/max_abs":0.130859375,"train/tensor_act_model_layers_0_mlp_up_proj/max_abs":7.90625,"train/tensor_act_model_layers_82_mlp_gate_proj/range":6.515625,"train/tensor_act_model_layers_79_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_82_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_52_mlp_up_proj_weight/max":0.1337890625,"train/tensor_act_model_layers_7_mlp_up_proj/max_abs":4.34375,"train/tensor_param_model_layers_97_mlp_down_proj_weight/std":0.0517578125,"train/tensor_act_model_layers_35_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_39_mlp/range":0,"train/tensor_param_model_layers_13_mlp_up_proj_weight/max_abs":0.1171875,"train/tensor_param_model_layers_39_mlp_down_proj_weight/mean":-0.00238037109375,"train/tensor_param_model_layers_4_mlp_up_proj_weight/mean":-2.765655517578125e-05,"train/tensor_act_model_layers_31_residual_post_attn/min":122.5,"train/tensor_act_model_layers_96_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_51_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/max":0.1318359375,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_40_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_36_mlp_down_proj_weight/max":0.11376953125,"train/tensor_act_model_layers_26_mlp_up_proj/std":1.9296875,"train/tensor_act_model_layers_60_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_44_mlp_gate_proj/mean":0.76953125,"train/tensor_act_model_layers_67_mlp/range":0,"train/tensor_act_model_layers_84_residual_post_mlp/norm":696285.25,"train/tensor_act_model_layers_13_mlp_up_proj/mean":0.0264892578125,"train/tensor_act_model_layers_78_mlp_gate_proj/std":1.6484375,"train/tensor_act_model_layers_81_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_82_residual_pre_attn/max":328,"train/tensor_act_model_layers_47_residual_pre_attn/range":3,"train/tensor_act_model_layers_47_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_20_residual_post_attn/min":78.5,"train/tensor_act_model_layers_19_mlp/min":4,"train/tensor_act_model_layers_92_mlp_up_proj/std":1.84375,"train/tensor_param_model_layers_36_mlp_down_proj_weight/max_abs":0.1181640625,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_46_mlp_gate_proj/max_abs":4.59375,"train/tensor_act_model_layers_34_mlp_down_proj/max":16.125,"train/tensor_param_model_layers_48_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_95_mlp/min":4,"train/tensor_act_model_layers_62_residual_post_attn/mean":248,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_64_mlp_down_proj/std":0.50390625,"train/tensor_act_model_layers_1_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_residual_post_attn/max":352,"train/tensor_act_model_layers_41_residual_post_attn/max":165,"train/tensor_act_model_layers_2_mlp_gate_proj/mean":0.58203125,"train/tensor_act_model_layers_37_mlp_up_proj/range":8.59375,"train/tensor_act_model_layers_60_mlp_gate_proj/range":5.8203125,"train/tensor_act_model_layers_39_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_52_residual_pre_attn/range":3,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/mean":0.0042724609375,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_87_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_3_residual_post_mlp/max":16.625,"train/tensor_act_model_layers_14_residual_pre_attn/max_abs":57,"train/tensor_act_model_layers_29_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_20_mlp/range":0,"train/tensor_act_model_layers_36_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_66_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_87_mlp_down_proj/norm":31247.595703125,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_20_residual_post_mlp/norm":171793.453125,"train/tensor_act_model_layers_21_mlp_gate_proj/norm":5278.9501953125,"train/tensor_act_model_layers_75_residual_post_mlp/min":302,"train/tensor_act_model_layers_32_residual_post_attn/range":2.5,"train/tensor_act_model_layers_47_residual_pre_attn/min":186,"train/tensor_param_model_layers_67_mlp_up_proj_weight/mean":-0.0009613037109375,"train/tensor_act_model_layers_84_residual_post_attn/min":334,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_18_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_61_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_mlp_up_proj/max":4.21875,"train/tensor_act_model_layers_2_residual_post_mlp/range":0.9375,"train/tensor_act_model_layers_22_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_6_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_28_residual_pre_attn/min":110.5,"train/tensor_param_model_layers_63_mlp_down_proj_weight/std":0.036865234375,"train/tensor_act_model_layers_81_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_90_residual_pre_attn/max":360,"train/tensor_act_model_layers_16_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_28_mlp_up_proj_weight/range":0.23974609375,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_61_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_55_mlp_up_proj_weight/max":0.125,"train/tensor_act_model_layers_6_residual_pre_attn/max_abs":24.625,"train/tensor_act_model_layers_9_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_53_residual_pre_attn/norm":434028.53125,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_97_mlp/max_abs":4,"train/tensor_param_model_layers_89_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_mlp_up_proj/norm":5478.96630859375,"train/tensor_act_model_layers_7_residual_pre_attn/range":1.5,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_59_mlp_up_proj/range":8.921875,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/std":344,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/min":-0.11376953125,"train/tensor_act_model_layers_17_residual_post_attn/std":0.34375,"train/tensor_act_model_layers_96_mlp_up_proj/mean":0.193359375,"train/tensor_act_model_layers_0_residual_post_attn/frac_near_user_limit":0,"train/tensor_act_model_layers_26_mlp_gate_proj/std":1.7109375,"train/tensor_act_model_layers_15_mlp_up_proj/min":-3.953125,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_57_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_28_residual_pre_attn/std":0.359375,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_51_mlp/std":0,"train/tensor_act_model_layers_66_mlp_gate_proj/norm":4918.56591796875,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_47_mlp_up_proj/min":-4.28125,"train/tensor_act_model_layers_29_mlp/frac_near_user_limit":0,"eval/steps_per_second":0.474,"train/tensor_act_model_layers_81_residual_pre_attn/min":322,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_19_mlp_up_proj_weight/std":0.0311279296875,"train/tensor_act_model_layers_79_residual_post_mlp/min":318,"train/tensor_act_model_layers_94_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_6_mlp_up_proj/max_abs":4.4375,"train/tensor_act_model_layers_92_mlp_gate_proj/norm":5167.69091796875,"train/tensor_act_model_layers_77_residual_post_attn/range":2,"train/tensor_act_model_layers_80_mlp_up_proj/min":-4.40625,"train/tensor_act_model_layers_34_residual_pre_attn/std":0.31640625,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_78_mlp_down_proj_weight/std":0.041015625,"train/tensor_act_model_layers_71_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_54_mlp_down_proj/std":0.50390625,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_74_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/min":-0.125,"train/tensor_param_model_layers_97_mlp_up_proj_weight/max":0.1953125,"train/tensor_param_model_layers_92_mlp_up_proj_weight/range":0.322265625,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/max":0.12890625,"train/tensor_act_model_layers_12_residual_pre_attn/range":2,"train/tensor_act_model_layers_78_residual_post_attn/range":2,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_87_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_25_mlp_down_proj_weight/norm":5.6875,"train/tensor_act_model_layers_34_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/min":-8.3125,"train/tensor_act_model_layers_0_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_20_mlp_up_proj_weight/min":-0.125,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_5_mlp/max":4,"train/tensor_act_model_layers_74_mlp_down_proj/norm":30778.408203125,"train/tensor_act_model_layers_96_residual_post_attn/range":2,"train/tensor_act_model_layers_17_residual_pre_attn/mean":68,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_9_mlp_gate_proj/std":1.7265625,"train/tensor_act_model_layers_40_residual_post_attn/mean":160,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/norm":7.25,"train/tensor_act_model_layers_61_residual_pre_attn/max":245,"train/tensor_act_model_layers_30_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_72_mlp_up_proj_weight/min":-0.1328125,"train/tensor_act_model_layers_85_mlp_down_proj/max":18.75,"train/tensor_act_model_layers_6_mlp_up_proj/norm":5447.45263671875,"train/tensor_act_model_layers_23_residual_pre_attn/max":93,"train/tensor_act_model_layers_43_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_68_mlp_down_proj/norm":30859.70703125,"train/tensor_act_model_layers_59_mlp_gate_proj/max":4.78125,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_56_mlp_up_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_24_residual_post_attn/max_abs":97,"train/tensor_act_model_layers_34_residual_pre_attn/min":134,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/norm":6.40625,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_51_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_pre_attn/std":0.408203125,"train/tensor_act_model_layers_96_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_50_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_10_mlp_down_proj_weight/max_abs":0.11767578125,"train/tensor_act_model_layers_76_mlp/std":0,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_mlp_down_proj/range":2.125,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_80_mlp_gate_proj/mean":0.60546875,"train/tensor_act_model_layers_60_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_60_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_mlp_up_proj/norm":5337.6865234375,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/max_abs":0.212890625,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_75_residual_post_attn/range":2,"train/tensor_param_model_layers_99_mlp_up_proj_weight/max":0.169921875,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_32_residual_pre_attn/max":129,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_57_residual_pre_attn/norm":466732.78125,"train/tensor_act_model_layers_37_mlp_gate_proj/max":4.78125,"train/tensor_act_model_layers_12_residual_post_attn/max":49,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/range":0.3125,"train/tensor_act_model_layers_1_mlp_gate_proj/std":1.6328125,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_12_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_94_mlp_down_proj/mean":15.125,"train/tensor_act_model_layers_21_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/min":-0.150390625,"train/tensor_param_model_layers_34_mlp_down_proj_weight/range":0.25244140625,"train/tensor_act_model_layers_79_residual_pre_attn/range":2,"train/tensor_act_model_layers_43_mlp_down_proj/range":2.3125,"train/tensor_act_model_layers_66_mlp/max_abs":4,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_pre_attn/min":222,"train/tensor_param_model_layers_42_mlp_down_proj_weight/min":-0.1259765625,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/min":-0.12890625,"train/tensor_param_model_layers_8_mlp_down_proj_weight/range":0.21337890625,"train/tensor_act_model_layers_60_mlp_down_proj/max_abs":16.5,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_37_mlp/norm":8192,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/max_abs":0.1640625,"train/tensor_act_model_layers_50_mlp/max":4,"train/tensor_act_model_layers_35_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_47_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_62_mlp_down_proj/min":14.1875,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_66_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_73_residual_post_attn/range":2,"train/tensor_act_model_layers_7_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/max":0.13671875,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_69_mlp_down_proj/range":2.9375,"train/tensor_param_model_layers_88_mlp_up_proj_weight/max_abs":0.158203125,"train/tensor_act_model_layers_63_residual_post_mlp/max_abs":256,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_20_mlp/max":4,"train/tensor_act_model_layers_52_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_80_mlp/mean":4,"train/tensor_act_model_layers_93_mlp_down_proj/range":3.5,"train/tensor_act_model_layers_94_mlp_down_proj/min":14.0625,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/mean":0.004241943359375,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_18_mlp/max":4,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_77_mlp_down_proj/min":14.1875,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/range":0.2529296875,"train/tensor_act_model_layers_50_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_72_mlp_up_proj/mean":-0.28125,"train/tensor_act_model_layers_32_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_19_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_5_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_50_residual_pre_attn/min":198,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_27_mlp_down_proj_weight/mean":0.0010223388671875,"train/tensor_act_model_layers_69_residual_pre_attn/range":2,"train/tensor_act_model_layers_12_mlp_gate_proj/max":4.78125,"train/tensor_act_model_layers_27_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_mlp_gate_proj/mean":0.640625,"train/tensor_act_model_layers_66_residual_post_mlp/range":2,"train/tensor_act_model_layers_10_mlp/range":0,"train/tensor_act_model_layers_0_mlp_down_proj/max":99.5,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_28_mlp_up_proj_weight/mean":-2.0712614059448242e-06,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_28_residual_post_attn/max_abs":113,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_79_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_35_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/max_abs":0.16015625,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/range":0.271484375,"train/tensor_act_model_layers_41_residual_post_mlp/range":3,"train/tensor_act_model_layers_73_residual_post_attn/mean":292,"train/tensor_act_model_layers_35_residual_post_attn/norm":286572.78125,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_88_mlp/max":4,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_71_mlp_up_proj_weight/min":-0.1337890625,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_45_mlp_down_proj/min":14.25,"train/tensor_act_model_layers_42_mlp_up_proj/mean":0.1171875,"train/tensor_act_model_layers_65_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_64_residual_post_attn/min":254,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_45_mlp_up_proj_weight/min":-0.1259765625,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_64_residual_post_mlp/norm":532445.3125,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_5_mlp_down_proj_weight/mean":0.001068115234375,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_residual_post_attn/mean":104,"train/tensor_act_model_layers_51_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_14_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_86_mlp_gate_proj/mean":0.60546875,"train/tensor_act_model_layers_77_mlp_gate_proj/mean":0.498046875,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_55_residual_pre_attn/max_abs":221,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_98_mlp_down_proj/max":20,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/max":0.1494140625,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/max":1448,"train/tensor_act_model_layers_87_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_71_mlp_down_proj_weight/max_abs":0.1435546875,"train/tensor_act_model_layers_37_mlp_up_proj/mean":0.0751953125,"train/tensor_act_model_layers_88_residual_pre_attn/mean":352,"train/tensor_act_model_layers_41_residual_pre_attn/mean":164,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_12_residual_post_attn/std":0.267578125,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_70_mlp_up_proj/range":8.6875,"train/tensor_act_model_layers_45_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_95_mlp_up_proj_weight/max_abs":0.1650390625,"train/tensor_param_model_layers_28_mlp_up_proj_weight/std":0.032470703125,"train/tensor_param_model_layers_50_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_mlp_down_proj/std":0.392578125,"train/tensor_act_model_layers_93_mlp/min":4,"train/tensor_act_model_layers_71_mlp_gate_proj/mean":0.478515625,"train/tensor_act_model_layers_20_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_81_mlp_up_proj_weight/mean":-0.0006103515625,"train/tensor_act_model_layers_79_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/range":3880,"train/tensor_act_model_layers_37_residual_pre_attn/max":149,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_7_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_82_mlp_gate_proj/max_abs":5.21875,"train/tensor_act_model_layers_49_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_59_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_0_mlp_up_proj/range":15.4375,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_93_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_mlp_gate_proj/max_abs":4.4375,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_55_mlp_down_proj_weight/std":0.036865234375,"train/tensor_act_model_layers_3_mlp_gate_proj/max":4.25,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_64_mlp_up_proj/norm":5209.9873046875,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_70_residual_pre_attn/norm":573405.3125,"train/tensor_act_model_layers_37_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_99_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_3_residual_post_attn/norm":24561.234375,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_11_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_57_mlp_up_proj_weight/max":0.1279296875,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_85_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_33_mlp_down_proj_weight/mean":-0.00055694580078125,"train/tensor_param_model_layers_77_mlp_down_proj_weight/norm":7.25,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/range":0.2548828125,"train/tensor_param_model_layers_93_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_32_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_22_residual_post_attn/max":89,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_78_mlp_down_proj_weight/norm":7.40625,"train/tensor_act_model_layers_52_mlp_up_proj/mean":-0.10400390625,"train/tensor_act_model_layers_2_residual_pre_attn/range":0.875,"train/tensor_act_model_layers_38_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_61_residual_post_mlp/std":0.416015625,"train/tensor_param_model_layers_42_mlp_down_proj_weight/norm":6.125,"train/tensor_act_model_layers_55_residual_post_mlp/max":225,"train/tensor_act_model_layers_23_residual_post_mlp/std":0.359375,"train/tensor_param_model_layers_66_mlp_up_proj_weight/max_abs":0.1494140625,"train/tensor_act_model_layers_13_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_24_mlp_down_proj_weight/max_abs":0.126953125,"train/tensor_param_model_layers_55_mlp_down_proj_weight/max_abs":0.134765625,"train/tensor_param_model_layers_50_mlp_down_proj_weight/norm":6.4375,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_41_mlp_down_proj_weight/std":0.033447265625,"train/tensor_param_model_layers_9_mlp_down_proj_weight/max":0.10595703125,"train/tensor_param_model_layers_5_mlp_down_proj_weight/range":0.22607421875,"train/tensor_act_model_layers_71_mlp_up_proj/range":8.59375,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/frac_near_user_limit":0.98876953125,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_42_mlp_down_proj_weight/range":0.2529296875,"train/tensor_param_model_layers_7_mlp_down_proj_weight/max":0.09521484375,"train/tensor_param_model_layers_47_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_74_residual_pre_attn/range":2,"train/tensor_act_model_layers_7_residual_post_mlp/range":1.625,"train/tensor_act_model_layers_49_mlp_down_proj/norm":30248.05859375,"train/tensor_act_model_layers_83_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_23_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_75_residual_post_attn/max":300,"train/tensor_act_model_layers_13_mlp_down_proj/norm":29899.013671875,"train/tensor_act_model_layers_92_mlp_gate_proj/max_abs":5.3125,"train/tensor_act_model_layers_33_residual_post_attn/max":133,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_4_mlp_down_proj_weight/range":0.22119140625,"train/tensor_act_model_layers_94_mlp_up_proj/norm":5089.162109375,"train/tensor_act_model_layers_74_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_48_mlp/norm":8192,"train/tensor_act_model_layers_6_residual_post_mlp/min":27.125,"train/tensor_act_model_layers_30_mlp_down_proj/range":2.0625,"train/tensor_act_model_layers_71_residual_pre_attn/max":284,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_65_mlp_up_proj/min":-4.9375,"train/tensor_act_model_layers_17_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_73_mlp_down_proj/norm":31050.55859375,"train/tensor_act_model_layers_37_mlp/range":0,"train/tensor_act_model_layers_20_residual_post_attn/range":2.5,"train/tensor_act_model_layers_26_residual_pre_attn/mean":104,"train/tensor_act_model_layers_56_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_68_residual_post_attn/max_abs":272,"train/tensor_act_model_layers_52_residual_pre_attn/max_abs":209,"train/tensor_act_model_layers_22_mlp_down_proj/range":1.375,"train/tensor_param_model_layers_17_mlp_up_proj_weight/std":0.0308837890625,"train/tensor_act_model_layers_59_residual_post_mlp/max":241,"train/tensor_param_model_layers_88_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_51_residual_pre_attn/max":205,"train/tensor_act_model_layers_22_residual_post_attn/mean":88,"train/tensor_act_model_layers_7_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_50_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_38_residual_post_attn/max_abs":153,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/mean":0.005096435546875,"train/tensor_act_model_layers_42_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_72_mlp/max":4,"train/tensor_act_model_layers_46_mlp/range":0,"train/tensor_act_model_layers_61_mlp_down_proj/std":0.470703125,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_87_residual_pre_attn/max":348,"train/tensor_act_model_layers_76_residual_post_attn/mean":304,"train/tensor_act_model_layers_88_residual_post_attn/range":2,"train/tensor_act_model_layers_41_mlp_gate_proj/mean":0.54296875,"train/tensor_act_model_layers_3_residual_post_attn/min":11.375,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_85_mlp_up_proj_weight/norm":8.25,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_98_mlp/norm":8192,"train/tensor_act_model_layers_85_mlp_down_proj/max_abs":18.75,"train/tensor_act_model_layers_5_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_14_mlp_down_proj_weight/norm":5.4375,"train/tensor_param_model_layers_63_mlp_down_proj_weight/max":0.142578125,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/norm":6.6875,"train/tensor_act_model_layers_28_mlp_up_proj/norm":5600.0107421875,"train/tensor_act_model_layers_29_residual_post_mlp/max":121,"train/tensor_act_model_layers_17_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_57_residual_post_attn/norm":466716.8125,"train/tensor_act_model_layers_70_residual_post_mlp/min":282,"train/tensor_param_model_layers_60_mlp_up_proj_weight/std":0.037109375,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/min":-0.158203125,"train/tensor_act_model_layers_88_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_52_mlp_up_proj_weight/norm":6.625,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_50_mlp_up_proj/range":8.3125,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_76_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_59_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_30_mlp_up_proj_weight/max_abs":0.11669921875,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_44_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/mean":0.0033721923828125,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_65_residual_pre_attn/max":260,"train/tensor_param_model_layers_30_mlp_up_proj_weight/min":-0.1142578125,"train/tensor_act_model_layers_65_mlp_down_proj/max":16.75,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_14_mlp_down_proj/max":15.6875,"train/tensor_act_model_layers_1_residual_pre_attn/norm":8190.14501953125,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_37_mlp_gate_proj/range":6.0703125,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_91_mlp_down_proj/max_abs":20,"train/tensor_act_model_layers_20_mlp_down_proj/max_abs":16.25,"train/tensor_act_model_layers_44_residual_pre_attn/mean":176,"train/tensor_act_model_layers_95_mlp_up_proj/norm":5158.568359375,"train/tensor_act_model_layers_33_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_75_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_19_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_residual_post_mlp/mean":288,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_mlp_down_proj/range":2.875,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_mlp_up_proj/range":8.96875,"train/tensor_act_model_layers_99_mlp_up_proj/std":1.8984375,"train/tensor_param_model_layers_0_mlp_down_proj_weight/mean":0.00372314453125,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/max_abs":0.154296875,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_8_mlp_up_proj/norm":5439.52294921875,"train/tensor_act_model_layers_85_mlp_up_proj/mean":0.09814453125,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_91_mlp_down_proj/norm":30938.615234375,"train/tensor_act_model_layers_82_residual_post_mlp/range":2,"train/tensor_param_model_layers_55_mlp_down_proj_weight/min":-0.130859375,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_59_mlp/norm":8192,"train/tensor_act_model_layers_48_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_22_residual_post_attn/std":0.359375,"train/tensor_act_model_layers_3_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_89_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_6_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_residual_post_mlp/range":2,"train/tensor_param_model_layers_50_mlp_up_proj_weight/norm":6.46875,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_40_residual_post_attn/range":3,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_19_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_97_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_20_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_33_mlp_gate_proj/max_abs":5.03125,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_83_residual_post_attn/max":332,"train/tensor_act_model_layers_49_residual_post_mlp/min":198,"train/tensor_act_model_layers_38_mlp_up_proj/range":8.84375,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_38_mlp_gate_proj/max":4.65625,"train/tensor_act_model_layers_33_residual_pre_attn/mean":132,"train/tensor_act_model_layers_44_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_99_mlp_down_proj/max":21.875,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_75_residual_pre_attn/mean":300,"train/tensor_act_model_layers_85_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_11_mlp_up_proj/max_abs":4.03125,"train/tensor_act_model_layers_77_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_87_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_26_residual_post_attn/max":105,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_68_mlp/min":4,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_24_residual_post_mlp/min":98.5,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_10_mlp/max_abs":4,"train/tensor_param_model_layers_57_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_63_mlp/min":4,"train/tensor_param_model_layers_38_mlp_up_proj_weight/norm":6.09375,"train/tensor_act_model_layers_68_residual_post_mlp/max":276,"train/tensor_act_model_layers_65_residual_post_attn/max_abs":260,"train/tensor_param_model_layers_67_mlp_up_proj_weight/max_abs":0.138671875,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_24_mlp_down_proj/norm":29917.611328125,"train/tensor_param_model_layers_20_mlp_down_proj_weight/min":-0.123046875,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_41_mlp/max":4,"train/tensor_act_model_layers_59_mlp/range":0,"train/tensor_act_model_layers_84_residual_post_attn/mean":336,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_37_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_22_residual_pre_attn/mean":88,"train/tensor_act_model_layers_87_residual_post_mlp/mean":352,"train/tensor_param_model_layers_23_mlp_down_proj_weight/max_abs":0.1259765625,"train/tensor_act_model_layers_61_mlp_gate_proj/range":5.734375,"train/tensor_act_model_layers_27_mlp_gate_proj/max":4.84375,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_29_mlp_up_proj/std":1.9375,"train/tensor_param_model_layers_26_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_20_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_54_mlp_up_proj/max":4,"train/tensor_param_model_layers_96_mlp_down_proj_weight/norm":9.375,"train/tensor_act_model_layers_51_residual_post_mlp/range":3,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_95_mlp_up_proj_weight/max":0.1650390625,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_73_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_15_mlp_up_proj/mean":0.0556640625,"train/tensor_param_model_layers_28_mlp_down_proj_weight/norm":5.8125,"train/tensor_act_model_layers_42_residual_post_mlp/mean":172,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_57_mlp_down_proj_weight/mean":-0.000732421875,"train/tensor_act_model_layers_92_residual_post_attn/min":366,"train/tensor_act_model_layers_50_mlp_gate_proj/max":4.125,"train/tensor_act_model_layers_33_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_45_mlp/max":4,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_63_mlp_down_proj_weight/mean":2.1576881408691406e-05,"train/tensor_act_model_layers_7_mlp/max":4,"train/tensor_act_model_layers_74_residual_post_attn/range":2,"train/tensor_act_model_layers_95_mlp_gate_proj/mean":0.6796875,"train/tensor_act_model_layers_74_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_53_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_64_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_31_mlp_up_proj/mean":0.154296875,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_residual_post_attn/max":221,"train/tensor_act_model_layers_32_residual_pre_attn/norm":261786.84375,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_31_mlp_gate_proj/norm":5248.2548828125,"train/tensor_act_model_layers_92_residual_post_attn/mean":368,"train/tensor_act_model_layers_26_mlp_up_proj/mean":0.0634765625,"train/tensor_act_model_layers_0_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_88_mlp_down_proj_weight/range":0.2900390625,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/min":-0.11669921875,"train/tensor_act_model_layers_81_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_83_mlp_down_proj_weight/min":-0.1484375,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/std":0.038818359375,"train/tensor_act_model_layers_27_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_92_residual_post_mlp/norm":761821.25,"train/tensor_act_model_layers_43_residual_post_mlp/max":177,"train/tensor_param_model_layers_99_mlp_down_proj_weight/norm":9.5,"train/tensor_param_model_layers_78_mlp_down_proj_weight/max":0.142578125,"train/tensor_param_model_layers_48_mlp_up_proj_weight/norm":6.4375,"train/tensor_param_model_layers_68_mlp_up_proj_weight/std":0.038330078125,"train/tensor_param_model_layers_29_mlp_up_proj_weight/mean":0.0003757476806640625,"train/tensor_act_model_layers_49_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_99_mlp_down_proj_weight/min":-0.1689453125,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_37_mlp_gate_proj/mean":0.515625,"train/tensor_act_model_layers_77_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_mlp_gate_proj/min":-1.3125,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_43_residual_pre_attn/max_abs":173,"train/tensor_act_model_layers_76_residual_pre_attn/min":302,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_83_residual_pre_attn/mean":332,"train/tensor_act_model_layers_25_mlp_down_proj/max":16.375,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/max":2000,"train/tensor_act_model_layers_98_mlp/std":0,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/std":0.0419921875,"train/tensor_param_model_layers_4_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_52_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_post_attn/range":2,"train/tensor_act_model_layers_76_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_26_mlp_gate_proj/max_abs":5.03125,"train/tensor_act_model_layers_79_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_97_mlp_down_proj_weight/range":0.3583984375,"train/tensor_act_model_layers_47_mlp_down_proj/range":1.875,"train/tensor_act_model_layers_71_mlp_gate_proj/max":4.3125,"train/tensor_act_model_layers_44_residual_post_attn/max":177,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/range":0.24853515625,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_74_mlp_down_proj/max":17,"train/tensor_act_model_layers_90_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_26_residual_post_attn/std":0.359375,"train/tensor_act_model_layers_76_mlp_up_proj/max_abs":5,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_39_residual_post_mlp/max_abs":161,"train/tensor_act_model_layers_86_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_23_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_96_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_74_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_20_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_1_mlp_gate_proj/max_abs":4.9375,"train/tensor_act_model_layers_19_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_55_residual_post_attn/std":0.408203125,"train/tensor_act_model_layers_12_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_77_mlp_down_proj/range":4.8125,"train/tensor_act_model_layers_13_residual_post_attn/min":51,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_11_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_51_mlp_down_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_26_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_69_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_48_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_71_mlp_up_proj_weight/norm":7.09375,"train/tensor_param_model_layers_84_mlp_down_proj_weight/norm":8.125,"train/tensor_act_model_layers_37_residual_post_attn/range":3,"train/tensor_param_model_layers_48_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_89_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_43_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_84_mlp_up_proj_weight/max":0.1484375,"train/tensor_act_model_layers_41_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_1_mlp/min":3.984375,"train/tensor_act_model_layers_78_mlp_gate_proj/min":-1.2578125,"train/tensor_act_model_layers_91_mlp/min":4,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/std":0.04150390625,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/min":-0.125,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_91_mlp/max":4,"train/tensor_act_model_layers_82_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_80_mlp_down_proj/max_abs":17.25,"train/tensor_act_model_layers_63_mlp_up_proj/std":1.84375,"train/tensor_act_model_layers_53_mlp_up_proj/std":1.8515625,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_13_residual_post_mlp/norm":114513.6796875,"train/tensor_act_model_layers_67_residual_pre_attn/norm":548829.3125,"train/tensor_act_model_layers_53_mlp_down_proj/mean":14.8125,"train/tensor_act_model_layers_40_mlp_up_proj/range":9.0625,"train/tensor_param_model_layers_61_mlp_up_proj_weight/std":0.037353515625,"train/tensor_act_model_layers_15_residual_post_attn/max_abs":61,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_86_mlp_up_proj_weight/min":-0.1435546875,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_93_mlp_up_proj/mean":0.09375,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/max":0.1611328125,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_69_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_33_mlp_gate_proj/max":5.03125,"train/tensor_act_model_layers_40_residual_post_mlp/max":165,"train/tensor_param_model_layers_90_mlp_down_proj_weight/mean":-0.00064849853515625,"train/tensor_act_model_layers_14_mlp_gate_proj/range":5.5625,"train/tensor_param_model_layers_63_mlp_down_proj_weight/min":-0.12890625,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_42_mlp_gate_proj/range":5.75,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_56_mlp/std":0,"train/tensor_act_model_layers_8_mlp_gate_proj/range":5.8046875,"train/tensor_act_model_layers_76_mlp/min":4,"train/tensor_act_model_layers_87_mlp_down_proj/min":14.0625,"train/tensor_act_model_layers_38_residual_pre_attn/max_abs":153,"train/tensor_act_model_layers_82_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_0_mlp_up_proj/mean":0.028564453125,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/std":0.031982421875,"train/tensor_act_model_layers_61_mlp/mean":4,"train/tensor_param_model_layers_96_mlp_up_proj_weight/std":0.0537109375,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_23_mlp_gate_proj/range":6.0390625,"train/tensor_param_model_layers_47_mlp_up_proj_weight/max":0.12109375,"train/tensor_param_model_layers_86_mlp_down_proj_weight/range":0.2900390625,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/norm":5.375,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_39_mlp_up_proj_weight/mean":-0.00238037109375,"train/tensor_act_model_layers_51_mlp_gate_proj/min":-1.25,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_64_mlp_up_proj_weight/min":-0.126953125,"train/tensor_act_model_layers_99_residual_pre_attn/mean":396,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_9_mlp_down_proj_weight/mean":-0.00058746337890625,"train/tensor_act_model_layers_90_mlp_up_proj/std":1.7578125,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_48_mlp/range":0,"train/tensor_act_model_layers_58_residual_post_mlp/mean":236,"train/tensor_act_model_layers_4_residual_post_attn/std":0.1640625,"train/tensor_act_model_layers_4_residual_post_mlp/std":0.16796875,"train/tensor_act_model_layers_59_residual_post_attn/max_abs":237,"train/tensor_act_model_layers_53_mlp_gate_proj/min":-1.265625,"train/tensor_act_model_layers_48_mlp_down_proj/std":0.455078125,"train/tensor_act_model_layers_5_residual_post_attn/mean":20,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_91_mlp_down_proj/mean":15.0625,"train/loss":8.109441375732422,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_84_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_17_mlp_gate_proj/min":-1.28125,"train/tensor_act_model_layers_12_residual_pre_attn/std":0.26171875,"train/tensor_act_model_layers_30_mlp_up_proj/range":8.46875,"train/tensor_act_model_layers_51_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_88_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_79_mlp_up_proj_weight/max_abs":0.1455078125,"train/tensor_act_model_layers_78_mlp_up_proj/norm":5166.52001953125,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_35_mlp_gate_proj/max_abs":4.46875,"train/tensor_act_model_layers_47_residual_post_mlp/max_abs":193,"train/tensor_act_model_layers_10_mlp_gate_proj/std":1.71875,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/norm":6.03125,"train/tensor_act_model_layers_76_mlp_down_proj/std":0.59765625,"train/tensor_act_model_layers_12_residual_pre_attn/min":47,"train/tensor_act_model_layers_88_residual_pre_attn/min":350,"train/tensor_act_model_layers_84_mlp_up_proj/std":1.7890625,"train/tensor_param_model_layers_55_mlp_down_proj_weight/max":0.134765625,"train/tensor_act_model_layers_72_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_72_mlp_up_proj_weight/norm":7.0625,"train/tensor_act_model_layers_56_mlp_up_proj/mean":-0.037353515625,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_31_residual_post_mlp/max":129,"train/tensor_param_model_layers_15_mlp_up_proj_weight/mean":0.00043487548828125,"train/tensor_act_model_layers_24_mlp_up_proj/min":-4.375,"train/tensor_act_model_layers_23_mlp_up_proj/min":-4.875,"train/tensor_act_model_layers_74_mlp/range":0,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_0_residual_pre_attn/norm":186.0115966796875,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/range":0.2587890625,"train/tensor_act_model_layers_42_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_20_mlp_gate_proj/std":1.6796875,"train/tensor_act_model_layers_53_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_58_residual_post_mlp/std":0.416015625,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_67_mlp/std":0,"train/tensor_act_model_layers_78_mlp_gate_proj/norm":4931.880859375,"train/tensor_param_model_layers_27_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_90_residual_pre_attn/range":2,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_65_residual_pre_attn/max_abs":260,"train/tensor_act_model_layers_86_residual_post_attn/mean":344,"train/tensor_act_model_layers_52_mlp_down_proj/max":16.625,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/norm":7.0625,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_97_mlp_down_proj_weight/min":-0.1767578125,"train/tensor_act_model_layers_19_residual_pre_attn/mean":76,"train/tensor_act_model_layers_8_mlp_gate_proj/mean":0.45703125,"train/tensor_act_model_layers_80_mlp_gate_proj/range":6.1171875,"train/tensor_act_model_layers_42_mlp/norm":8192,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_20_mlp/max_abs":4,"train/tensor_param_model_layers_6_mlp_down_proj_weight/std":0.029296875,"train/tensor_act_model_layers_95_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_14_mlp_down_proj_weight/max":0.1142578125,"train/tensor_act_model_layers_2_mlp_gate_proj/min":-1.375,"train/tensor_param_model_layers_40_mlp_down_proj_weight/max_abs":0.126953125,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_48_mlp_gate_proj/max_abs":5.3125,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/norm":5.6875,"train/tensor_act_model_layers_91_residual_pre_attn/max":364,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_36_residual_pre_attn/mean":144,"train/tensor_act_model_layers_12_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_80_residual_post_attn/max_abs":320,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_66_mlp_up_proj_weight/mean":-0.0003337860107421875,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_34_mlp/max":4,"train/tensor_param_model_layers_13_mlp_up_proj_weight/max":0.1171875,"train/tensor_act_model_layers_9_residual_pre_attn/norm":73657.7421875,"train/tensor_act_model_layers_49_mlp_down_proj/max":16.125,"train/tensor_param_model_layers_98_mlp_down_proj_weight/std":0.052490234375,"train/tensor_param_model_layers_91_mlp_up_proj_weight/norm":8.8125,"train/tensor_act_model_layers_31_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_mlp_down_proj/mean":14.75,"train/tensor_act_model_layers_0_residual_pre_attn/range":0.5859375,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_21_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_89_mlp_down_proj/range":2.5625,"train/tensor_act_model_layers_67_residual_post_mlp/max":272,"train/tensor_act_model_layers_73_mlp_gate_proj/range":5.6640625,"train/tensor_act_model_layers_81_mlp_up_proj/norm":5279.68310546875,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_16_residual_pre_attn/mean":63.75,"train/tensor_act_model_layers_89_mlp_down_proj/max":16.625,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_29_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_13_mlp_down_proj/mean":14.625,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_15_mlp_gate_proj/max_abs":4.34375,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/norm":5.375,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_66_mlp_up_proj/min":-4.15625,"train/tensor_param_model_layers_3_mlp_down_proj_weight/min":-0.10693359375,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_95_residual_pre_attn/mean":380,"train/tensor_act_model_layers_26_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_86_mlp_down_proj_weight/norm":8.1875,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/norm":5.75,"train/tensor_act_model_layers_17_mlp_down_proj/norm":29927.61328125,"train/tensor_act_model_layers_73_mlp_down_proj/mean":15.125,"train/tensor_act_model_layers_3_mlp_gate_proj/min":-1.3984375,"train/tensor_act_model_layers_32_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_50_residual_pre_attn/max_abs":201,"train/tensor_act_model_layers_63_residual_pre_attn/max":253,"train/tensor_act_model_layers_41_mlp_down_proj/std":0.41015625,"train/tensor_act_model_layers_66_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_25_mlp/mean":4,"train/tensor_act_model_layers_55_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_post_attn/min":274,"train/tensor_act_model_layers_33_mlp_up_proj/mean":-0.2099609375,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/norm":9.8125,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_48_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_residual_post_attn/min":242,"train/tensor_act_model_layers_17_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_94_mlp/range":0,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/mean":0.005828857421875,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_17_mlp_down_proj/range":2.25,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_3_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_1_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_0_mlp_down_proj_weight/max_abs":0.125,"train/tensor_param_model_layers_99_mlp_down_proj_weight/max":0.1796875,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_83_residual_pre_attn/max":332,"train/tensor_act_model_layers_61_residual_post_mlp/norm":507676.75,"train/tensor_act_model_layers_19_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/norm":7.6875,"train/tensor_act_model_layers_16_residual_post_attn/std":0.31640625,"train/tensor_param_model_layers_87_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_10_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_7_residual_pre_attn/mean":28,"train/tensor_param_model_layers_85_mlp_up_proj_weight/max_abs":0.1474609375,"train/tensor_param_model_layers_83_mlp_up_proj_weight/max_abs":0.140625,"train/tensor_act_model_layers_69_residual_post_mlp/norm":573405.3125,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_95_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_31_residual_post_attn/range":2.5,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/max":0.115234375,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_residual_post_attn/range":3,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_6_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_43_mlp_down_proj/max_abs":16.5,"train/tensor_act_model_layers_99_residual_post_attn/max":396,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_6_mlp_gate_proj/min":-1.34375,"train/tensor_act_model_layers_19_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_94_mlp_down_proj_weight/norm":9.125,"train/tensor_act_model_layers_84_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_87_mlp_up_proj_weight/min":-0.1591796875,"train/tensor_param_model_layers_51_mlp_down_proj_weight/max":0.12890625,"train/tensor_act_model_layers_29_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_35_mlp_up_proj/norm":5382.1103515625,"train/tensor_act_model_layers_12_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_19_residual_post_attn/std":0.357421875,"train/tensor_param_model_layers_59_mlp_up_proj_weight/range":0.255859375,"train/tensor_param_model_layers_30_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_79_mlp_down_proj/max":16.5,"train/tensor_act_model_layers_53_mlp_down_proj/norm":30408.0234375,"train/tensor_act_model_layers_36_residual_post_attn/std":0.31640625,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_37_residual_post_mlp/std":0.31640625,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/std":0.038330078125,"train/tensor_act_model_layers_69_mlp_up_proj/min":-4.1875,"train/tensor_act_model_layers_52_mlp_gate_proj/range":6.0625,"train/tensor_act_model_layers_27_mlp/mean":4,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_56_mlp_up_proj_weight/mean":-0.0002899169921875,"train/tensor_act_model_layers_95_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_81_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_16_residual_post_attn/max":65,"train/tensor_act_model_layers_86_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_20_mlp_down_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_56_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_33_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp/norm":8192,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_95_residual_post_attn/max":380,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_post_attn/norm":245497.171875,"train/tensor_act_model_layers_24_mlp_down_proj/mean":14.625,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_64_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_62_residual_pre_attn/std":0.416015625,"train/tensor_act_model_layers_30_mlp_gate_proj/max":4.40625,"train/tensor_param_model_layers_16_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/range":0.259765625,"train/tensor_act_model_layers_19_residual_post_attn/norm":155418.359375,"train/tensor_act_model_layers_62_residual_pre_attn/min":246,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/std":0.048095703125,"train/tensor_act_model_layers_15_residual_post_attn/mean":60,"train/tensor_act_model_layers_88_mlp_down_proj/std":0.5859375,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_25_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/max_abs":0.1376953125,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/std":0.040771484375,"train/tensor_param_model_layers_42_mlp_down_proj_weight/std":0.033935546875,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/norm":5.59375,"train/tensor_param_model_layers_0_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/mean":0.004058837890625,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/norm":5.84375,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_42_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_55_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_88_mlp_up_proj_weight/std":0.046875,"train/tensor_act_model_layers_80_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_32_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_73_mlp_up_proj/norm":5118.564453125,"train/tensor_act_model_layers_19_residual_post_attn/max":77,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_74_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_78_mlp_down_proj_weight/min":-0.1357421875,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/range":0.28125,"train/tensor_act_model_layers_63_mlp_gate_proj/std":1.65625,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/std":0.032958984375,"train/tensor_act_model_layers_36_residual_post_attn/max":145,"eval/loss":8.10374927520752,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_83_mlp_gate_proj/min":-1.3125,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_83_mlp_down_proj_weight/max":0.1376953125,"train/tensor_act_model_layers_91_mlp_down_proj/std":0.79296875,"train/tensor_param_model_layers_22_mlp_down_proj_weight/max_abs":0.11572265625,"train/tensor_act_model_layers_41_residual_pre_attn/max_abs":165,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_11_mlp_gate_proj/norm":5172.32470703125,"train/tensor_act_model_layers_20_mlp_down_proj/range":2.25,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_83_residual_post_attn/max_abs":332,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_11_mlp_gate_proj/std":1.7265625,"train/tensor_act_model_layers_82_mlp_gate_proj/std":1.6171875,"train/tensor_param_model_layers_13_mlp_down_proj_weight/range":0.220703125,"train/tensor_act_model_layers_31_mlp/range":0,"train/tensor_act_model_layers_99_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_63_mlp_down_proj_weight/range":0.271484375,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_27_mlp_down_proj/norm":29899.71484375,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_33_mlp_up_proj/range":9.4375,"train/tensor_act_model_layers_59_mlp_down_proj/max_abs":17.25,"train/tensor_act_model_layers_77_residual_pre_attn/norm":630749.3125,"train/tensor_act_model_layers_30_mlp_up_proj/std":1.90625,"train/tensor_act_model_layers_14_residual_pre_attn/norm":114513.6796875,"train/tensor_act_model_layers_40_mlp_gate_proj/max":4.84375,"train/tensor_act_model_layers_19_mlp/std":0,"train/tensor_act_model_layers_75_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_89_residual_post_mlp/range":2,"train/tensor_param_model_layers_32_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_7_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_55_residual_pre_attn/mean":220,"train/tensor_param_model_layers_51_mlp_up_proj_weight/range":0.2548828125,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_6_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_32_mlp/min":4,"train/tensor_act_model_layers_47_mlp_down_proj/max":16,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_64_mlp_down_proj_weight/range":0.2578125,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_36_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_41_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_97_mlp_down_proj_weight/max":0.181640625,"train/tensor_param_model_layers_14_mlp_down_proj_weight/min":-0.1083984375,"train/tensor_param_model_layers_46_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/mean":0.00518798828125,"train/tensor_act_model_layers_0_mlp_down_proj/frac_near_user_limit":0.9976546764373779,"train/tensor_param_model_layers_36_mlp_down_proj_weight/std":0.032958984375,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/std":0.036865234375,"train/tensor_act_model_layers_55_mlp/range":0,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_66_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_54_residual_post_mlp/mean":220,"train/tensor_act_model_layers_41_mlp_down_proj/max":16.25,"train/tensor_act_model_layers_65_mlp/max_abs":4,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_88_mlp_up_proj/min":-4.5625,"train/tensor_act_model_layers_72_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_mlp_down_proj/std":0.41796875,"train/tensor_act_model_layers_56_residual_post_mlp/max":229,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_24_mlp/norm":8192,"train/tensor_act_model_layers_35_mlp/std":0,"train/tensor_act_model_layers_66_mlp/max":4,"train/tensor_act_model_layers_6_residual_post_mlp/std":0.1953125,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_residual_post_mlp/range":3,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/std":0.04736328125,"train/tensor_act_model_layers_12_mlp_up_proj/norm":5473.5078125,"train/tensor_act_model_layers_52_mlp_down_proj/std":0.490234375,"train/tensor_act_model_layers_7_mlp/max_abs":4,"train/tensor_param_model_layers_54_mlp_down_proj_weight/range":0.2509765625,"train/tensor_act_model_layers_51_mlp_gate_proj/norm":5012.34814453125,"train/tensor_act_model_layers_16_mlp/std":0,"train/tensor_act_model_layers_60_mlp_up_proj/norm":5380.9833984375,"train/tensor_param_model_layers_72_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_5_residual_post_attn/norm":40926.046875,"train/tensor_act_model_layers_7_mlp_up_proj/mean":0.014404296875,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_36_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_69_mlp_down_proj/norm":30594.291015625,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_28_mlp_down_proj_weight/max":0.11767578125,"train/tensor_param_model_layers_17_mlp_up_proj_weight/norm":5.59375,"train/tensor_act_model_layers_24_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_34_residual_post_mlp/max":141,"train/tensor_act_model_layers_60_residual_pre_attn/range":3,"train/tensor_act_model_layers_53_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_99_mlp/max_abs":4,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_61_mlp_up_proj/max":3.984375,"train/tensor_param_model_layers_39_mlp_up_proj_weight/range":0.24853515625,"train/tensor_act_model_layers_5_residual_post_attn/range":1.375,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_88_mlp_gate_proj/range":6.0390625,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_8_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_78_mlp_up_proj/min":-4.53125,"train/tensor_act_model_layers_60_mlp_gate_proj/min":-1.3828125,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_41_mlp_down_proj/max_abs":16.25,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/max_abs":0.166015625,"train/tensor_act_model_layers_43_mlp/norm":8192,"train/tensor_act_model_layers_94_mlp/max":4,"train/tensor_act_model_layers_75_mlp/range":0,"train/tensor_act_model_layers_93_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_mlp_down_proj/max":16.75,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_47_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_96_mlp_down_proj_weight/std":0.052001953125,"train/tensor_param_model_layers_10_mlp_down_proj_weight/min":-0.11767578125,"train/tensor_act_model_layers_12_mlp_gate_proj/mean":0.376953125,"train/tensor_param_model_layers_13_mlp_down_proj_weight/std":0.030029296875,"train/tensor_act_model_layers_41_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_7_residual_post_attn/min":27.125,"train/tensor_act_model_layers_2_residual_pre_attn/min":7.5,"train/tensor_act_model_layers_31_mlp/max_abs":4,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_18_mlp_up_proj_weight/range":0.2119140625,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_48_mlp_down_proj/max":17,"train/tensor_act_model_layers_97_residual_post_mlp/mean":392,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/max":0.14453125,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_11_residual_pre_attn/max":44.75,"train/tensor_act_model_layers_60_residual_pre_attn/max":241,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_17_residual_post_mlp/range":2,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_17_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_36_mlp_down_proj/max_abs":15.6875,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_1_mlp_down_proj_weight/min":-0.125,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/norm":5.96875,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_82_residual_post_mlp/mean":332,"train/tensor_act_model_layers_41_residual_post_mlp/min":166,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_16_mlp_down_proj_weight/range":0.22998046875,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_13_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_37_mlp_up_proj/max":4.71875,"train/tensor_act_model_layers_37_residual_post_mlp/range":3,"train/tensor_param_model_layers_8_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_17_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_15_residual_post_attn/norm":122671.5546875,"train/tensor_param_model_layers_91_mlp_down_proj_weight/max":0.142578125,"train/tensor_act_model_layers_24_mlp_gate_proj/max":4.46875,"train/tensor_act_model_layers_79_mlp/max":4,"train/tensor_act_model_layers_3_residual_pre_attn/max":12.4375,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_30_residual_post_attn/mean":120,"train/tensor_param_model_layers_22_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_84_residual_post_mlp/min":338,"train/tensor_act_model_layers_34_residual_post_attn/max_abs":137,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/std":0.03564453125,"train/tensor_act_model_layers_42_mlp_up_proj/min":-4.625,"train/tensor_act_model_layers_19_residual_post_attn/mean":76,"train/tensor_act_model_layers_50_residual_post_mlp/mean":204,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_3_mlp_up_proj_weight/std":0.0303955078125,"train/tensor_param_model_layers_5_mlp_up_proj_weight/norm":5.40625,"train/tensor_act_model_layers_71_mlp/min":4,"train/tensor_act_model_layers_68_residual_post_mlp/range":2,"train/tensor_act_model_layers_23_mlp/norm":8192,"train/tensor_act_model_layers_98_mlp_down_proj/norm":30972.515625,"train/tensor_act_model_layers_50_residual_post_mlp/max":205,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_93_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_57_mlp_up_proj/max":5,"train/tensor_param_model_layers_17_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/max":0.13671875,"train/tensor_act_model_layers_17_residual_pre_attn/max":69,"train/tensor_act_model_layers_5_residual_post_attn/max":20.625,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_18_mlp_down_proj/norm":29843.671875,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_72_mlp_down_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_27_residual_post_mlp/std":0.359375,"train/tensor_act_model_layers_5_mlp_up_proj/norm":5333.0947265625,"train/tensor_act_model_layers_98_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_14_residual_post_mlp/range":2,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_42_residual_post_attn/range":3,"train/tensor_act_model_layers_72_mlp_down_proj/mean":14.9375,"train/tensor_act_model_layers_82_mlp/std":0,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_89_mlp_down_proj_weight/norm":8.375,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/std":0.031494140625,"train/tensor_act_model_layers_24_mlp_up_proj/range":8.53125,"train/tensor_param_model_layers_81_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/max":0.125,"train/tensor_param_model_layers_69_mlp_up_proj_weight/mean":-0.0006866455078125,"train/tensor_act_model_layers_28_mlp/max":4,"train/tensor_act_model_layers_15_mlp_down_proj/mean":14.6875,"train/tensor_param_model_layers_19_mlp_down_proj_weight/mean":0.00063323974609375,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_84_mlp_up_proj_weight/max_abs":0.1484375,"train/tensor_act_model_layers_51_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/max":0.1328125,"train/tensor_act_model_layers_68_residual_post_mlp/mean":276,"train/tensor_act_model_layers_54_mlp_gate_proj/mean":0.546875,"train/tensor_act_model_layers_45_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_79_mlp/std":0,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_91_mlp_down_proj/max":20,"train/tensor_act_model_layers_28_residual_post_attn/range":2.5,"train/tensor_param_model_layers_21_mlp_up_proj_weight/max_abs":0.12353515625,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_35_mlp/max_abs":4,"train/tensor_act_model_layers_0_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_94_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_88_mlp_down_proj/max":18.125,"train/tensor_act_model_layers_19_residual_post_mlp/max":81,"train/tensor_act_model_layers_75_mlp_gate_proj/mean":0.60546875,"train/tensor_param_model_layers_29_mlp_down_proj_weight/norm":5.75,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_26_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_90_mlp_down_proj/min":14.0625,"train/tensor_act_model_layers_52_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_90_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_14_mlp_down_proj/norm":29934.05859375,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_46_mlp_down_proj_weight/std":0.034423828125,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/std":0.037353515625,"train/tensor_act_model_layers_69_residual_pre_attn/norm":565213.3125,"train/tensor_param_model_layers_38_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_param_model_layers_50_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_36_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_54_mlp_down_proj_weight/max_abs":0.1259765625,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_63_residual_post_attn/max":253,"train/tensor_act_model_layers_50_mlp_gate_proj/range":5.3828125,"train/tensor_param_model_layers_57_mlp_up_proj_weight/norm":6.6875,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_14_mlp_down_proj/max_abs":15.6875,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_1_mlp/max_abs":4,"train/tensor_act_model_layers_61_residual_pre_attn/norm":499484.78125,"train/tensor_act_model_layers_22_mlp/range":0,"train/tensor_act_model_layers_34_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_residual_pre_attn/max_abs":348,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_29_mlp_up_proj/mean":0.04833984375,"train/tensor_param_model_layers_15_mlp_up_proj_weight/range":0.2333984375,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_24_mlp_up_proj_weight/min":-0.11181640625,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/min":-0.1376953125,"train/tensor_act_model_layers_15_mlp_up_proj/max_abs":4.15625,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_79_mlp_down_proj/std":0.48046875,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_13_residual_pre_attn/range":2,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/range":0.2763671875,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_15_mlp_down_proj_weight/range":0.21337890625,"train/tensor_act_model_layers_28_residual_post_mlp/max":117,"train/tensor_act_model_layers_38_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_77_residual_post_attn/min":306,"train/tensor_act_model_layers_14_residual_post_attn/min":55,"train/tensor_act_model_layers_93_mlp_down_proj/norm":30910.78125,"train/tensor_act_model_layers_81_residual_post_mlp/max":328,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_37_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_mlp/max":4,"train/tensor_act_model_layers_14_residual_post_mlp/norm":122679.5234375,"train/tensor_act_model_layers_77_residual_post_mlp/norm":638941.25,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/norm":39.5,"train/tensor_act_model_layers_72_mlp/min":4,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_31_mlp_down_proj_weight/max_abs":0.1328125,"train/tensor_act_model_layers_28_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_mlp_down_proj/range":3.25,"train/tensor_act_model_layers_3_mlp/max_abs":4,"train/tensor_act_model_layers_89_residual_post_mlp/mean":360,"train/tensor_param_model_layers_2_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_7_residual_post_attn/max_abs":28.75,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_50_mlp_gate_proj/norm":5259.24365234375,"train/tensor_param_model_layers_73_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_89_residual_post_mlp/min":358,"train/tensor_param_model_layers_74_mlp_up_proj_weight/range":0.2548828125,"train/tensor_act_model_layers_80_mlp/std":0,"train/tensor_act_model_layers_2_residual_post_attn/std":0.12890625,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_97_mlp_gate_proj/range":6.859375,"train/tensor_param_model_layers_18_mlp_up_proj_weight/max_abs":0.10986328125,"train/tensor_act_model_layers_63_residual_post_attn/std":0.416015625,"train/tensor_param_model_layers_24_mlp_down_proj_weight/std":0.031494140625,"train/tensor_param_model_layers_98_mlp_down_proj_weight/mean":-0.000820159912109375,"train/tensor_param_model_layers_79_mlp_down_proj_weight/max_abs":0.1455078125,"train/tensor_act_model_layers_24_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_72_mlp_up_proj_weight/range":0.2626953125,"train/tensor_act_model_layers_32_mlp_up_proj/norm":5519.72998046875,"train/tensor_act_model_layers_86_residual_post_attn/norm":704477.25,"train/tensor_param_model_layers_81_mlp_down_proj_weight/std":0.043701171875,"train/tensor_param_model_layers_76_mlp_down_proj_weight/max_abs":0.1357421875,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_16_mlp_down_proj_weight/norm":5.5,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_27_residual_post_attn/max_abs":109,"train/tensor_param_model_layers_45_mlp_down_proj_weight/range":0.25146484375,"train/tensor_act_model_layers_47_residual_post_mlp/max":193,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/range":0.251953125,"train/tensor_act_model_layers_23_mlp/max_abs":4,"train/tensor_act_model_layers_25_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_97_mlp_up_proj_weight/mean":0.000759124755859375,"train/tensor_param_model_layers_48_mlp_up_proj_weight/max_abs":0.1259765625,"train/tensor_act_model_layers_17_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_20_mlp/std":0,"train/tensor_param_model_layers_15_mlp_down_proj_weight/max":0.09716796875,"train/tensor_act_model_layers_50_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_residual_post_attn/max_abs":340,"train/tensor_act_model_layers_84_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_15_mlp/mean":4,"train/tensor_act_model_layers_0_mlp_down_proj/mean":27.75,"train/tensor_act_model_layers_72_residual_post_mlp/max_abs":292,"train/tensor_param_model_layers_68_mlp_down_proj_weight/std":0.0380859375,"train/tensor_act_model_layers_31_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/max_abs":0.1376953125,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_8_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_21_mlp_up_proj_weight/max":0.11181640625,"train/tensor_act_model_layers_5_mlp_gate_proj/range":5.859375,"train/tensor_act_model_layers_29_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_48_mlp_up_proj/min":-3.703125,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/min":-0.146484375,"train/tensor_act_model_layers_4_residual_post_mlp/min":19.25,"train/tensor_act_model_layers_34_residual_pre_attn/max_abs":137,"train/tensor_act_model_layers_7_mlp/range":0,"train/tensor_param_model_layers_11_mlp_up_proj_weight/max":0.10791015625,"train/tensor_param_model_layers_42_mlp_up_proj_weight/min":-0.130859375,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_9_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_38_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_mlp/mean":4,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/std":0.035888671875,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_77_mlp_up_proj_weight/std":0.04052734375,"train/tensor_param_model_layers_69_mlp_up_proj_weight/max_abs":0.1279296875,"train/tensor_param_model_layers_34_mlp_down_proj_weight/mean":-0.000308990478515625,"train/tensor_act_model_layers_3_mlp_down_proj/mean":14.8125,"train/tensor_act_model_layers_92_residual_post_attn/norm":753629.25,"train/tensor_act_model_layers_0_residual_post_attn/max_abs":0.33203125,"train/tensor_param_model_layers_58_mlp_up_proj_weight/max":0.13671875,"train/tensor_param_model_layers_20_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_25_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_73_residual_pre_attn/max":292,"train/tensor_act_model_layers_14_mlp_down_proj/std":0.35546875,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_20_residual_post_mlp/std":0.359375,"train/tensor_param_model_layers_61_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/std":0.03759765625,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/std":0.044677734375,"train/tensor_act_model_layers_3_residual_post_attn/mean":12,"train/tensor_act_model_layers_57_mlp_up_proj/range":9.5,"train/tensor_act_model_layers_60_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_52_residual_pre_attn/mean":208,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_10_mlp_down_proj_weight/max":0.09423828125,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_71_mlp/range":0,"train/tensor_act_model_layers_86_mlp_down_proj/min":13.9375,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_46_residual_pre_attn/mean":184,"train/tensor_act_model_layers_70_residual_post_mlp/norm":581597.3125,"train/tensor_act_model_layers_36_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_10_mlp_gate_proj/norm":5139.6689453125,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_96_residual_post_attn/max_abs":384,"train/tensor_param_model_layers_65_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_79_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_5_residual_post_mlp/mean":24,"train/tensor_act_model_layers_65_mlp_down_proj/std":0.5234375,"train/tensor_act_model_layers_75_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_84_residual_post_attn/norm":688093.3125,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_89_mlp/max":4,"train/tensor_act_model_layers_13_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_18_mlp_down_proj/mean":14.5625,"train/tensor_act_model_layers_24_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_60_residual_post_attn/max_abs":241,"train/tensor_act_model_layers_49_mlp_up_proj/max":4.40625,"train/tensor_param_model_layers_95_mlp_down_proj_weight/norm":9.125,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_1_residual_post_mlp/max":8.375,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_mlp/min":4,"train/tensor_act_model_layers_36_residual_post_mlp/min":146,"train/tensor_act_model_layers_7_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_mlp_down_proj/max":16.75,"train/tensor_param_model_layers_91_mlp_down_proj_weight/min":-0.1416015625,"train/tensor_act_model_layers_1_residual_post_mlp/max_abs":8.375,"train/tensor_act_model_layers_78_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_24_mlp/min":4,"train/tensor_param_model_layers_23_mlp_up_proj_weight/norm":5.75,"train/tensor_act_model_layers_23_mlp_gate_proj/std":1.7421875,"train/tensor_param_model_layers_59_mlp_down_proj_weight/norm":6.65625,"train/tensor_param_model_layers_62_mlp_up_proj_weight/max":0.1396484375,"train/tensor_act_model_layers_36_mlp_down_proj/range":1.5625,"train/tensor_act_model_layers_98_mlp_down_proj/mean":15.0625,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_79_residual_pre_attn/max_abs":316,"train/tensor_act_model_layers_70_residual_post_attn/min":278,"train/tensor_act_model_layers_13_mlp_up_proj/std":1.921875,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_18_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_8_residual_pre_attn/min":31.125,"train/tensor_act_model_layers_6_residual_post_mlp/norm":57297.72265625,"train/tensor_act_model_layers_32_residual_post_mlp/max_abs":133,"train/tensor_act_model_layers_47_mlp/std":0,"train/tensor_act_model_layers_99_mlp_up_proj/norm":5506.41845703125,"train/tensor_act_model_layers_97_residual_pre_attn/range":2,"train/tensor_act_model_layers_72_residual_pre_attn/min":286,"train/tensor_act_model_layers_10_residual_pre_attn/std":0.2265625,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_82_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_92_mlp/mean":4,"train/tensor_act_model_layers_74_mlp_up_proj/norm":5305.05517578125,"train/tensor_act_model_layers_79_mlp_up_proj/norm":5218.3759765625,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_88_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_31_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp_up_proj/mean":0.1904296875,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/std":0.038818359375,"train/tensor_act_model_layers_33_residual_post_mlp/min":134,"train/tensor_act_model_layers_10_residual_pre_attn/min":39,"train/tensor_act_model_layers_98_mlp_up_proj/max":5.09375,"train/tensor_param_model_layers_88_mlp_up_proj_weight/mean":-0.00054168701171875,"train/tensor_act_model_layers_21_mlp_up_proj/mean":0.0625,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_94_mlp/max_abs":4,"train/tensor_param_model_layers_29_mlp_down_proj_weight/min":-0.11962890625,"train/tensor_act_model_layers_43_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_65_mlp/max":4,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/mean":0.00537109375,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_31_mlp_down_proj_weight/range":0.255859375,"train/tensor_act_model_layers_74_residual_post_attn/max_abs":296,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_72_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_54_mlp_up_proj_weight/mean":-0.00080108642578125,"train/tensor_act_model_layers_84_mlp_down_proj/max_abs":17.25,"train/tensor_act_model_layers_83_mlp_up_proj/norm":5295.78173828125,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_90_mlp_gate_proj/max_abs":4.75,"train/tensor_act_model_layers_99_residual_post_mlp/range":2,"train/tensor_act_model_layers_24_residual_post_mlp/mean":100,"train/tensor_act_model_layers_14_mlp_up_proj/mean":0.01129150390625,"train/tensor_param_model_layers_37_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_residual_pre_attn/max":340,"train/tensor_act_model_layers_95_residual_post_attn/max_abs":380,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_69_mlp_down_proj/mean":14.9375,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_50_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_mlp_up_proj/max":4.34375,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp_up_proj/max":4.53125,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/range":0.23779296875,"train/tensor_param_model_layers_5_mlp_up_proj_weight/mean":0.00075531005859375,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/norm":6,"train/tensor_param_model_layers_80_mlp_up_proj_weight/mean":-1.2636184692382812e-05,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/norm":9.9375,"train/tensor_param_model_layers_86_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_31_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_47_mlp_down_proj/std":0.404296875,"train/tensor_act_model_layers_5_mlp_gate_proj/std":1.6640625,"train/tensor_act_model_layers_6_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_28_mlp_down_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_49_mlp_up_proj_weight/range":0.2578125,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_8_residual_post_mlp/max_abs":36.75,"train/tensor_act_model_layers_88_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_41_mlp_up_proj/min":-4.4375,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_22_mlp_up_proj_weight/max":0.11767578125,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_53_mlp_gate_proj/range":6.328125,"train/tensor_act_model_layers_0_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_32_residual_post_attn/min":126.5,"train/tensor_param_model_layers_59_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_98_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_mlp_up_proj/max_abs":4.21875,"train/tensor_act_model_layers_92_residual_pre_attn/min":366,"train/tensor_act_model_layers_93_mlp_down_proj/mean":15.0625,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_58_mlp_up_proj/mean":0.0732421875,"train/tensor_act_model_layers_87_mlp_gate_proj/range":6.40625,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_33_mlp_gate_proj/mean":0.52734375,"train/tensor_act_model_layers_73_mlp_up_proj/std":1.765625,"train/tensor_act_model_layers_96_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_41_mlp_up_proj/max_abs":4.4375,"train/tensor_act_model_layers_81_mlp/range":0,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/std":0.038330078125,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_18_mlp/max_abs":4,"train/tensor_act_model_layers_89_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_92_residual_pre_attn/norm":753629.25,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_2_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_46_residual_post_attn/max":185,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/max":0.1416015625,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_96_mlp_up_proj_weight/mean":0.0015106201171875,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_3_residual_pre_attn/mean":12,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_19_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_32_residual_post_mlp/norm":270188.8125,"train/tensor_act_model_layers_11_mlp/norm":8192,"train/tensor_act_model_layers_92_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/max_abs":0.11669921875,"train/tensor_act_model_layers_74_mlp_down_proj/std":0.578125,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_92_mlp_gate_proj/range":6.8046875,"train/tensor_act_model_layers_21_mlp_up_proj/std":1.9375,"train/tensor_act_model_layers_42_residual_post_mlp/range":3,"train/tensor_act_model_layers_3_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_1_mlp_up_proj/std":1.8125,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/range":0.4189453125,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_34_mlp_up_proj_weight/norm":5.90625,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_28_residual_pre_attn/range":2.5,"train/tensor_param_model_layers_52_mlp_down_proj_weight/mean":-0.00015163421630859375,"train/tensor_param_model_layers_15_mlp_down_proj_weight/min":-0.1162109375,"train/tensor_param_model_layers_14_mlp_up_proj_weight/norm":5.5,"train/tensor_param_model_layers_30_mlp_up_proj_weight/mean":-0.00115203857421875,"train/tensor_param_model_layers_90_mlp_down_proj_weight/std":0.046875,"train/tensor_act_model_layers_56_mlp_down_proj/max_abs":16.375,"train/tensor_act_model_layers_52_residual_post_attn/max":209,"train/tensor_act_model_layers_80_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_27_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_22_residual_post_mlp/std":0.359375,"train/tensor_act_model_layers_39_mlp_gate_proj/mean":0.55078125,"train/tensor_param_model_layers_50_mlp_up_proj_weight/max_abs":0.1259765625,"train/tensor_param_model_layers_33_mlp_up_proj_weight/mean":-0.00164031982421875,"train/tensor_act_model_layers_23_residual_post_mlp/mean":96,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_80_residual_pre_attn/min":318,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_37_mlp/std":0,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_62_mlp_up_proj_weight/max_abs":0.1396484375,"train/tensor_act_model_layers_83_mlp_gate_proj/norm":5179.40771484375,"train/tensor_act_model_layers_34_mlp/norm":8192,"train/tensor_act_model_layers_79_mlp_up_proj/min":-4.4375,"train/tensor_act_model_layers_25_mlp_down_proj/std":0.3984375,"train/tensor_act_model_layers_16_mlp/range":0,"train/tensor_act_model_layers_74_residual_post_mlp/min":298,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_mlp_gate_proj/std":1.6875,"train/tensor_act_model_layers_64_mlp_gate_proj/range":5.625,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_98_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/norm":7.1875,"train/tensor_act_model_layers_16_residual_pre_attn/max":65,"train/tensor_param_model_layers_1_mlp_down_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_62_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_residual_pre_attn/range":3,"train/tensor_param_model_layers_94_mlp_up_proj_weight/norm":9.3125,"train/tensor_act_model_layers_43_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_98_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_67_mlp_down_proj/min":14.1875,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_29_mlp_up_proj_weight/max":0.11279296875,"train/tensor_act_model_layers_38_residual_post_mlp/max":157,"train/tensor_act_model_layers_95_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_18_mlp_up_proj/std":1.8828125,"train/tensor_act_model_layers_94_residual_pre_attn/max":376,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/min":-0.12353515625,"train/tensor_act_model_layers_67_residual_post_attn/min":266,"train/tensor_act_model_layers_46_residual_post_attn/std":0.31640625,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/min":-0.134765625,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_97_mlp_down_proj/max_abs":19.75,"train/tensor_act_model_layers_72_mlp_down_proj/norm":30626.21484375,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_83_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_50_residual_pre_attn/range":3,"train/tensor_act_model_layers_15_mlp_down_proj/norm":30054.994140625,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/min":-0.14453125,"train/tensor_act_model_layers_43_mlp_up_proj/mean":0.0267333984375,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_13_residual_pre_attn/norm":106329.78125,"train/tensor_act_model_layers_86_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_25_mlp_down_proj_weight/std":0.03125,"train/tensor_act_model_layers_67_residual_post_mlp/norm":557021.3125,"train/tensor_param_model_layers_91_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_16_mlp_up_proj_weight/max":0.10693359375,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_96_residual_pre_attn/max":384,"train/tensor_act_model_layers_18_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train_loss":2.2044668284824915,"train/tensor_act_model_layers_26_mlp/min":4,"train/tensor_param_model_layers_1_mlp_up_proj_weight/std":0.03564453125,"train/tensor_act_model_layers_71_mlp/max":4,"train/tensor_param_model_layers_89_mlp_up_proj_weight/std":0.04736328125,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_88_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/norm":9.1875,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_44_mlp_up_proj_weight/range":0.24462890625,"train/tensor_act_model_layers_26_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_mlp_gate_proj/max":5.0625,"train/tensor_act_model_layers_49_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_14_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_57_residual_post_attn/mean":228,"train/tensor_act_model_layers_84_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_6_mlp_down_proj_weight/norm":5.28125,"train/tensor_act_model_layers_77_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_32_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_31_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_12_mlp_up_proj/max":4.875,"train/tensor_act_model_layers_99_mlp_down_proj/std":1.4453125,"train/tensor_act_model_layers_19_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_2_residual_pre_attn/max_abs":8.375,"train/tensor_act_model_layers_76_residual_post_attn/norm":622557.3125,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/norm":7.34375,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_70_mlp_up_proj_weight/max":0.13671875,"train/tensor_act_model_layers_84_mlp_up_proj/norm":5196.33642578125,"train/tensor_act_model_layers_18_mlp/norm":8192,"train/tensor_act_model_layers_74_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/norm":5.5625,"train/tensor_act_model_layers_5_mlp/range":0,"train/tensor_act_model_layers_5_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_79_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_22_residual_post_attn/norm":179985.390625,"train/tensor_act_model_layers_52_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_64_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/mean":0.00555419921875,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_89_mlp_down_proj/std":0.58984375,"train/tensor_act_model_layers_63_residual_pre_attn/mean":252,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/max_abs":0.1728515625,"train/tensor_act_model_layers_50_mlp_up_proj/max":4.21875,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_54_mlp/max_abs":4,"train/tensor_param_model_layers_99_mlp_up_proj_weight/norm":10,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_82_mlp/min":4,"train/tensor_act_model_layers_25_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_85_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_69_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_78_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_12_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_83_mlp/norm":8192,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_63_mlp_up_proj_weight/min":-0.1279296875,"train/tensor_param_model_layers_12_mlp_up_proj_weight/min":-0.125,"train/tensor_param_model_layers_45_mlp_down_proj_weight/std":0.03466796875,"train/tensor_param_model_layers_29_mlp_down_proj_weight/max":0.125,"train/tensor_act_model_layers_26_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/frac_near_user_limit":0.98876953125,"train/tensor_act_model_layers_2_mlp/max_abs":4,"train/tensor_act_model_layers_90_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_34_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_43_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_7_mlp_down_proj_weight/range":0.19384765625,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/norm":7.28125,"train/tensor_act_model_layers_37_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_6_residual_post_mlp/mean":28,"train/tensor_act_model_layers_29_mlp/min":4,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_act_model_layers_58_mlp_gate_proj/range":6.46875,"train/tensor_act_model_layers_19_mlp/max":4,"train/tensor_param_model_layers_51_mlp_up_proj_weight/norm":6.5,"train/tensor_param_model_layers_7_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_94_residual_post_attn/max":376,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_25_residual_post_mlp/range":2.5,"train/tensor_param_model_layers_68_mlp_up_proj_weight/mean":-0.00193023681640625,"train/tensor_act_model_layers_33_residual_post_mlp/max_abs":137,"train/tensor_act_model_layers_22_residual_pre_attn/max":89,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/range":0.23388671875,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_9_mlp/max":4,"train/tensor_act_model_layers_56_residual_pre_attn/norm":458540.8125,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/max":0.1396484375,"train/tensor_act_model_layers_79_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/max":0.12890625,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_26_mlp_up_proj/max_abs":5.0625,"train/tensor_act_model_layers_1_mlp_gate_proj/range":6.5,"train/tensor_act_model_layers_87_mlp/range":0,"train/tensor_act_model_layers_35_mlp_down_proj/range":2.125,"train/tensor_act_model_layers_13_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_21_mlp_gate_proj/range":5.7578125,"train/tensor_act_model_layers_98_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_74_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/range":0.22705078125,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/mean":0.0037841796875,"train/tensor_act_model_layers_81_residual_post_mlp/max_abs":328,"train/tensor_param_model_layers_10_mlp_down_proj_weight/std":0.029541015625,"train/tensor_param_model_layers_96_mlp_down_proj_weight/min":-0.1591796875,"train/tensor_act_model_layers_46_mlp_up_proj/std":1.859375,"train/tensor_act_model_layers_0_residual_pre_attn/min":-0.30078125,"train/tensor_act_model_layers_58_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_36_mlp_up_proj/norm":5406.5390625,"train/tensor_act_model_layers_94_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_53_mlp_down_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_5_mlp_down_proj/mean":14.625,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_91_residual_post_mlp/max_abs":368,"train/tensor_act_model_layers_52_mlp_down_proj/mean":14.8125,"train/tensor_act_model_layers_35_residual_post_attn/max_abs":141,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_40_residual_post_attn/norm":327532.6875,"train/tensor_act_model_layers_43_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_18_mlp_down_proj_weight/max_abs":0.11767578125,"train/tensor_act_model_layers_89_residual_post_attn/range":2,"train/tensor_act_model_layers_91_residual_pre_attn/mean":364,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_6_mlp/range":0,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_13_mlp_down_proj_weight/min":-0.103515625,"train/tensor_act_model_layers_43_mlp/max_abs":4,"train/tensor_act_model_layers_34_mlp_down_proj/std":0.416015625,"train/tensor_act_model_layers_48_mlp_up_proj/max_abs":4.8125,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_97_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/max":0.126953125,"train/tensor_param_model_layers_19_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_64_mlp_down_proj/max":16.625,"train/tensor_act_model_layers_46_mlp/std":0,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_23_mlp/max":4,"train/tensor_param_model_layers_91_mlp_down_proj_weight/mean":0.00010538101196289062,"train/tensor_act_model_layers_57_mlp_gate_proj/mean":0.55078125,"train/tensor_act_model_layers_89_mlp/range":0,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_25_mlp_down_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_16_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_97_mlp_up_proj_weight/max_abs":0.1953125,"train/tensor_param_model_layers_31_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_mlp_down_proj/max_abs":16.375,"train/tensor_act_model_layers_53_residual_post_mlp/norm":442156.84375,"train/tensor_param_model_layers_18_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_27_mlp_up_proj_weight/mean":0.0004730224609375,"train/tensor_act_model_layers_9_mlp_gate_proj/mean":0.40625,"train/tensor_param_model_layers_57_mlp_up_proj_weight/max_abs":0.1328125,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/norm":5.375,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_26_residual_post_attn/norm":212753.15625,"train/tensor_act_model_layers_81_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_94_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_28_mlp_down_proj/std":0.40625,"train/tensor_param_model_layers_99_mlp_up_proj_weight/max_abs":0.169921875,"train/tensor_act_model_layers_79_residual_pre_attn/mean":316,"train/tensor_act_model_layers_24_mlp_gate_proj/max_abs":4.46875,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/range":0.2685546875,"train/tensor_act_model_layers_64_mlp_gate_proj/norm":5038.33203125,"train/tensor_param_model_layers_84_mlp_up_proj_weight/norm":8.125,"train/tensor_act_model_layers_55_mlp_up_proj/max":4.5,"train/tensor_act_model_layers_78_residual_pre_attn/mean":312,"train/tensor_act_model_layers_84_mlp_down_proj/norm":31083.40234375,"train/tensor_act_model_layers_53_residual_post_attn/min":210,"train/tensor_act_model_layers_64_mlp/std":0,"train/tensor_act_model_layers_73_residual_post_attn/norm":597981.3125,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_1_mlp_up_proj/max":4.3125,"train/tensor_param_model_layers_12_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_35_residual_post_attn/mean":140,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_73_mlp/range":0,"train/tensor_act_model_layers_64_residual_post_mlp/range":2,"train/tensor_act_model_layers_17_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_89_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_mlp_gate_proj/mean":0.50390625,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/norm":6.0625,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/max":0.1328125,"train/tensor_act_model_layers_63_mlp/mean":4,"train/tensor_param_model_layers_16_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_85_residual_post_mlp/range":2,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_96_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_45_mlp_down_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_47_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_84_residual_pre_attn/norm":688093.3125,"train/tensor_act_model_layers_35_mlp_up_proj/min":-4.9375,"train/tensor_act_model_layers_21_mlp_down_proj/max":16.5,"train/tensor_act_model_layers_72_residual_pre_attn/range":2,"train/tensor_act_model_layers_16_residual_post_mlp/std":0.34375,"train/tensor_act_model_layers_5_residual_post_mlp/norm":49117.7578125,"train/tensor_act_model_layers_20_residual_post_attn/max_abs":81,"train/tensor_act_model_layers_38_mlp_down_proj/norm":30227.251953125,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_20_residual_pre_attn/norm":163610.28125,"train/tensor_act_model_layers_98_residual_pre_attn/max":392,"train/tensor_act_model_layers_15_mlp_up_proj/range":8.109375,"train/tensor_param_model_layers_5_mlp_up_proj_weight/range":0.23046875,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_91_mlp_down_proj_weight/range":0.2841796875,"train/tensor_act_model_layers_93_mlp_gate_proj/max_abs":5.46875,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_75_mlp_up_proj_weight/norm":7.25,"train/tensor_act_model_layers_62_residual_post_mlp/max":253,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_94_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_10_mlp_up_proj/norm":5630.46533203125,"train/tensor_act_model_layers_97_mlp_down_proj/norm":31035.3125,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/min":-0.1435546875,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_20_mlp_up_proj/std":1.890625,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/mean":0.0033416748046875,"train/tensor_act_model_layers_57_residual_post_mlp/norm":474908.8125,"train/tensor_act_model_layers_60_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_57_mlp/max":4,"train/tensor_act_model_layers_7_residual_pre_attn/min":27.125,"train/tensor_act_model_layers_91_residual_post_attn/max":364,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_35_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_90_mlp_down_proj_weight/max":0.154296875,"train/tensor_param_model_layers_73_mlp_up_proj_weight/std":0.03857421875,"train/tensor_act_model_layers_10_mlp/min":4,"train/tensor_param_model_layers_24_mlp_down_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_90_mlp_up_proj_weight/max":0.158203125,"train/learning_rate":9e-05,"train/tensor_param_model_layers_9_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_98_mlp_down_proj/std":0.96484375,"train/tensor_param_model_layers_79_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_74_mlp_up_proj_weight/max":0.1279296875,"train/tensor_act_model_layers_23_mlp_gate_proj/min":-1.2890625,"train/tensor_act_model_layers_15_mlp_gate_proj/max":4.34375,"train/tensor_param_model_layers_56_mlp_down_proj_weight/norm":6.625,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_41_mlp_up_proj_weight/min":-0.1220703125,"train/tensor_param_model_layers_60_mlp_up_proj_weight/range":0.2578125,"train/tensor_act_model_layers_90_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_81_mlp_down_proj_weight/min":-0.1337890625,"train/tensor_act_model_layers_46_mlp_down_proj/min":14.1875,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_19_mlp_down_proj/max_abs":16.375,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_98_mlp_up_proj_weight/std":0.0546875,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_81_mlp_up_proj_weight/min":-0.1416015625,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_78_mlp_up_proj/max_abs":4.53125,"train/tensor_act_model_layers_80_residual_post_attn/norm":655325.25,"train/tensor_act_model_layers_36_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_88_residual_post_mlp/range":2,"train/tensor_act_model_layers_23_residual_post_attn/max_abs":93,"train/tensor_act_model_layers_88_residual_post_mlp/max":356,"train/tensor_act_model_layers_12_residual_post_mlp/max":53,"train/tensor_act_model_layers_57_residual_post_attn/min":226,"train/tensor_act_model_layers_36_residual_post_attn/range":3,"train/tensor_param_model_layers_6_mlp_down_proj_weight/max":0.10693359375,"train/tensor_param_model_layers_68_mlp_down_proj_weight/min":-0.130859375,"train/tensor_act_model_layers_79_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_21_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/std":0.03759765625,"train/tensor_act_model_layers_46_residual_post_mlp/min":186,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_73_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_98_residual_post_mlp/range":2,"train/tensor_act_model_layers_9_residual_post_mlp/std":0.2265625,"train/tensor_act_model_layers_63_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_39_mlp_down_proj/max_abs":16,"train/tensor_act_model_layers_11_residual_post_mlp/max_abs":49,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_20_residual_pre_attn/max":81,"train/tensor_act_model_layers_60_mlp_up_proj/max":3.96875,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/range":0.24267578125,"train/tensor_act_model_layers_68_residual_post_attn/range":2,"train/tensor_act_model_layers_94_residual_pre_attn/mean":376,"train/tensor_act_model_layers_94_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_16_mlp_gate_proj/norm":5107.9638671875,"train/tensor_act_model_layers_40_residual_pre_attn/min":158,"train/tensor_act_model_layers_81_mlp_down_proj/norm":30958.36328125,"train/tensor_act_model_layers_58_mlp/mean":4,"train/tensor_param_model_layers_69_mlp_down_proj_weight/max_abs":0.1416015625,"train/tensor_act_model_layers_33_residual_post_attn/mean":132,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/mean":0.004486083984375,"train/tensor_act_model_layers_46_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_86_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/norm":5.4375,"train/tensor_act_model_layers_58_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_40_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_79_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_25_mlp_up_proj_weight/std":0.03173828125,"train/tensor_act_model_layers_67_mlp_gate_proj/norm":5099.82861328125,"train/tensor_act_model_layers_79_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_88_mlp_gate_proj/max_abs":4.59375,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/max_abs":0.1357421875,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/max":0.1376953125,"train/tensor_act_model_layers_34_residual_post_attn/max":137,"train/tensor_param_model_layers_95_mlp_down_proj_weight/max_abs":0.189453125,"train/tensor_act_model_layers_43_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_85_residual_post_mlp/mean":344,"train/tensor_act_model_layers_97_residual_post_attn/range":2,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_34_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_78_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_67_residual_pre_attn/min":266,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_11_mlp_up_proj_weight/max_abs":0.1220703125,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_21_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_30_mlp_down_proj_weight/mean":-3.504753112792969e-05,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_17_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_3_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_40_mlp_down_proj_weight/range":0.24853515625,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/min":-0.1357421875,"train/tensor_act_model_layers_80_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_75_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_46_mlp_down_proj_weight/min":-0.125,"train/tensor_param_model_layers_99_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_91_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_58_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/mean":0.00537109375,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/max_abs":0.2197265625,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_34_mlp_gate_proj/mean":0.5625,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_66_residual_post_mlp/mean":268,"train/tensor_act_model_layers_95_mlp_up_proj/min":-4.8125,"train/tensor_param_model_layers_87_mlp_up_proj_weight/norm":8.3125,"train/tensor_act_model_layers_85_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_31_mlp_gate_proj/range":5.890625,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_64_residual_pre_attn/norm":524001.8125,"train/tensor_act_model_layers_99_mlp_down_proj/mean":15.375,"train/tensor_act_model_layers_67_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_25_mlp_down_proj_weight/max":0.125,"train/tensor_act_model_layers_98_residual_pre_attn/max_abs":392,"train/tensor_act_model_layers_7_mlp/min":4,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_29_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_79_residual_post_attn/max":316,"train/tensor_act_model_layers_89_mlp_down_proj/mean":15.125,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_68_mlp_down_proj/min":14.1875,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_mlp_down_proj/std":0.7265625,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_34_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_31_mlp_gate_proj/max_abs":4.59375,"train/tensor_param_model_layers_54_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_21_mlp_down_proj_weight/norm":5.59375,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_18_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_90_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_78_residual_post_attn/max":312,"train/tensor_act_model_layers_31_residual_pre_attn/min":122.5,"train/tensor_param_model_layers_73_mlp_down_proj_weight/max_abs":0.1357421875,"train/tensor_act_model_layers_71_mlp_down_proj/range":2.8125,"train/tensor_act_model_layers_73_mlp/mean":4,"train/tensor_act_model_layers_4_mlp/mean":4,"train/tensor_act_model_layers_68_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/min":-0.12109375,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_62_residual_post_attn/norm":507676.75,"train/tensor_act_model_layers_31_residual_pre_attn/std":0.388671875,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/max_abs":2000,"train/tensor_act_model_layers_32_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_55_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_54_mlp_down_proj/max":16.75,"train/tensor_act_model_layers_69_mlp_up_proj/mean":-0.08837890625,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/mean":0.004730224609375,"train/tensor_act_model_layers_8_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_66_mlp_up_proj_weight/range":0.2763671875,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_92_mlp_up_proj_weight/mean":-0.00010204315185546875,"train/tensor_act_model_layers_40_mlp/norm":8192,"train/tensor_act_model_layers_37_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_15_mlp_down_proj_weight/mean":0.00011491775512695312,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_83_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_77_mlp_up_proj_weight/norm":7.375,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/std":0.0308837890625,"train/tensor_act_model_layers_56_mlp/min":4,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_81_residual_post_attn/norm":663517.3125,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_81_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_29_residual_pre_attn/mean":116,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_residual_pre_attn/std":0.31640625,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_43_residual_pre_attn/min":170,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/max_abs":0.1318359375,"train/tensor_act_model_layers_50_residual_post_attn/range":3,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_20_mlp_gate_proj/mean":0.61328125,"train/tensor_param_model_layers_28_mlp_down_proj_weight/std":0.031982421875,"train/tensor_act_model_layers_11_mlp_up_proj/min":-3.734375,"train/tensor_param_model_layers_92_mlp_up_proj_weight/max_abs":0.171875,"train/tensor_act_model_layers_34_residual_pre_attn/range":3,"train/tensor_act_model_layers_52_residual_pre_attn/max":209,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_mlp/max_abs":4,"train/tensor_act_model_layers_84_mlp_down_proj/mean":15.1875,"train/tensor_act_model_layers_68_residual_pre_attn/min":270,"train/tensor_act_model_layers_10_residual_pre_attn/range":1.75,"train/tensor_param_model_layers_22_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_residual_post_mlp/max_abs":201,"train/tensor_act_model_layers_26_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_30_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_2_mlp_down_proj_weight/norm":5.53125,"train/tensor_act_model_layers_39_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_mlp_down_proj/std":0.4609375,"train/tensor_act_model_layers_82_mlp_gate_proj/norm":5007.25244140625,"train/tensor_act_model_layers_19_residual_post_mlp/max_abs":81,"train/tensor_act_model_layers_72_mlp_gate_proj/max":4.875,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_65_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/max":0.208984375,"train/tensor_act_model_layers_76_residual_pre_attn/mean":304,"train/tensor_act_model_layers_39_mlp_down_proj/max":16,"train/tensor_act_model_layers_18_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_70_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_post_mlp/max_abs":280,"train/tensor_act_model_layers_70_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_62_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_24_mlp_up_proj/mean":0.05859375,"train/tensor_act_model_layers_30_mlp_up_proj/norm":5544.7373046875,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_act_model_layers_17_residual_post_mlp/std":0.34375,"train/tensor_act_model_layers_19_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/norm":7.5625,"train/tensor_act_model_layers_96_mlp_gate_proj/max":5.53125,"train/tensor_param_model_layers_93_mlp_up_proj_weight/range":0.3359375,"train/tensor_act_model_layers_24_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_mlp_down_proj/mean":14.75,"train/tensor_act_model_layers_27_residual_pre_attn/max":109,"train/tensor_act_model_layers_95_mlp_gate_proj/max":6.15625,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_68_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_45_mlp_down_proj_weight/max":0.12353515625,"train/tensor_param_model_layers_29_mlp_down_proj_weight/std":0.03173828125,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_47_mlp_gate_proj/mean":0.62890625,"train/tensor_act_model_layers_21_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_67_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_mlp_down_proj/mean":15.125,"train/tensor_param_model_layers_77_mlp_up_proj_weight/range":0.2880859375,"train/tensor_act_model_layers_31_mlp/min":4,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_48_mlp_down_proj_weight/norm":6.375,"train/tensor_param_model_layers_80_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_19_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_46_mlp_down_proj_weight/mean":0.0009307861328125,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/max":0.125,"train/tensor_param_model_layers_80_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/range":0.251953125,"train/tensor_param_model_layers_91_mlp_up_proj_weight/std":0.048583984375,"train/tensor_act_model_layers_90_mlp_up_proj/max_abs":4.90625,"train/tensor_act_model_layers_46_mlp_gate_proj/range":5.8671875,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_45_mlp_up_proj/max_abs":4.75,"train/tensor_act_model_layers_96_residual_post_mlp/norm":794589.25,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_97_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_68_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_82_mlp_up_proj_weight/max":0.138671875,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_83_mlp_down_proj_weight/norm":8.0625,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_73_mlp_down_proj/max_abs":17.5,"train/tensor_act_model_layers_5_residual_post_mlp/max_abs":24.625,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_25_residual_post_mlp/mean":104,"train/tensor_act_model_layers_27_mlp/std":0,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_86_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_16_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_6_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_36_mlp_gate_proj/max":4.8125,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/mean":0.0050048828125,"train/tensor_act_model_layers_60_mlp/range":0,"train/tensor_param_model_layers_62_mlp_up_proj_weight/range":0.275390625,"train/tensor_param_model_layers_91_mlp_down_proj_weight/std":0.047607421875,"train/tensor_act_model_layers_2_mlp/range":0,"train/tensor_act_model_layers_96_residual_post_attn/min":382,"train/tensor_act_model_layers_78_mlp/norm":8192,"train/tensor_param_model_layers_64_mlp_up_proj_weight/max_abs":0.1318359375,"train/tensor_act_model_layers_33_mlp_down_proj/range":2.375,"train/tensor_act_model_layers_98_residual_pre_attn/norm":802781.25,"train/tensor_act_model_layers_70_mlp/std":0,"train/tensor_act_model_layers_40_mlp_up_proj/max":4.8125,"train/tensor_act_model_layers_30_mlp/range":0,"train/tensor_act_model_layers_30_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_15_mlp/max_abs":4,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/max_abs":0.13671875,"train/tensor_act_model_layers_27_residual_post_attn/norm":220945.109375,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_74_residual_pre_attn/mean":296,"train/tensor_act_model_layers_62_residual_pre_attn/norm":507676.75,"train/tensor_act_model_layers_58_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_84_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_14_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_residual_post_mlp/max":308,"train/tensor_act_model_layers_28_mlp_gate_proj/range":5.7265625,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_7_mlp_up_proj_weight/max":0.09912109375,"train/tensor_act_model_layers_70_mlp_gate_proj/norm":5054.52197265625,"train/tensor_act_model_layers_1_residual_post_attn/frac_near_user_limit":0,"train/tensor_act_model_layers_97_mlp_down_proj/min":13.875,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/norm":8.625,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_30_mlp/min":4,"train/tensor_act_model_layers_93_residual_post_mlp/min":374,"train/tensor_param_model_layers_29_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/std":0.045654296875,"train/tensor_param_model_layers_62_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_55_mlp_up_proj_weight/mean":-0.00023651123046875,"train/tensor_act_model_layers_41_residual_post_attn/max_abs":165,"train/tensor_act_model_layers_39_residual_post_attn/norm":319340.71875,"train/tensor_param_model_layers_37_mlp_up_proj_weight/max":0.12890625,"train/tensor_act_model_layers_18_residual_post_attn/max_abs":73,"train/tensor_param_model_layers_85_mlp_down_proj_weight/mean":0.00148773193359375,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_52_mlp_down_proj/norm":30363.6640625,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_mlp_down_proj/max_abs":19.5,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_20_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_31_mlp_up_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_46_mlp/max":4,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_59_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/min":-0.1416015625,"train/tensor_param_model_layers_96_mlp_down_proj_weight/max":0.17578125,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_10_mlp_up_proj/min":-4.0625,"train/tensor_act_model_layers_56_mlp_gate_proj/mean":0.49609375,"train/tensor_param_model_layers_13_mlp_down_proj_weight/max":0.1171875,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_47_residual_post_mlp/norm":393068.59375,"train/tensor_act_model_layers_67_mlp/min":4,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_62_mlp_down_proj/range":2.6875,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_74_mlp_up_proj_weight/max_abs":0.1279296875,"train/tensor_act_model_layers_1_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_50_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_36_mlp/max_abs":4,"train/tensor_act_model_layers_63_mlp/range":0,"train/tensor_act_model_layers_5_residual_pre_attn/std":0.16796875,"train/tensor_act_model_layers_74_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_36_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_residual_pre_attn/min":218,"train/tensor_act_model_layers_58_mlp_gate_proj/mean":0.44921875,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_9_mlp_up_proj_weight/std":0.0299072265625,"train/tensor_act_model_layers_55_residual_pre_attn/std":0.408203125,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/range":0.265625,"train/tensor_act_model_layers_31_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_99_mlp_down_proj/norm":31601.3046875,"train/tensor_param_model_layers_3_mlp_up_proj_weight/range":0.2099609375,"train/tensor_param_model_layers_46_mlp_down_proj_weight/max":0.1279296875,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/max":0.12890625,"train/tensor_act_model_layers_76_residual_post_attn/max":304,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/mean":0.004486083984375,"train/tensor_param_model_layers_22_mlp_up_proj_weight/range":0.2265625,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_47_residual_post_attn/min":186,"train/tensor_act_model_layers_72_mlp_down_proj/range":2.5,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_93_residual_post_attn/norm":761821.25,"train/tensor_act_model_layers_89_mlp_down_proj/norm":30959.48828125,"train/tensor_act_model_layers_86_residual_post_mlp/max_abs":348,"train/tensor_act_model_layers_36_mlp_up_proj/std":1.859375,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_54_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/max":0.1162109375,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/norm":6.96875,"train/tensor_param_model_layers_49_mlp_up_proj_weight/min":-0.125,"train/tensor_act_model_layers_75_mlp_down_proj/range":1.8125,"train/tensor_act_model_layers_17_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_13_residual_post_mlp/min":55,"train/tensor_act_model_layers_35_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/std":0.033935546875,"train/tensor_act_model_layers_27_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_post_mlp/norm":253689.125,"train/tensor_act_model_layers_74_mlp_up_proj/min":-4.71875,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_40_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/std":0.039794921875,"train/tensor_act_model_layers_70_mlp_up_proj/max":4.375,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_90_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_54_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_post_attn/min":118.5,"train/tensor_act_model_layers_6_residual_post_attn/max":24.625,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/max_abs":0.12255859375,"train/tensor_act_model_layers_70_residual_pre_attn/max_abs":280,"train/tensor_act_model_layers_29_residual_post_attn/max":117,"train/tensor_param_model_layers_60_mlp_down_proj_weight/norm":6.6875,"train/tensor_act_model_layers_78_residual_post_attn/max_abs":312,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_10_mlp_gate_proj/range":5.453125,"train/tensor_act_model_layers_93_mlp_down_proj/std":0.7109375,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/std":0.043701171875,"train/tensor_param_model_layers_25_mlp_down_proj_weight/min":-0.1162109375,"train/tensor_act_model_layers_48_residual_post_mlp/norm":401260.59375,"train/tensor_act_model_layers_36_mlp_up_proj/range":8.875,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_78_residual_pre_attn/max":312,"train/tensor_act_model_layers_70_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_47_mlp_down_proj_weight/norm":6.21875,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_45_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_32_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_70_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_61_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_95_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_98_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_87_mlp/max":4,"train/tensor_param_model_layers_26_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_35_residual_post_mlp/min":142,"train/tensor_act_model_layers_2_mlp_up_proj/norm":5767.90673828125,"train/tensor_act_model_layers_36_residual_post_attn/norm":294764.78125,"train/tensor_act_model_layers_93_mlp_up_proj/min":-5.40625,"train/tensor_act_model_layers_99_mlp_down_proj/range":8,"train/tensor_act_model_layers_48_residual_post_attn/min":190,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_mlp_up_proj/min":-4.03125,"train/tensor_act_model_layers_85_mlp_gate_proj/range":6.0390625,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/mean":0.004791259765625,"train/tensor_act_model_layers_84_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_mlp/mean":4,"train/tensor_act_model_layers_95_mlp_up_proj/range":10.8125,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_68_mlp_down_proj_weight/max":0.1328125,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_15_mlp_down_proj_weight/std":0.0301513671875,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/mean":0.0035400390625,"train/tensor_act_model_layers_96_mlp_up_proj/range":10.71875,"train/tensor_act_model_layers_75_mlp_down_proj/max_abs":16,"train/tensor_param_model_layers_40_mlp_up_proj_weight/std":0.03369140625,"train/tensor_act_model_layers_76_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_16_residual_post_mlp/range":2,"train/tensor_param_model_layers_24_mlp_down_proj_weight/max":0.11962890625,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_35_mlp_up_proj_weight/max":0.1220703125,"train/tensor_param_model_layers_90_mlp_down_proj_weight/min":-0.1435546875,"train/tensor_param_model_layers_2_mlp_up_proj_weight/max":0.11328125,"train/tensor_act_model_layers_35_residual_post_mlp/mean":144,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_74_residual_post_attn/max":296,"train/tensor_act_model_layers_3_mlp/min":4,"train/tensor_param_model_layers_20_mlp_down_proj_weight/mean":-6.437301635742188e-05,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/norm":5.65625,"train/tensor_act_model_layers_72_mlp_up_proj/range":9.875,"train/tensor_act_model_layers_1_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_40_mlp_gate_proj/range":6.109375,"train/tensor_act_model_layers_1_mlp_up_proj/min":-4.625,"train/tensor_act_model_layers_67_residual_pre_attn/range":2,"train/tensor_act_model_layers_74_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_81_mlp_down_proj/std":0.6171875,"train/tensor_param_model_layers_82_mlp_down_proj_weight/norm":8,"train/tensor_act_model_layers_4_mlp_down_proj/mean":14.625,"train/tensor_param_model_layers_47_mlp_up_proj_weight/norm":6.28125,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_12_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_85_mlp_up_proj_weight/min":-0.1474609375,"train/tensor_param_model_layers_81_mlp_down_proj_weight/max":0.1474609375,"train/tensor_act_model_layers_58_residual_pre_attn/range":3,"train/tensor_param_model_layers_78_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_67_residual_post_attn/max_abs":268,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_29_residual_post_attn/std":0.388671875,"train/tensor_act_model_layers_32_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_95_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_80_residual_pre_attn/norm":655325.25,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_17_mlp_up_proj/max":4.5625,"train/tensor_param_model_layers_17_mlp_down_proj_weight/range":0.21875,"train/tensor_param_model_layers_93_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_mlp_up_proj/std":1.828125,"train/tensor_act_model_layers_29_residual_pre_attn/std":0.388671875,"train/tensor_act_model_layers_95_mlp_down_proj/range":4.4375,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_24_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_79_mlp/norm":8192,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_9_mlp_down_proj_weight/std":0.029541015625,"train/tensor_act_model_layers_11_residual_post_mlp/std":0.26171875,"train/tensor_act_model_layers_37_mlp/min":4,"train/tensor_act_model_layers_42_mlp_gate_proj/norm":5100.998046875,"train/tensor_param_model_layers_11_mlp_down_proj_weight/max_abs":0.111328125,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/min":-0.1337890625,"train/tensor_act_model_layers_54_mlp/norm":8192,"train/tensor_act_model_layers_60_residual_post_mlp/std":0.416015625,"train/tensor_act_model_layers_91_residual_post_attn/mean":364,"train/tensor_act_model_layers_59_residual_pre_attn/min":234,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/range":0.3486328125,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_mlp/max_abs":4,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_60_mlp_gate_proj/std":1.65625,"train/tensor_param_model_layers_77_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/mean":0.004913330078125,"train/tensor_act_model_layers_92_mlp_down_proj/min":14.0625,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_7_mlp_down_proj_weight/std":0.0294189453125,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_59_residual_pre_attn/std":0.416015625,"train/tensor_act_model_layers_81_mlp_down_proj/mean":15.125,"train/tensor_param_model_layers_74_mlp_up_proj_weight/frac_near_user_limit":0,"eval/runtime":21.0785,"train/tensor_act_model_layers_72_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_82_mlp_up_proj_weight/mean":5.221366882324219e-05,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_63_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_29_residual_pre_attn/min":114.5,"train/tensor_act_model_layers_62_residual_pre_attn/max_abs":249,"train/tensor_param_model_layers_79_mlp_up_proj_weight/max":0.1455078125,"train/tensor_param_model_layers_15_mlp_up_proj_weight/min":-0.1259765625,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/norm":7.46875,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_48_residual_pre_attn/max_abs":193,"train/tensor_act_model_layers_31_residual_pre_attn/norm":253689.125,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_31_mlp_up_proj_weight/norm":5.84375,"train/tensor_act_model_layers_56_mlp_gate_proj/min":-1.2734375,"train/tensor_act_model_layers_34_residual_pre_attn/mean":136,"train/tensor_act_model_layers_48_mlp_down_proj/mean":14.875,"train/tensor_act_model_layers_94_mlp_down_proj/std":0.79296875,"train/tensor_act_model_layers_61_mlp_up_proj/max_abs":4.625,"train/tensor_act_model_layers_28_mlp_up_proj/max":4.8125,"train/tensor_param_model_layers_28_mlp_up_proj_weight/max":0.11376953125,"train/tensor_act_model_layers_27_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/max":0.1298828125,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_13_mlp_up_proj/norm":5567.89013671875,"train/tensor_act_model_layers_38_mlp_down_proj/max_abs":16,"train/tensor_act_model_layers_89_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_95_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_77_mlp_down_proj/std":0.703125,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_89_residual_post_mlp/max_abs":360,"train/tensor_act_model_layers_90_mlp_up_proj/min":-4.90625,"train/tensor_act_model_layers_93_residual_pre_attn/max":372,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/max_abs":0.12890625,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_79_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_93_residual_post_mlp/max_abs":376,"_wandb":{"runtime":389},"train/tensor_param_model_layers_29_mlp_down_proj_weight/range":0.24462890625,"train/tensor_param_model_layers_12_mlp_up_proj_weight/range":0.2470703125,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/std":0.0380859375,"train/tensor_act_model_layers_91_mlp/std":0,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_2_mlp_up_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_67_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_2_mlp_gate_proj/norm":5188.33837890625,"train/tensor_param_model_layers_7_mlp_down_proj_weight/min":-0.0986328125,"train/tensor_act_model_layers_96_mlp_up_proj/norm":5225.73095703125,"train/tensor_act_model_layers_57_mlp/norm":8192,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_55_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_28_mlp_down_proj_weight/range":0.24462890625,"train/tensor_act_model_layers_81_residual_post_attn/range":2,"train/tensor_act_model_layers_24_mlp_gate_proj/std":1.7109375,"train/tensor_act_model_layers_74_mlp_down_proj/range":2.8125,"train/tensor_act_model_layers_45_mlp_gate_proj/max_abs":4.84375,"train/tensor_act_model_layers_2_mlp/min":4,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_24_residual_post_attn/max":97,"train/tensor_param_model_layers_9_mlp_down_proj_weight/min":-0.11865234375,"train/tensor_act_model_layers_17_mlp_up_proj/mean":-0.044921875,"train/tensor_act_model_layers_94_residual_pre_attn/max_abs":376,"train/tensor_act_model_layers_46_mlp_down_proj/max":16.5,"train/tensor_act_model_layers_53_mlp/std":0,"train/tensor_param_model_layers_86_mlp_down_proj_weight/max_abs":0.1455078125,"train/tensor_param_model_layers_15_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_50_mlp_down_proj_weight/range":0.2568359375,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_62_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_84_mlp_up_proj/min":-4.75,"train/tensor_act_model_layers_52_mlp/mean":4,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_6_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_85_mlp_down_proj_weight/min":-0.146484375,"train/tensor_act_model_layers_80_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_88_mlp_gate_proj/norm":5079.67138671875,"train/tensor_param_model_layers_75_mlp_down_proj_weight/std":0.03955078125,"train/tensor_act_model_layers_41_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_mlp_up_proj/std":1.90625,"train/tensor_act_model_layers_24_mlp_up_proj/std":1.921875,"train/tensor_act_model_layers_68_mlp_down_proj/max":17.125,"train/tensor_act_model_layers_27_residual_post_mlp/min":110.5,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_45_mlp_down_proj/std":0.466796875,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_90_residual_post_attn/norm":737245.25,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_42_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_16_mlp/norm":8192,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_11_mlp_up_proj/mean":-0.0546875,"train/tensor_act_model_layers_16_mlp_down_proj/range":2.375,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_90_mlp_down_proj/max_abs":17.875,"train/tensor_act_model_layers_23_mlp_gate_proj/mean":0.427734375,"train/tensor_act_model_layers_76_mlp_gate_proj/range":6.765625,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/min":-0.1787109375,"train/tensor_act_model_layers_42_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_66_mlp_down_proj/norm":30653.9453125,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_33_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/min":-0.134765625,"train/tensor_param_model_layers_19_mlp_up_proj_weight/norm":5.625,"train/tensor_act_model_layers_47_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/mean":0.004058837890625,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/norm":5.46875,"train/tensor_param_model_layers_66_mlp_down_proj_weight/max_abs":0.12890625,"train/tensor_act_model_layers_23_residual_pre_attn/mean":92,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_54_mlp_down_proj/min":14.25,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_17_mlp_down_proj_weight/mean":-0.00052642822265625,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_96_residual_pre_attn/max_abs":384,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/max":0.10888671875,"train/tensor_param_model_layers_35_mlp_up_proj_weight/mean":0.000461578369140625,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_58_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_54_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_24_residual_post_attn/std":0.359375,"train/tensor_act_model_layers_22_residual_post_attn/range":2.5,"train/tensor_act_model_layers_62_mlp_down_proj/std":0.5,"train/tensor_act_model_layers_78_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_80_mlp_down_proj/max":17.25,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_47_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_50_mlp_down_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_52_residual_post_mlp/min":210,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_3_mlp_gate_proj/range":5.6484375,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_50_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_mlp/mean":4,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_31_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_mlp/norm":8192,"train/tensor_act_model_layers_62_mlp/mean":4,"train/tensor_param_model_layers_83_mlp_up_proj_weight/norm":8.125,"train/tensor_act_model_layers_0_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_mlp_up_proj/std":1.828125,"train/tensor_act_model_layers_87_mlp/max_abs":4,"train/tensor_act_model_layers_13_residual_post_attn/mean":52,"train/tensor_param_model_layers_80_mlp_up_proj_weight/range":0.28515625,"train/tensor_act_model_layers_71_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_87_mlp_down_proj_weight/max":0.1376953125,"train/tensor_act_model_layers_87_mlp_gate_proj/norm":4893.646484375,"train/tensor_act_model_layers_30_mlp_down_proj/max_abs":16.25,"train/tensor_act_model_layers_21_mlp_down_proj/norm":30108.583984375,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_24_residual_pre_attn/mean":96,"train/tensor_act_model_layers_60_residual_post_attn/mean":240,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_25_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/range":0.23291015625,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_56_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_72_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_67_mlp_down_proj_weight/max_abs":0.130859375,"train/tensor_act_model_layers_72_mlp_up_proj/max":5.125,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_89_mlp_down_proj_weight/max_abs":0.1455078125,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_64_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_72_mlp_up_proj/norm":5102.84619140625,"train/tensor_act_model_layers_2_residual_pre_attn/frac_near_user_limit":0,"train/tensor_param_model_layers_3_mlp_up_proj_weight/max_abs":0.10693359375,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_20_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_71_mlp_up_proj_weight/range":0.26953125,"train/tensor_param_model_layers_59_mlp_up_proj_weight/max_abs":0.12890625,"train/tensor_act_model_layers_41_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_92_mlp_down_proj/norm":31245.439453125,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/max":0.125,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/mean":0.0029449462890625,"train/tensor_act_model_layers_28_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_37_mlp_up_proj/norm":5420.33935546875,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_79_mlp_gate_proj/max_abs":4.59375,"train/tensor_param_model_layers_13_mlp_down_proj_weight/norm":5.4375,"train/tensor_act_model_layers_24_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_mlp/std":0,"train/tensor_act_model_layers_42_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_7_residual_post_mlp/norm":65435.83203125,"train/tensor_act_model_layers_42_mlp_up_proj/norm":5381.84228515625,"train/tensor_act_model_layers_84_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_16_mlp_down_proj/std":0.4140625,"train/tensor_act_model_layers_69_mlp/std":0,"train/tensor_act_model_layers_74_mlp_gate_proj/norm":5116.66845703125,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_68_mlp_gate_proj/norm":4906.59619140625,"train/tensor_act_model_layers_13_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_92_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_78_mlp/min":4,"train/tensor_act_model_layers_36_mlp_gate_proj/min":-1.2578125,"train/tensor_act_model_layers_23_residual_post_attn/std":0.359375,"train/tensor_act_model_layers_23_residual_post_mlp/max_abs":97,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/std":0.03515625,"train/tensor_act_model_layers_69_mlp_up_proj/std":1.8046875,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/norm":5.5,"train/tensor_act_model_layers_19_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_59_mlp_gate_proj/mean":0.515625,"train/tensor_act_model_layers_94_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_99_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_72_mlp_gate_proj/min":-1.2734375,"train/tensor_param_model_layers_57_mlp_down_proj_weight/max":0.1298828125,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_60_mlp_down_proj_weight/min":-0.1318359375,"train/tensor_act_model_layers_46_residual_post_mlp/mean":188,"train/tensor_act_model_layers_81_mlp_gate_proj/max_abs":4.4375,"train/tensor_param_model_layers_72_mlp_down_proj_weight/norm":7,"train/tensor_act_model_layers_78_residual_post_mlp/min":314,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_59_mlp_up_proj_weight/norm":6.75,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/norm":7.4375,"train/tensor_param_model_layers_93_mlp_up_proj_weight/min":-0.173828125,"train/tensor_act_model_layers_43_residual_pre_attn/norm":352108.65625,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/std":0.0303955078125,"train/tensor_act_model_layers_46_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_71_mlp_gate_proj/max_abs":4.3125,"train/tensor_act_model_layers_27_mlp_up_proj/max":4.8125,"train/tensor_act_model_layers_35_residual_pre_attn/max":141,"train/tensor_act_model_layers_1_mlp_gate_proj/min":-1.5625,"train/tensor_act_model_layers_50_mlp/norm":8192,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_mlp_gate_proj/min":-1.7109375,"train/tensor_param_model_layers_79_mlp_up_proj_weight/norm":7.5625,"train/tensor_act_model_layers_33_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_70_mlp/max":4,"train/tensor_param_model_layers_54_mlp_up_proj_weight/min":-0.126953125,"train/tensor_act_model_layers_21_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_71_mlp_up_proj_weight/max_abs":0.1357421875,"train/tensor_act_model_layers_11_residual_post_attn/range":2,"train/tensor_act_model_layers_96_mlp_gate_proj/min":-1.8359375,"train/tensor_act_model_layers_13_residual_pre_attn/max":53,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_15_residual_pre_attn/max_abs":61,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_82_mlp_down_proj_weight/std":0.044189453125,"train/tensor_act_model_layers_3_mlp_down_proj/norm":30383.61328125,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/std":0.03466796875,"train/tensor_act_model_layers_95_mlp/max_abs":4,"train/tensor_act_model_layers_14_mlp/min":4,"train/tensor_act_model_layers_37_mlp_up_proj/max_abs":4.71875,"train/tensor_act_model_layers_42_residual_post_mlp/max":173,"train/tensor_act_model_layers_17_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_61_mlp_down_proj_weight/std":0.037109375,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_25_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_36_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_23_residual_post_attn/max":93,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/std":0.041015625,"train/tensor_act_model_layers_64_mlp_up_proj/max":4.0625,"train/tensor_act_model_layers_9_mlp_gate_proj/min":-1.3671875,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_23_mlp_up_proj_weight/std":0.03173828125,"train/tensor_param_model_layers_30_mlp_down_proj_weight/std":0.031982421875,"train/tensor_act_model_layers_9_mlp_gate_proj/max_abs":4.53125,"train/tensor_act_model_layers_25_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_5_mlp_up_proj/max":4.25,"train/tensor_act_model_layers_54_residual_pre_attn/min":214,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_mlp_up_proj/mean":-0.07861328125,"train/tensor_act_model_layers_65_mlp_down_proj/mean":14.9375,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_52_residual_post_mlp/mean":212,"train/tensor_act_model_layers_13_mlp_down_proj/std":0.39453125,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_6_mlp_up_proj_weight/min":-0.12158203125,"train/tensor_act_model_layers_44_residual_post_attn/range":3,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_54_residual_pre_attn/std":0.408203125,"train/tensor_param_model_layers_41_mlp_down_proj_weight/mean":-0.000514984130859375,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_93_mlp_down_proj/max_abs":17.625,"train/tensor_act_model_layers_99_residual_post_attn/max_abs":396,"train/tensor_act_model_layers_95_residual_post_mlp/max_abs":384,"train/tensor_act_model_layers_54_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_30_mlp_up_proj/min":-4.53125,"train/tensor_act_model_layers_1_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_98_residual_post_mlp/min":394,"train/tensor_act_model_layers_53_residual_pre_attn/std":0.31640625,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_61_mlp_down_proj_weight/max_abs":0.13671875,"train/tensor_param_model_layers_37_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/max":0.240234375,"train/tensor_param_model_layers_14_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_46_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_16_mlp_down_proj_weight/std":0.0303955078125,"train/tensor_act_model_layers_86_residual_post_attn/min":342,"train/tensor_act_model_layers_32_mlp_down_proj/max":15.9375,"train/tensor_act_model_layers_9_mlp_gate_proj/range":5.8984375,"train/tensor_param_model_layers_33_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_85_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_mlp/range":0,"train/tensor_param_model_layers_9_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/range":0.251953125,"train/tensor_param_model_layers_35_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/max":0.12890625,"train/tensor_act_model_layers_35_mlp_down_proj/max":16.25,"train/tensor_act_model_layers_46_mlp/mean":4,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_85_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_40_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_19_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_86_residual_post_mlp/max":348,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_17_mlp_down_proj_weight/std":0.03076171875,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_97_residual_post_attn/max":388,"train/tensor_act_model_layers_0_mlp_down_proj/norm":64239.03125,"train/tensor_act_model_layers_86_residual_post_mlp/norm":712669.25,"train/tensor_act_model_layers_53_residual_pre_attn/min":210,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_51_mlp/norm":8192,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/max":0.16015625,"train/tensor_act_model_layers_11_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_mlp_gate_proj/mean":0.357421875,"train/tensor_act_model_layers_66_mlp_down_proj/mean":14.9375,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_28_residual_post_attn/mean":112,"train/tensor_act_model_layers_89_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_60_mlp/mean":4,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/mean":0.0034027099609375,"train/tensor_act_model_layers_54_residual_post_attn/std":0.408203125,"train/tensor_act_model_layers_41_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_3_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_37_residual_post_attn/max_abs":149,"train/tensor_param_model_layers_52_mlp_up_proj_weight/std":0.03662109375,"train/tensor_param_model_layers_36_mlp_up_proj_weight/mean":-0.001220703125,"train/tensor_act_model_layers_56_mlp_up_proj/range":8.46875,"train/tensor_act_model_layers_74_residual_post_attn/norm":606173.3125,"train/tensor_act_model_layers_50_residual_post_attn/max":201,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_1_residual_pre_attn/max_abs":4.3125,"train/tensor_param_model_layers_87_mlp_down_proj_weight/min":-0.1396484375,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_94_residual_post_mlp/range":2,"train/tensor_act_model_layers_80_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp/max":4,"train/tensor_act_model_layers_73_mlp/norm":8192,"train/tensor_param_model_layers_78_mlp_down_proj_weight/max_abs":0.142578125,"train/tensor_act_model_layers_24_residual_pre_attn/max_abs":97,"train/tensor_param_model_layers_76_mlp_up_proj_weight/min":-0.1318359375,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_4_mlp/min":4,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_65_mlp_up_proj/std":1.796875,"train/tensor_act_model_layers_10_mlp_gate_proj/max_abs":4.09375,"train/tensor_param_model_layers_27_mlp_up_proj_weight/max":0.1181640625,"train/tensor_param_model_layers_36_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_29_mlp_gate_proj/norm":5207.7958984375,"train/tensor_act_model_layers_7_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_21_mlp_down_proj/mean":14.6875,"train/tensor_param_model_layers_50_mlp_down_proj_weight/min":-0.130859375,"train/tensor_act_model_layers_44_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_52_mlp_up_proj/max_abs":4.8125,"train/tensor_act_model_layers_18_mlp_down_proj/range":1.4375,"train/tensor_act_model_layers_43_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_57_mlp_up_proj/mean":-0.042724609375,"train/tensor_act_model_layers_69_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_12_mlp_down_proj_weight/range":0.2353515625,"train/tensor_act_model_layers_56_residual_post_attn/max_abs":225,"train/tensor_act_model_layers_82_residual_post_attn/max":328,"train/tensor_param_model_layers_65_mlp_up_proj_weight/max":0.1357421875,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_89_residual_pre_attn/mean":356,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_56_mlp_down_proj/range":2.1875,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_91_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_67_mlp_down_proj_weight/std":0.037841796875,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/range":0.240234375,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/min":-0.1181640625,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_2_mlp/std":0,"train/tensor_act_model_layers_46_residual_post_attn/range":3,"train/tensor_act_model_layers_16_mlp/mean":4,"train/tensor_param_model_layers_34_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_residual_post_mlp/max_abs":245,"train/tensor_act_model_layers_33_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_8_mlp/norm":8192,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_86_mlp_down_proj/max_abs":17.25,"train/tensor_param_model_layers_34_mlp_down_proj_weight/std":0.0322265625,"train/tensor_param_model_layers_25_mlp_down_proj_weight/range":0.2412109375,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_63_mlp_down_proj/norm":30376.05078125,"train/tensor_param_model_layers_89_mlp_up_proj_weight/norm":8.5625,"train/tensor_act_model_layers_54_residual_post_attn/max":217,"train/tensor_act_model_layers_99_residual_pre_attn/max_abs":396,"train/tensor_act_model_layers_77_mlp_up_proj/max":4.71875,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/mean":0.00445556640625,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_92_mlp_down_proj/max":17.875,"train/tensor_act_model_layers_18_mlp/min":4,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_72_mlp/norm":8192,"train/tensor_param_model_layers_56_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_0_residual_post_mlp/std":0.091796875,"train/tensor_param_model_layers_83_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_30_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_48_mlp_down_proj/norm":30427.515625,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/std":0.03662109375,"train/tensor_act_model_layers_87_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_97_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_47_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_residual_post_attn/max_abs":105,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_57_residual_pre_attn/range":3,"train/tensor_act_model_layers_90_residual_pre_attn/max_abs":360,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_3_mlp_down_proj_weight/max":0.10791015625,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_48_residual_pre_attn/min":190,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_residual_pre_attn/max_abs":32.75,"train/tensor_param_model_layers_25_mlp_up_proj_weight/mean":-0.000579833984375,"train/tensor_act_model_layers_50_residual_post_attn/mean":200,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_18_mlp_down_proj_weight/mean":-0.000514984130859375,"train/tensor_act_model_layers_23_mlp_gate_proj/max":4.75,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/max":1952,"train/tensor_act_model_layers_15_mlp_gate_proj/min":-1.3203125,"train/tensor_act_model_layers_59_residual_post_attn/range":3,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_22_residual_post_mlp/min":90.5,"train/tensor_param_model_layers_85_mlp_down_proj_weight/max":0.1474609375,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_46_mlp_up_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_99_mlp_up_proj/min":-5.5625,"train/tensor_act_model_layers_5_mlp_up_proj/min":-4.25,"train/tensor_param_model_layers_59_mlp_down_proj_weight/max_abs":0.1396484375,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_22_mlp_down_proj/std":0.3203125,"train/tensor_act_model_layers_82_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_75_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/mean":0.00469970703125,"train/tensor_param_model_layers_4_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_68_residual_pre_attn/mean":272,"train/tensor_act_model_layers_86_mlp/max":4,"train/tensor_act_model_layers_43_residual_post_attn/mean":172,"train/tensor_act_model_layers_66_mlp/mean":4,"train/tensor_act_model_layers_50_residual_post_attn/std":0.31640625,"train/tensor_param_model_layers_94_mlp_down_proj_weight/std":0.050537109375,"train/tensor_act_model_layers_96_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_68_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_55_mlp_up_proj_weight/std":0.037109375,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/std":0.0478515625,"train/tensor_act_model_layers_48_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_59_mlp/mean":4,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_9_residual_post_mlp/range":1.75,"train/tensor_act_model_layers_74_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_31_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_42_mlp_up_proj/max_abs":4.625,"train/tensor_act_model_layers_98_mlp/max":4,"train/tensor_act_model_layers_14_mlp_up_proj/max_abs":4.46875,"train/tensor_act_model_layers_16_mlp/min":4,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_75_mlp_down_proj/std":0.416015625,"train/tensor_param_model_layers_53_mlp_up_proj_weight/max_abs":0.1337890625,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_5_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_92_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_89_residual_post_attn/max":356,"train/tensor_act_model_layers_82_residual_pre_attn/mean":328,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_39_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_38_residual_pre_attn/norm":311148.71875,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/range":0.3134765625,"train/tensor_act_model_layers_4_mlp_gate_proj/min":-1.4609375,"train/tensor_act_model_layers_31_mlp_down_proj/std":0.44140625,"train/tensor_act_model_layers_98_mlp_up_proj/mean":-0.2412109375,"train/tensor_act_model_layers_40_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_23_mlp_up_proj/max_abs":4.875,"train/tensor_param_model_layers_57_mlp_down_proj_weight/min":-0.125,"train/tensor_act_model_layers_68_mlp/max_abs":4,"train/tensor_act_model_layers_63_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_50_residual_pre_attn/mean":200,"train/tensor_act_model_layers_93_residual_pre_attn/norm":761821.25,"train/tensor_param_model_layers_60_mlp_down_proj_weight/range":0.259765625,"train/tensor_act_model_layers_68_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_33_residual_post_mlp/std":0.31640625,"train/tensor_param_model_layers_35_mlp_down_proj_weight/std":0.032470703125,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_28_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_23_mlp_up_proj/range":9.71875,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/max_abs":0.158203125,"train/tensor_act_model_layers_29_mlp_up_proj/min":-3.984375,"train/tensor_act_model_layers_16_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_71_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_60_mlp_down_proj_weight/max":0.1279296875,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_90_mlp/max":4,"train/tensor_param_model_layers_97_mlp_up_proj_weight/std":0.053466796875,"train/tensor_act_model_layers_61_mlp_gate_proj/mean":0.5703125,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_93_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_26_mlp_gate_proj_weight/mean":0.00518798828125,"train/tensor_act_model_layers_97_mlp/mean":4,"train/tensor_act_model_layers_76_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_51_mlp_down_proj_weight/mean":0.00021839141845703125,"train/tensor_param_model_layers_44_mlp_up_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_48_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_19_mlp_up_proj_weight/max":0.1162109375,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_26_residual_pre_attn/min":102.5,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_70_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_54_mlp_down_proj_weight/mean":-0.0015716552734375,"train/tensor_param_model_layers_61_mlp_up_proj_weight/mean":-0.0006103515625,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_85_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_53_mlp_up_proj_weight/min":-0.12451171875,"train/tensor_act_model_layers_19_mlp_gate_proj/norm":5045.552734375,"train/tensor_act_model_layers_57_mlp_gate_proj/std":1.671875,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_26_residual_pre_attn/std":0.359375,"train/tensor_act_model_layers_65_mlp_down_proj/norm":30665.39453125,"train/tensor_act_model_layers_45_mlp_down_proj/max_abs":16.5,"train/tensor_param_model_layers_88_mlp_down_proj_weight/max_abs":0.1484375,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_6_residual_post_attn/min":23.125,"train/tensor_act_model_layers_21_residual_post_mlp/max_abs":89,"train/tensor_act_model_layers_38_mlp/max":4,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_92_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_96_mlp_down_proj_weight/max_abs":0.17578125,"train/tensor_act_model_layers_97_residual_pre_attn/mean":388,"train/tensor_param_model_layers_99_mlp_up_proj_weight/std":0.05517578125,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_74_mlp/max":4,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_mlp_down_proj/max":19.75,"train/tensor_act_model_layers_29_residual_post_mlp/max_abs":121,"train/tensor_act_model_layers_16_mlp_up_proj/mean":-0.0771484375,"train/tensor_act_model_layers_68_residual_post_attn/max":272,"train/tensor_act_model_layers_60_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_4_residual_post_attn/range":1.3125,"train/tensor_param_model_layers_40_mlp_down_proj_weight/std":0.03369140625,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_25_mlp/range":0,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_residual_pre_attn/max_abs":185,"train/tensor_param_model_layers_2_mlp_down_proj_weight/mean":-0.00146484375,"train/tensor_act_model_layers_1_residual_pre_attn/std":0.091796875,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_33_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_6_mlp_gate_proj/norm":5148.900390625,"train/tensor_act_model_layers_17_residual_pre_attn/max_abs":69,"train/tensor_act_model_layers_80_mlp_up_proj/max_abs":4.40625,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_20_residual_post_mlp/min":82.5,"train/tensor_act_model_layers_39_residual_post_attn/range":3,"train/tensor_act_model_layers_99_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_residual_pre_attn/max":185,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_94_mlp_gate_proj/norm":4993.33154296875,"train/tensor_param_model_layers_82_mlp_up_proj_weight/max_abs":0.14453125,"train/tensor_act_model_layers_89_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_66_mlp_up_proj_weight/min":-0.126953125,"train/tensor_act_model_layers_9_residual_pre_attn/min":35,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_10_mlp_down_proj_weight/mean":0.0004558563232421875,"train/tensor_act_model_layers_56_mlp_gate_proj/max":4.1875,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_84_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_81_residual_post_mlp/norm":671709.25,"train/tensor_param_model_layers_74_mlp_down_proj_weight/std":0.038818359375,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/std":282,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_92_mlp_down_proj/mean":15.25,"train/tensor_act_model_layers_63_residual_post_mlp/mean":256,"train/tensor_param_model_layers_49_mlp_down_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_73_mlp_up_proj/max_abs":4.78125,"train/tensor_act_model_layers_44_mlp/mean":4,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_36_mlp_down_proj_weight/mean":-0.00122833251953125,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_51_residual_post_mlp/norm":425836.53125,"train/tensor_act_model_layers_2_residual_post_mlp/max_abs":12.4375,"train/tensor_act_model_layers_18_residual_post_mlp/max":77,"train/tensor_act_model_layers_73_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_6_residual_post_mlp/range":1.5,"train/tensor_act_model_layers_36_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_58_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_residual_post_attn/min":202,"train/tensor_param_model_layers_1_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_62_residual_post_attn/max_abs":249,"train/tensor_act_model_layers_63_mlp_up_proj/min":-4.09375,"train/tensor_act_model_layers_53_mlp_up_proj/min":-4.71875,"train/tensor_act_model_layers_97_residual_post_attn/mean":388,"train/tensor_act_model_layers_38_mlp_up_proj/max":4.3125,"train/tensor_act_model_layers_65_residual_pre_attn/norm":532445.3125,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/std":0.04833984375,"train/tensor_act_model_layers_44_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_residual_post_attn/frac_near_user_limit":0,"train/tensor_act_model_layers_54_residual_pre_attn/mean":216,"train/tensor_act_model_layers_7_residual_post_mlp/std":0.1962890625,"train/tensor_act_model_layers_18_mlp_down_proj/std":0.330078125,"train/tensor_act_model_layers_81_residual_post_attn/min":322,"train/tensor_act_model_layers_11_mlp/range":0,"train/tensor_act_model_layers_76_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_11_mlp_gate_proj/range":5.1484375,"train/tensor_act_model_layers_7_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_mlp/norm":8192,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_68_mlp_down_proj/max_abs":17.125,"train/tensor_grad_model_layers_49_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_63_mlp_down_proj/mean":14.8125,"train/tensor_param_model_layers_46_mlp_up_proj_weight/range":0.2509765625,"train/tensor_act_model_layers_52_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_0_mlp_up_proj_weight/min":-0.125,"train/tensor_act_model_layers_30_residual_pre_attn/std":0.388671875,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_residual_post_attn/max_abs":264,"train/tensor_param_model_layers_19_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_34_mlp_up_proj/max":4.59375,"train/tensor_act_model_layers_24_mlp_up_proj/max_abs":4.375,"train/tensor_act_model_layers_87_mlp_up_proj/std":1.765625,"train/tensor_act_model_layers_83_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_82_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_58_mlp_down_proj/std":0.5546875,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_94_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_mlp_gate_proj/norm":5077.09033203125,"train/tensor_act_model_layers_28_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_33_residual_pre_attn/range":3,"train/tensor_act_model_layers_85_residual_post_mlp/min":342,"train/tensor_param_model_layers_51_mlp_up_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_21_residual_post_attn/min":82.5,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_77_mlp/max_abs":4,"train/tensor_act_model_layers_91_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_37_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_33_residual_pre_attn/max":133,"train/tensor_act_model_layers_24_mlp/max_abs":4,"train/tensor_act_model_layers_55_residual_post_mlp/range":3,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_17_mlp_up_proj_weight/max":0.11572265625,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_32_mlp/std":0,"train/tensor_act_model_layers_81_mlp_up_proj/mean":-0.078125,"train/tensor_act_model_layers_15_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/range":0.25048828125,"train/tensor_act_model_layers_77_residual_pre_attn/min":306,"train/tensor_param_model_layers_17_mlp_down_proj_weight/max":0.10205078125,"train/tensor_act_model_layers_85_mlp_gate_proj/mean":0.6328125,"train/tensor_param_model_layers_7_mlp_up_proj_weight/norm":5.40625,"train/tensor_act_model_layers_81_mlp/mean":4,"train/tensor_param_model_layers_93_mlp_down_proj_weight/max_abs":0.1630859375,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_63_mlp_gate_proj/range":6.1484375,"train/tensor_param_model_layers_23_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_11_residual_pre_attn/norm":90037.5546875,"train/tensor_act_model_layers_4_mlp_down_proj/norm":29986.625,"train/tensor_act_model_layers_77_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_2_mlp_gate_proj/std":1.6953125,"train/tensor_param_model_layers_16_mlp_down_proj_weight/max":0.1083984375,"train/tensor_param_model_layers_65_mlp_down_proj_weight/norm":6.78125,"train/tensor_act_model_layers_75_mlp_gate_proj/max_abs":4.3125,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/std":0.0537109375,"train/tensor_act_model_layers_8_residual_post_mlp/std":0.2197265625,"train/tensor_act_model_layers_91_mlp_down_proj/min":13.875,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/std":0.03662109375,"train/tensor_param_model_layers_88_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_26_mlp_up_proj/min":-4.15625,"train/tensor_act_model_layers_35_residual_pre_attn/mean":140,"train/tensor_param_model_layers_33_mlp_up_proj_weight/range":0.24853515625,"train/tensor_act_model_layers_49_mlp_up_proj/max_abs":4.40625,"train/tensor_act_model_layers_21_residual_post_mlp/norm":179985.390625,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/mean":0.0040283203125,"train/tensor_act_model_layers_9_residual_pre_attn/range":1.75,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_90_mlp_down_proj_weight/max_abs":0.154296875,"train/tensor_act_model_layers_36_residual_pre_attn/max":145,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_95_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_94_mlp_down_proj_weight/max":0.1767578125,"train/tensor_param_model_layers_58_mlp_down_proj_weight/norm":6.6875,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_act_model_layers_7_residual_post_attn/max":28.75,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_62_mlp_up_proj/max":4.9375,"train/tensor_act_model_layers_97_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_60_mlp_up_proj_weight/max":0.1298828125,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_92_mlp_up_proj/min":-5.125,"train/tensor_act_model_layers_41_mlp/max_abs":4,"train/tensor_act_model_layers_43_mlp/max":4,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_47_residual_post_attn/norm":384876.59375,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/max":0.1435546875,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_29_mlp_gate_proj/range":6.375,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_89_mlp/mean":4,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_45_mlp_up_proj_weight/mean":-0.00116729736328125,"train/tensor_act_model_layers_88_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_66_mlp_down_proj/std":0.5390625,"train/tensor_act_model_layers_95_residual_pre_attn/min":378,"train/tensor_act_model_layers_56_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_17_mlp_gate_proj/norm":5211.57421875,"train/tensor_act_model_layers_52_residual_post_attn/norm":425836.53125,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_75_mlp_down_proj_weight/min":-0.14453125,"train/tensor_param_model_layers_85_mlp_down_proj_weight/std":0.044921875,"train/tensor_act_model_layers_84_mlp_down_proj/max":17.25,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_25_residual_post_attn/std":0.359375,"train/tensor_act_model_layers_29_mlp_down_proj/std":0.419921875,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_38_mlp_gate_proj/max_abs":4.65625,"train/tensor_act_model_layers_90_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_40_residual_pre_attn/norm":327532.6875,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_3_mlp/range":0,"train/tensor_act_model_layers_56_mlp_up_proj/norm":5319.07666015625,"train/tensor_act_model_layers_46_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_73_residual_post_mlp/min":294,"train/tensor_act_model_layers_2_mlp/norm":8192,"train/tensor_act_model_layers_45_residual_post_attn/range":3,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/max":0.12890625,"train/tensor_act_model_layers_16_residual_post_attn/range":2.25,"train/tensor_act_model_layers_37_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_39_mlp/mean":4,"train/tensor_act_model_layers_11_mlp/min":4,"train/tensor_act_model_layers_78_mlp_down_proj/norm":30810.072265625,"train/tensor_act_model_layers_78_mlp_up_proj/mean":0.2119140625,"train/tensor_act_model_layers_44_mlp/range":0,"train/tensor_act_model_layers_74_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_35_mlp_down_proj_weight/mean":-0.000667572021484375,"train/tensor_act_model_layers_2_residual_pre_attn/norm":16366.814453125,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_59_mlp_up_proj/max_abs":5.03125,"train/tensor_act_model_layers_83_mlp_down_proj/range":5.5,"train/tensor_act_model_layers_15_mlp_up_proj/std":1.8671875,"train/tensor_act_model_layers_85_mlp/max":4,"train/tensor_param_model_layers_67_mlp_down_proj_weight/max":0.126953125,"train/tensor_param_model_layers_88_mlp_down_proj_weight/mean":-0.0004596710205078125,"train/tensor_act_model_layers_96_mlp/range":0,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/mean":0.0029449462890625,"train/tensor_param_model_layers_34_mlp_down_proj_weight/max":0.12890625,"train/tensor_act_model_layers_47_mlp_up_proj/std":1.828125,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/mean":0.0062255859375,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_83_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_53_mlp_up_proj_weight/mean":-0.00054931640625,"train/tensor_param_model_layers_82_mlp_down_proj_weight/min":-0.1376953125,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_61_mlp_down_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_94_mlp_gate_proj/max_abs":5.34375,"train/tensor_act_model_layers_80_mlp_up_proj/std":1.765625,"train/tensor_act_model_layers_81_mlp_down_proj/range":3.4375,"train/tensor_act_model_layers_12_residual_post_attn/norm":98137.8984375,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_53_mlp_up_proj_weight/range":0.25830078125,"train/tensor_act_model_layers_84_mlp_down_proj/range":3.0625,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/min":-0.169921875,"train/tensor_act_model_layers_74_residual_post_mlp/mean":300,"train/tensor_act_model_layers_1_mlp_down_proj/std":1.2109375,"train/tensor_act_model_layers_5_residual_post_attn/min":19.25,"train/tensor_act_model_layers_64_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_95_mlp_up_proj_weight/std":0.052001953125,"train/tensor_param_model_layers_30_mlp_down_proj_weight/min":-0.109375,"train/tensor_act_model_layers_35_residual_post_attn/max":141,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/mean":0.004302978515625,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_94_mlp_up_proj_weight/max_abs":0.16796875,"train/tensor_param_model_layers_81_mlp_down_proj_weight/norm":7.90625,"train/tensor_act_model_layers_28_residual_post_mlp/max_abs":117,"train/tensor_act_model_layers_61_residual_pre_attn/max_abs":245,"train/tensor_act_model_layers_79_mlp_up_proj/max_abs":4.4375,"train/tensor_act_model_layers_22_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_53_mlp/max_abs":4,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_37_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_53_mlp_gate_proj/norm":5119.3583984375,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/norm":7.0625,"train/tensor_act_model_layers_29_mlp/std":0,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_78_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_35_mlp_down_proj_weight/max":0.111328125,"train/tensor_act_model_layers_25_residual_pre_attn/min":98.5,"train/tensor_act_model_layers_3_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_95_mlp_up_proj_weight/mean":0.0014495849609375,"train/tensor_act_model_layers_63_residual_post_attn/max_abs":253,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_31_mlp_up_proj/max":3.90625,"train/tensor_act_model_layers_14_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_57_mlp_down_proj_weight/range":0.2548828125,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_42_mlp_up_proj_weight/mean":0.000919342041015625,"train/tensor_act_model_layers_41_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_99_residual_pre_attn/min":394,"train/tensor_act_model_layers_86_mlp_up_proj/max":4.65625,"train/tensor_param_model_layers_60_mlp_down_proj_weight/std":0.036865234375,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_85_mlp/std":0,"train/tensor_param_model_layers_68_mlp_up_proj_weight/max":0.134765625,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_90_residual_pre_attn/norm":737245.25,"train/tensor_act_model_layers_55_mlp_gate_proj/max_abs":4.6875,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_96_residual_post_mlp/min":386,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/norm":6.90625,"train/tensor_act_model_layers_9_mlp_down_proj/std":0.4296875,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_residual_pre_attn/range":1.375,"train/tensor_act_model_layers_43_mlp_gate_proj/max_abs":4.65625,"train/tensor_act_model_layers_23_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/max_abs":0.14453125,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_39_residual_post_mlp/max":161,"train/tensor_act_model_layers_16_residual_post_mlp/min":67,"train/tensor_act_model_layers_66_mlp_down_proj/max_abs":17,"train/tensor_act_model_layers_99_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/min":-0.12158203125,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_25_residual_pre_attn/std":0.359375,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_25_residual_pre_attn/mean":100,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/norm":6.875,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_39_residual_pre_attn/norm":319340.71875,"train/tensor_act_model_layers_73_mlp_up_proj/mean":0.11474609375,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_92_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_79_residual_post_attn/norm":647133.3125,"train/tensor_act_model_layers_95_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_72_residual_post_attn/mean":288,"train/tensor_act_model_layers_32_mlp/max":4,"train/tensor_act_model_layers_13_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_76_residual_post_mlp/mean":308,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_64_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_31_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_51_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/norm":8.5,"train/tensor_act_model_layers_27_mlp/norm":8192,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_30_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_78_mlp/std":0,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_91_residual_post_attn/max_abs":364,"train/tensor_act_model_layers_94_residual_post_mlp/min":378,"train/tensor_param_model_layers_58_mlp_down_proj_weight/max_abs":0.1337890625,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/max":0.1435546875,"train/tensor_act_model_layers_47_mlp/max_abs":4,"train/tensor_act_model_layers_74_residual_pre_attn/max":296,"train/tensor_param_model_layers_16_mlp_up_proj_weight/range":0.21826171875,"train/tensor_param_model_layers_85_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_72_mlp/max_abs":4,"train/tensor_param_model_layers_74_mlp_down_proj_weight/max":0.1328125,"train/tensor_act_model_layers_6_mlp_gate_proj/std":1.6953125,"train/tensor_act_model_layers_59_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_25_mlp_gate_proj/mean":0.57421875,"train/tensor_act_model_layers_11_mlp_down_proj/norm":29886.126953125,"train/tensor_act_model_layers_31_mlp_down_proj/mean":14.8125,"train/tensor_act_model_layers_21_mlp_gate_proj/max_abs":4.46875,"train/tensor_param_model_layers_5_mlp_down_proj_weight/max":0.10302734375,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_52_mlp/max":4,"train/tensor_act_model_layers_80_mlp/max_abs":4,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_27_mlp_gate_proj/norm":5141.8818359375,"train/tensor_act_model_layers_44_residual_post_mlp/max":181,"train/tensor_act_model_layers_15_mlp_gate_proj/std":1.7265625,"train/tensor_param_model_layers_94_mlp_up_proj_weight/min":-0.16796875,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_87_mlp_down_proj/std":0.78125,"train/tensor_act_model_layers_33_mlp_up_proj/norm":5550.23876953125,"train/tensor_act_model_layers_69_residual_post_attn/norm":565213.3125,"train/tensor_act_model_layers_35_residual_post_attn/std":0.31640625,"train/tensor_param_model_layers_51_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_43_mlp_up_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_76_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_9_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_84_mlp_up_proj_weight/std":0.044921875,"train/tensor_act_model_layers_5_mlp_down_proj/min":14.0625,"train/tensor_act_model_layers_91_mlp/max_abs":4,"train/tensor_act_model_layers_78_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_68_mlp/mean":4,"train/tensor_act_model_layers_28_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_47_mlp_up_proj_weight/range":0.2578125,"train/tensor_param_model_layers_8_mlp_up_proj_weight/range":0.21630859375,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_49_mlp_down_proj_weight/mean":-0.0010986328125,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_64_mlp/max":4,"train/tensor_act_model_layers_35_mlp_up_proj/mean":0.05908203125,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_35_residual_pre_attn/range":3,"train/tensor_act_model_layers_29_residual_post_mlp/mean":120,"train/tensor_act_model_layers_82_mlp_up_proj/max":4.25,"train/tensor_act_model_layers_48_mlp_down_proj/min":14.25,"train/tensor_act_model_layers_6_mlp/min":4,"train/tensor_param_model_layers_61_mlp_up_proj_weight/max":0.1298828125,"train/tensor_param_model_layers_93_mlp_down_proj_weight/norm":8.9375,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_25_residual_pre_attn/max_abs":101,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/range":0.2626953125,"train/tensor_param_model_layers_69_mlp_up_proj_weight/range":0.2548828125,"train/tensor_param_model_layers_26_mlp_up_proj_weight/norm":5.75,"train/tensor_param_model_layers_2_mlp_up_proj_weight/mean":0.0003681182861328125,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_40_mlp_down_proj_weight/max":0.12158203125,"train/tensor_act_model_layers_32_mlp_up_proj/max":4.0625,"train/tensor_act_model_layers_88_residual_pre_attn/max_abs":352,"train/tensor_act_model_layers_32_residual_post_mlp/range":3,"train/tensor_act_model_layers_4_mlp_gate_proj/max":3.953125,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_96_mlp_up_proj_weight/max":0.1689453125,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/max_abs":1952,"train/tensor_act_model_layers_84_residual_pre_attn/max":336,"train/tensor_act_model_layers_71_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_74_mlp_down_proj_weight/norm":7.03125,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_92_mlp/range":0,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_22_mlp_down_proj_weight/mean":-0.00124359130859375,"train/tensor_param_model_layers_6_mlp_down_proj_weight/mean":-0.00022220611572265625,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_47_mlp_up_proj_weight/std":0.03466796875,"train/tensor_act_model_layers_36_mlp_gate_proj/std":1.65625,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_65_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_96_mlp_down_proj_weight/range":0.3349609375,"train/tensor_act_model_layers_2_mlp_up_proj/max_abs":4.25,"train/tensor_param_model_layers_75_mlp_down_proj_weight/max":0.12890625,"train/tensor_act_model_layers_75_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_49_mlp_gate_proj/norm":5049.44873046875,"train/tensor_act_model_layers_28_mlp_up_proj/mean":-0.0005645751953125,"train/tensor_act_model_layers_25_mlp_down_proj/range":2.25,"train/tensor_act_model_layers_54_residual_post_mlp/max_abs":221,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/mean":0.0050048828125,"train/tensor_act_model_layers_85_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_41_residual_pre_attn/max":165,"train/tensor_act_model_layers_13_mlp_gate_proj/norm":5055.96484375,"train/tensor_act_model_layers_71_residual_post_attn/range":2,"train/tensor_param_model_layers_94_mlp_down_proj_weight/min":-0.1572265625,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/max":0.1279296875,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/min":-0.158203125,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_69_mlp_down_proj_weight/norm":6.84375,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_5_mlp_up_proj/std":1.8359375,"train/tensor_act_model_layers_21_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_5_mlp_up_proj_weight/min":-0.1201171875,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_74_residual_pre_attn/max_abs":296,"train/tensor_param_model_layers_40_mlp_up_proj_weight/mean":0.000518798828125,"train/tensor_act_model_layers_73_mlp_gate_proj/norm":4798.0166015625,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/std":0.039794921875,"train/tensor_param_model_layers_7_mlp_up_proj_weight/mean":0.00011110305786132812,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/range":0.26953125,"train/tensor_act_model_layers_66_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/mean":0.0027923583984375,"train/tensor_param_model_layers_67_mlp_down_proj_weight/min":-0.130859375,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_85_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_residual_post_mlp/min":242,"train/tensor_act_model_layers_65_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_34_mlp_down_proj_weight/min":-0.12353515625,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/norm":5.96875,"train/tensor_param_model_layers_75_mlp_up_proj_weight/max_abs":0.1318359375,"train/tensor_act_model_layers_6_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_78_mlp_down_proj/mean":15.0625,"train/tensor_param_model_layers_13_mlp_up_proj_weight/mean":0.000209808349609375,"train/tensor_act_model_layers_11_residual_post_mlp/min":47,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_92_mlp_down_proj_weight/max_abs":0.15234375,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/max_abs":0.130859375,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_24_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_14_mlp_gate_proj/max_abs":4.28125,"train/tensor_act_model_layers_66_residual_pre_attn/mean":264,"train/tensor_act_model_layers_85_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_residual_post_mlp/std":0.1298828125,"train/tensor_act_model_layers_91_mlp/norm":8192,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/norm":7.125,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/min":-0.12890625,"train/tensor_act_model_layers_85_mlp_up_proj/norm":5154.626953125,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_51_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_mlp/std":0,"train/tensor_param_model_layers_8_mlp_up_proj_weight/min":-0.0966796875,"train/tensor_act_model_layers_9_mlp_up_proj/mean":-0.2197265625,"train/tensor_act_model_layers_95_mlp_up_proj/mean":0.185546875,"train/tensor_act_model_layers_7_mlp_gate_proj/max_abs":4.34375,"train/tensor_act_model_layers_56_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_21_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_35_mlp_down_proj/max_abs":16.25,"train/tensor_act_model_layers_32_mlp_gate_proj/max_abs":4.59375,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_94_residual_post_mlp/mean":380,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_91_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_3_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_62_mlp_up_proj_weight/norm":6.8125,"train/tensor_act_model_layers_80_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_39_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_63_mlp/std":0,"train/tensor_param_model_layers_87_mlp_down_proj_weight/std":0.044921875,"train/tensor_act_model_layers_64_mlp_up_proj/mean":-0.1728515625,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_91_mlp_gate_proj/norm":5170.48828125,"train/tensor_param_model_layers_9_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_63_mlp_up_proj_weight/range":0.2568359375,"train/tensor_act_model_layers_37_residual_pre_attn/norm":302956.75,"train/tensor_act_model_layers_21_residual_post_attn/max_abs":85,"train/tensor_param_model_layers_76_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_60_mlp_up_proj/range":8.1875,"train/tensor_param_model_layers_72_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_5_residual_pre_attn/max_abs":20.625,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/std":0.030517578125,"train/tensor_act_model_layers_44_mlp/max_abs":4,"train/tensor_act_model_layers_85_residual_pre_attn/mean":340,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/range":0.28125,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_84_residual_post_attn/max_abs":336,"train/tensor_param_model_layers_50_mlp_down_proj_weight/std":0.03564453125,"train/tensor_param_model_layers_66_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_63_mlp/max_abs":4,"train/tensor_grad_model_layers_89_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_91_residual_pre_attn/range":2,"train/tensor_act_model_layers_67_mlp_down_proj/max":17.625,"train/tensor_param_model_layers_48_mlp_down_proj_weight/range":0.25,"train/tensor_act_model_layers_6_residual_pre_attn/min":23.25,"train/tensor_act_model_layers_15_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_15_residual_post_mlp/max":65,"train/tensor_act_model_layers_35_residual_post_mlp/std":0.31640625,"train/tensor_param_model_layers_27_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_58_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_95_mlp_gate_proj/std":1.625,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_92_mlp_down_proj_weight/range":0.3037109375,"train/tensor_act_model_layers_86_mlp_down_proj/range":3.3125,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_16_mlp_down_proj_weight/min":-0.12158203125,"train/tensor_act_model_layers_56_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_46_mlp_up_proj_weight/mean":-5.793571472167969e-05,"train/tensor_act_model_layers_67_residual_post_attn/mean":268,"train/tensor_act_model_layers_91_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_59_residual_post_attn/min":234,"train/tensor_act_model_layers_18_residual_post_mlp/norm":155434.3125,"train/tensor_act_model_layers_43_residual_post_attn/range":3,"train/tensor_act_model_layers_94_mlp_up_proj/mean":0.1103515625,"train/tensor_act_model_layers_96_mlp_gate_proj/std":1.6171875,"train/tensor_act_model_layers_13_residual_post_mlp/std":0.267578125,"train/tensor_act_model_layers_58_residual_post_attn/mean":232,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_75_residual_post_mlp/range":2,"train/tensor_param_model_layers_3_mlp_down_proj_weight/max_abs":0.10791015625,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_13_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_17_mlp_down_proj_weight/min":-0.11669921875,"train/tensor_act_model_layers_12_mlp_down_proj/range":2,"train/tensor_act_model_layers_72_mlp_gate_proj/std":1.6171875,"train/tensor_act_model_layers_1_residual_post_attn/mean":3.984375,"train/tensor_param_model_layers_32_mlp_up_proj_weight/norm":5.84375,"train/tensor_act_model_layers_59_residual_post_mlp/norm":491292.78125,"train/tensor_param_model_layers_36_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_18_residual_post_attn/min":70.5,"train/tensor_param_model_layers_13_mlp_up_proj_weight/std":0.0301513671875,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_52_mlp_up_proj/norm":5342.94677734375,"train/tensor_param_model_layers_41_mlp_down_proj_weight/max_abs":0.1259765625,"train/tensor_act_model_layers_5_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_7_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_residual_post_mlp/max":392,"train/tensor_act_model_layers_9_mlp/max_abs":4,"train/tensor_act_model_layers_5_residual_post_mlp/max":24.625,"train/tensor_act_model_layers_80_mlp_gate_proj/max_abs":4.5,"train/tensor_act_model_layers_65_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_24_mlp_gate_proj/min":-1.3046875,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/mean":0.00457763671875,"train/tensor_act_model_layers_40_mlp/std":0,"train/tensor_param_model_layers_74_mlp_down_proj_weight/min":-0.126953125,"train/tensor_act_model_layers_66_residual_post_attn/max":264,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_33_mlp/max":4,"train/tensor_act_model_layers_44_mlp_up_proj/min":-4.40625,"train/tensor_param_model_layers_76_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_21_mlp/std":0,"train/tensor_act_model_layers_53_residual_post_mlp/range":3,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_73_residual_post_attn/max_abs":292,"train/tensor_param_model_layers_44_mlp_down_proj_weight/mean":0.00054168701171875,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_13_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_86_mlp_down_proj/max":17.25,"train/tensor_act_model_layers_38_residual_post_mlp/range":3,"train/tensor_act_model_layers_47_residual_post_attn/max_abs":189,"train/tensor_act_model_layers_11_mlp_down_proj/mean":14.5625,"train/tensor_act_model_layers_61_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_51_mlp_down_proj/max_abs":16.5,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_10_residual_post_attn/mean":40,"train/tensor_act_model_layers_97_mlp/std":0,"train/tensor_act_model_layers_72_residual_pre_attn/norm":589789.25,"train/tensor_act_model_layers_95_mlp_down_proj/max_abs":18.375,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_19_mlp/norm":8192,"train/tensor_param_model_layers_75_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_mlp_gate_proj/max_abs":4.40625,"train/tensor_act_model_layers_31_mlp_down_proj/norm":30346.072265625,"train/tensor_act_model_layers_9_mlp_gate_proj/max":4.53125,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/min":-0.14453125,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_42_mlp_down_proj/std":0.435546875,"train/tensor_act_model_layers_73_residual_pre_attn/norm":597981.3125,"train/tensor_act_model_layers_1_mlp_gate_proj/max":4.9375,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_45_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_act_model_layers_89_mlp_gate_proj/max_abs":4.53125,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_93_mlp_up_proj/norm":5298.93896484375,"train/tensor_act_model_layers_82_mlp_up_proj/max_abs":4.8125,"train/tensor_act_model_layers_89_residual_pre_attn/max":356,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_39_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_40_mlp_down_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_67_mlp_up_proj_weight/min":-0.138671875,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_10_residual_post_attn/min":39,"train/tensor_act_model_layers_68_mlp/std":0,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_60_mlp_gate_proj/max":4.4375,"train/tensor_act_model_layers_24_mlp_gate_proj/norm":5159.1923828125,"train/tensor_act_model_layers_93_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_61_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_34_residual_post_attn/mean":136,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/std":0.029541015625,"train/tensor_act_model_layers_73_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_83_mlp/min":4,"train/tensor_act_model_layers_56_residual_post_mlp/range":3,"train/tensor_act_model_layers_78_residual_post_mlp/mean":316,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_77_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_post_attn/max_abs":121,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_87_mlp/min":4,"train/tensor_act_model_layers_10_residual_post_mlp/norm":90037.5546875,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_43_mlp_down_proj_weight/mean":0.0009918212890625,"train/tensor_act_model_layers_85_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_11_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_79_residual_post_mlp/mean":320,"train/tensor_act_model_layers_52_mlp_up_proj/std":1.84375,"train/tensor_act_model_layers_64_residual_post_mlp/max_abs":260,"train/tensor_param_model_layers_81_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_1_mlp_down_proj_weight/std":0.0341796875,"train/tensor_act_model_layers_78_residual_post_mlp/range":2,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/norm":6.34375,"train/tensor_act_model_layers_5_mlp_down_proj/norm":30009.12109375,"train/tensor_act_model_layers_13_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_65_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_34_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_48_mlp_up_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_89_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_mlp/range":0,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_post_mlp/max":280,"train/tensor_act_model_layers_68_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_residual_pre_attn/min":178,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/max_abs":0.11474609375,"train/tensor_act_model_layers_14_residual_pre_attn/min":55,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_46_residual_post_attn/norm":376684.59375,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_61_residual_post_attn/max_abs":245,"train/tensor_act_model_layers_45_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_14_residual_pre_attn/std":0.267578125,"train/tensor_param_model_layers_61_mlp_down_proj_weight/max":0.13671875,"train/tensor_param_model_layers_25_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_10_residual_post_mlp/max_abs":44.75,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/range":0.244140625,"train/tensor_param_model_layers_53_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/max":0.12353515625,"train/tensor_act_model_layers_95_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_mlp_up_proj/max_abs":4.21875,"train/tensor_act_model_layers_79_residual_post_attn/range":2,"train/tensor_param_model_layers_82_mlp_down_proj_weight/max":0.1396484375,"train/tensor_param_model_layers_30_mlp_down_proj_weight/max":0.1279296875,"train/tensor_act_model_layers_54_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_65_mlp/norm":8192,"train/tensor_param_model_layers_93_mlp_up_proj_weight/mean":0.000728607177734375,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp_down_proj/std":0.3984375,"train/tensor_act_model_layers_66_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_23_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_91_residual_post_mlp/mean":368,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/max":0.134765625,"train/tensor_act_model_layers_85_residual_post_mlp/max_abs":344,"train/tensor_param_model_layers_99_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_84_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_85_residual_post_attn/max":340,"train/tensor_act_model_layers_24_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_21_mlp/range":0,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_78_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_59_mlp_up_proj_weight/min":-0.12890625,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/max":0.11767578125,"train/tensor_act_model_layers_7_residual_post_mlp/mean":32,"train/tensor_act_model_layers_67_mlp_up_proj/max":4.28125,"train/tensor_act_model_layers_34_residual_post_mlp/mean":140,"train/tensor_act_model_layers_49_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_4_residual_post_attn/norm":32734.845703125,"train/tensor_param_model_layers_38_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_73_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_65_mlp_down_proj/max_abs":16.75,"train/tensor_act_model_layers_71_residual_post_attn/norm":581597.3125,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_26_mlp_up_proj/norm":5595.31103515625,"train/tensor_act_model_layers_43_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_70_mlp_up_proj_weight/norm":6.9375,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/norm":7.15625,"train/tensor_param_model_layers_88_mlp_up_proj_weight/max":0.146484375,"train/tensor_act_model_layers_91_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_33_mlp/std":0,"train/tensor_param_model_layers_57_mlp_down_proj_weight/std":0.03662109375,"train/tensor_param_model_layers_2_mlp_up_proj_weight/norm":5.75,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_35_mlp_down_proj_weight/min":-0.12158203125,"train/tensor_act_model_layers_17_residual_post_mlp/min":71,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_95_mlp_down_proj_weight/range":0.3603515625,"train/tensor_param_model_layers_83_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_31_mlp_up_proj/norm":5430.3173828125,"train/tensor_act_model_layers_15_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_15_mlp_up_proj/norm":5405.4189453125,"train/tensor_act_model_layers_45_residual_post_attn/norm":368492.625,"train/tensor_act_model_layers_43_mlp_down_proj/max":16.5,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_24_residual_post_attn/norm":196369.25,"train/tensor_act_model_layers_48_residual_post_attn/max_abs":193,"train/tensor_param_model_layers_32_mlp_down_proj_weight/mean":-0.00012683868408203125,"train/tensor_param_model_layers_2_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_3_mlp_gate_proj/std":1.671875,"train/tensor_act_model_layers_6_mlp/norm":8192,"train/tensor_act_model_layers_6_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_36_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_57_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_57_mlp_up_proj_weight/std":0.036865234375,"train/tensor_act_model_layers_57_mlp_gate_proj/min":-1.359375,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_36_mlp_gate_proj/max_abs":4.8125,"train/tensor_act_model_layers_18_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_34_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_16_mlp_up_proj/min":-5.0625,"train/tensor_act_model_layers_30_residual_pre_attn/min":118.5,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/norm":10.625,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_25_residual_post_attn/max":101,"train/tensor_act_model_layers_88_mlp_down_proj/min":13.9375,"train/tensor_act_model_layers_90_mlp_down_proj/max":17.875,"train/tensor_act_model_layers_67_mlp_down_proj/std":0.515625,"train/tensor_act_model_layers_41_residual_post_mlp/mean":168,"train/tensor_act_model_layers_88_residual_post_mlp/mean":356,"train/tensor_act_model_layers_18_mlp_down_proj/max_abs":15.5625,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_73_mlp/max_abs":4,"train/tensor_act_model_layers_12_mlp/mean":4,"train/tensor_act_model_layers_38_mlp/mean":4,"train/tensor_param_model_layers_76_mlp_down_proj_weight/max":0.1337890625,"train/tensor_param_model_layers_76_mlp_up_proj_weight/norm":7.34375,"train/tensor_act_model_layers_74_mlp_up_proj/mean":-0.2109375,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_79_mlp_up_proj/mean":-0.08203125,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_5_mlp/mean":4,"train/tensor_act_model_layers_21_residual_post_mlp/std":0.359375,"train/tensor_act_model_layers_8_residual_post_mlp/min":35,"train/tensor_act_model_layers_98_mlp_gate_proj/max_abs":5.75,"train/tensor_act_model_layers_11_mlp_up_proj/std":1.9140625,"train/tensor_param_model_layers_21_mlp_up_proj_weight/mean":0.000492095947265625,"train/tensor_act_model_layers_79_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_21_mlp/norm":8192,"train/tensor_act_model_layers_73_mlp_gate_proj/std":1.6171875,"train/tensor_act_model_layers_99_mlp_up_proj/mean":-0.1064453125,"train/tensor_act_model_layers_34_residual_post_attn/norm":278380.8125,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_39_mlp_down_proj/range":1.8125,"train/tensor_act_model_layers_66_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_4_mlp_gate_proj/range":5.4140625,"train/tensor_act_model_layers_19_mlp/max_abs":4,"train/tensor_act_model_layers_8_residual_pre_attn/range":1.625,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_48_mlp/mean":4,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_67_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_39_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_55_mlp_gate_proj/mean":0.44140625,"train/tensor_act_model_layers_99_mlp_down_proj/min":13.875,"train/tensor_param_model_layers_87_mlp_down_proj_weight/mean":0.00128173828125,"train/tensor_act_model_layers_50_mlp_down_proj/norm":30205.529296875,"train/tensor_act_model_layers_91_residual_post_mlp/min":366,"train/tensor_act_model_layers_40_mlp/range":0,"train/tensor_act_model_layers_16_residual_pre_attn/max_abs":65,"train/tensor_act_model_layers_10_residual_post_attn/norm":81845.671875,"train/tensor_act_model_layers_40_mlp_gate_proj/mean":0.416015625,"train/tensor_act_model_layers_71_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_27_mlp_down_proj_weight/std":0.031494140625,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/mean":0.005340576171875,"train/tensor_param_model_layers_92_mlp_down_proj_weight/std":0.048583984375,"train/tensor_act_model_layers_56_mlp/max_abs":4,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_5_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_32_mlp_down_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_72_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_14_mlp_up_proj/norm":5557.91845703125,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_76_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_53_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_43_mlp_gate_proj/mean":0.5625,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/norm":6.34375,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_48_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_53_mlp_up_proj/range":9.65625,"train/tensor_act_model_layers_42_residual_post_mlp/max_abs":173,"train/tensor_act_model_layers_45_residual_pre_attn/max_abs":181,"train/tensor_act_model_layers_69_mlp/range":0,"train/tensor_param_model_layers_45_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_17_mlp_up_proj/range":8.84375,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_91_mlp_down_proj/range":6.125,"train/tensor_param_model_layers_90_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/mean":0.003509521484375,"train/tensor_act_model_layers_83_mlp_up_proj/mean":-0.04052734375,"train/tensor_act_model_layers_30_mlp_gate_proj/min":-1.28125,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_51_mlp_down_proj/mean":14.875,"train/tensor_act_model_layers_45_residual_post_attn/mean":180,"train/tensor_act_model_layers_18_residual_pre_attn/min":71,"train/tensor_act_model_layers_30_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_24_residual_post_attn/mean":96,"train/tensor_act_model_layers_3_residual_post_attn/max":12.625,"train/tensor_act_model_layers_44_mlp_down_proj/range":1.9375,"train/tensor_act_model_layers_44_residual_post_mlp/range":3,"train/tensor_act_model_layers_18_residual_pre_attn/norm":147258.328125,"train/tensor_act_model_layers_40_mlp_gate_proj/max_abs":4.84375,"train/tensor_param_model_layers_48_mlp_down_proj_weight/min":-0.125,"train/tensor_act_model_layers_79_residual_post_attn/max_abs":316,"train/tensor_act_model_layers_67_mlp_down_proj/range":3.4375,"train/tensor_act_model_layers_4_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/mean":0.0037841796875,"train/tensor_act_model_layers_19_mlp/range":0,"train/tensor_act_model_layers_69_mlp_gate_proj/std":1.6328125,"train/tensor_param_model_layers_26_mlp_up_proj_weight/min":-0.11279296875,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_55_mlp/max":4,"train/tensor_act_model_layers_51_mlp/range":0,"train/tensor_act_model_layers_49_residual_post_attn/max":197,"train/tensor_act_model_layers_96_residual_post_mlp/range":2,"train/tensor_act_model_layers_42_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_61_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_71_residual_post_attn/mean":284,"train/tensor_param_model_layers_4_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_18_residual_pre_attn/std":0.34375,"train/tensor_act_model_layers_22_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_59_residual_pre_attn/norm":483100.78125,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_6_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_31_mlp_up_proj/max_abs":4.53125,"train/tensor_act_model_layers_4_residual_pre_attn/min":15.375,"train/tensor_act_model_layers_11_residual_post_mlp/range":2,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/min":-0.125,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/max_abs":0.12109375,"train/tensor_act_model_layers_88_mlp/range":0,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_97_mlp_gate_proj_weight/norm":10.375,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_97_mlp/range":0,"train/tensor_act_model_layers_9_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_37_residual_post_attn/min":146,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/mean":0.0042724609375,"train/tensor_act_model_layers_3_mlp/std":0,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_78_mlp/range":0,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/range":0.24462890625,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/max_abs":0.134765625,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_73_mlp_down_proj_weight/norm":6.90625,"train/tensor_act_model_layers_27_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_84_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_25_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_80_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_67_mlp_up_proj_weight/range":0.275390625,"train/tensor_act_model_layers_58_mlp/range":0,"train/tensor_act_model_layers_74_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_83_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_residual_post_attn/mean":184,"train/tensor_act_model_layers_2_mlp_down_proj/max_abs":17.375,"train/tensor_act_model_layers_34_mlp/mean":4,"train/tensor_param_model_layers_49_mlp_up_proj_weight/norm":6.46875,"train/tensor_act_model_layers_65_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_1_residual_post_attn/norm":8178.431640625,"train/tensor_act_model_layers_52_mlp/range":0,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_96_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_mlp/mean":4,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/range":0.3056640625,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_35_mlp/mean":4,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_39_residual_pre_attn/range":3,"train/tensor_act_model_layers_17_residual_post_mlp/mean":72,"train/tensor_act_model_layers_37_residual_post_mlp/norm":311148.71875,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_60_residual_post_attn/range":3,"train/tensor_param_model_layers_65_mlp_up_proj_weight/norm":6.875,"train/tensor_param_model_layers_75_mlp_up_proj_weight/range":0.26171875,"train/tensor_act_model_layers_20_residual_post_mlp/range":2.5,"train/tensor_param_model_layers_73_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/range":0.263671875,"train/tensor_param_model_layers_0_mlp_down_proj_weight/norm":11.25,"train/tensor_param_model_layers_6_mlp_down_proj_weight/min":-0.1015625,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_19_mlp_gate_proj/std":1.6875,"train/tensor_act_model_layers_68_mlp/range":0,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_31_residual_post_mlp/max_abs":129,"train/tensor_param_model_layers_90_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_17_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp_up_proj/max":4.28125,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_10_mlp_up_proj/std":1.9453125,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/std":0.031982421875,"train/tensor_act_model_layers_4_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_43_mlp_down_proj/std":0.443359375,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/norm":8.75,"train/tensor_act_model_layers_7_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_43_mlp_up_proj/norm":5354.46533203125,"train/tensor_param_model_layers_33_mlp_down_proj_weight/std":0.031982421875,"train/tensor_act_model_layers_64_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_37_mlp_down_proj_weight/mean":0.00020122528076171875,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_26_mlp_gate_proj/mean":0.6640625,"train/tensor_param_model_layers_52_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_33_residual_post_attn/range":3,"train/tensor_param_model_layers_42_mlp_up_proj_weight/range":0.2568359375,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/range":0.271484375,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_80_mlp_up_proj_weight/max_abs":0.1513671875,"train/tensor_grad_model_layers_5_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_10_residual_pre_attn/norm":81849.6953125,"train/tensor_act_model_layers_58_mlp_up_proj/max_abs":5.1875,"train/tensor_act_model_layers_83_mlp_down_proj/norm":31061.76953125,"train/tensor_act_model_layers_42_mlp/max_abs":4,"train/tensor_act_model_layers_13_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_61_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_47_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_21_mlp_up_proj/max_abs":4.84375,"train/tensor_act_model_layers_75_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_11_mlp_up_proj_weight/range":0.22998046875,"train/tensor_act_model_layers_72_residual_post_attn/min":286,"train/tensor_act_model_layers_86_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_0_residual_post_mlp/max_abs":4.3125,"train/tensor_param_model_layers_90_mlp_up_proj_weight/norm":8.625,"train/tensor_param_model_layers_3_mlp_down_proj_weight/std":0.02978515625,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_36_mlp_up_proj/min":-4.4375,"train/tensor_act_model_layers_26_mlp_gate_proj/range":6.34375,"train/tensor_act_model_layers_84_mlp_up_proj/mean":0.06689453125,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_15_mlp/max":4,"train/tensor_act_model_layers_25_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_12_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_93_residual_pre_attn/range":2,"train/tensor_grad_model_layers_50_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_73_residual_post_attn/min":290,"train/tensor_act_model_layers_74_mlp_up_proj/std":1.8203125,"train/tensor_param_model_layers_67_mlp_down_proj_weight/norm":6.875,"train/tensor_act_model_layers_4_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_39_mlp_up_proj_weight/min":-0.12255859375,"train/tensor_act_model_layers_73_residual_post_mlp/norm":606173.3125,"train/tensor_act_model_layers_53_residual_post_mlp/max":217,"train/tensor_act_model_layers_38_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_32_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_50_mlp_down_proj/mean":14.75,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_72_residual_post_mlp/norm":597981.3125,"train/tensor_act_model_layers_67_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_86_residual_pre_attn/min":342,"train/tensor_act_model_layers_48_mlp_gate_proj/range":6.546875,"train/tensor_param_model_layers_98_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_73_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/max":0.1357421875,"train/tensor_param_model_layers_32_mlp_up_proj_weight/min":-0.12158203125,"train/tensor_param_model_layers_75_mlp_up_proj_weight/min":-0.1298828125,"train/tensor_act_model_layers_93_mlp_up_proj/std":1.828125,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/norm":6.78125,"train/tensor_param_model_layers_70_mlp_down_proj_weight/std":0.0380859375,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_54_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_43_mlp_up_proj/std":1.8515625,"train/tensor_act_model_layers_17_mlp_gate_proj/max":4.625,"train/tensor_param_model_layers_53_mlp_down_proj_weight/std":0.0361328125,"train/tensor_act_model_layers_53_mlp_gate_proj/max":5.0625,"train/tensor_act_model_layers_36_residual_pre_attn/range":3,"train/tensor_act_model_layers_6_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/min":-0.12890625,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_77_mlp/norm":8192,"train/tensor_act_model_layers_83_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_1_residual_post_attn/std":0.1123046875,"train/tensor_act_model_layers_44_mlp/norm":8192,"train/tensor_act_model_layers_48_residual_post_mlp/max":197,"train/tensor_act_model_layers_3_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_residual_post_attn/norm":548829.3125,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_13_mlp_down_proj/range":2.0625,"train/tensor_param_model_layers_11_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_mlp_up_proj/norm":5356.2119140625,"train/tensor_act_model_layers_71_residual_post_attn/max_abs":284,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_93_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_44_mlp_up_proj_weight/max_abs":0.1259765625,"train/tensor_act_model_layers_10_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_residual_post_attn/norm":474908.8125,"train/tensor_param_model_layers_27_mlp_up_proj_weight/max_abs":0.12109375,"train/tensor_param_model_layers_74_mlp_up_proj_weight/mean":-0.00164794921875,"train/tensor_act_model_layers_44_mlp_gate_proj/std":1.65625,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/range":0.31640625,"train/tensor_grad_model_layers_3_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_61_residual_post_mlp/range":3,"train/tensor_act_model_layers_13_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_78_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/range":0.376953125,"train/tensor_act_model_layers_1_mlp_down_proj/mean":16.125,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_residual_post_mlp/min":334,"train/tensor_act_model_layers_57_mlp/max_abs":4,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_68_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_59_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_residual_pre_attn/mean":196,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_76_mlp_up_proj/min":-4.34375,"train/tensor_act_model_layers_72_mlp_gate_proj/norm":4844.53173828125,"train/tensor_act_model_layers_71_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_57_mlp_down_proj_weight/max_abs":0.1298828125,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_4_residual_post_attn/mean":16,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_42_mlp_down_proj/max":16.125,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_28_mlp/std":0,"train/tensor_act_model_layers_89_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_21_mlp_up_proj_weight/std":0.03125,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_65_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_61_mlp/norm":8192,"train/tensor_act_model_layers_82_mlp/norm":8192,"train/tensor_act_model_layers_18_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_68_residual_post_mlp/max_abs":276,"train/tensor_param_model_layers_71_mlp_down_proj_weight/mean":0.0007171630859375,"train/tensor_act_model_layers_33_mlp_gate_proj/min":-1.3046875,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_33_mlp_down_proj_weight/range":0.24853515625,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp_gate_proj/max":5.0625,"train/tensor_act_model_layers_15_mlp_gate_proj/norm":5174.7099609375,"train/tensor_act_model_layers_9_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_86_mlp_down_proj/std":0.5703125,"train/tensor_act_model_layers_12_residual_post_mlp/mean":52,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/max":0.126953125,"train/tensor_param_model_layers_30_mlp_up_proj_weight/range":0.23095703125,"train/tensor_param_model_layers_7_mlp_down_proj_weight/max_abs":0.0986328125,"train/tensor_param_model_layers_60_mlp_up_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_61_mlp/max_abs":4,"train/tensor_grad_model_layers_82_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_65_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_94_mlp_down_proj/range":4.9375,"train/tensor_act_model_layers_27_mlp_down_proj/range":1.6875,"train/tensor_act_model_layers_83_mlp_up_proj/range":8.4375,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/max_abs":0.142578125,"train/tensor_param_model_layers_98_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_2_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_11_residual_post_attn/min":43,"train/tensor_param_model_layers_52_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_residual_post_mlp/std":0.416015625,"train/tensor_param_model_layers_12_mlp_down_proj_weight/norm":5.46875,"train/tensor_act_model_layers_99_mlp/norm":8192,"train/tensor_act_model_layers_70_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_39_mlp_up_proj/max":4.1875,"train/tensor_act_model_layers_79_residual_post_attn/mean":316,"train/tensor_act_model_layers_51_mlp_down_proj/norm":30479.9375,"train/tensor_param_model_layers_35_mlp_gate_proj_weight/mean":0.004425048828125,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_mlp_gate_proj/max_abs":4.0625,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/max":0.12451171875,"train/tensor_act_model_layers_49_residual_pre_attn/max":197,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_4_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_15_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_residual_post_attn/max_abs":16.625,"train/tensor_act_model_layers_68_mlp_down_proj/range":2.9375,"train/tensor_act_model_layers_74_mlp_down_proj/max_abs":17,"train/tensor_act_model_layers_15_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_43_mlp_up_proj_weight/range":0.24853515625,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_2_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_mlp_gate_proj/norm":5022.44482421875,"train/tensor_act_model_layers_80_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_14_mlp_up_proj_weight/min":-0.10986328125,"train/tensor_act_model_layers_22_mlp_gate_proj/mean":0.5859375,"train/tensor_act_model_layers_29_mlp_down_proj/max":16.625,"train/tensor_act_model_layers_79_mlp_down_proj/min":14.25,"train/tensor_param_model_layers_58_mlp_up_proj_weight/std":0.037353515625,"train/tensor_act_model_layers_35_residual_pre_attn/max_abs":141,"train/tensor_act_model_layers_21_residual_pre_attn/norm":171793.453125,"train/tensor_act_model_layers_10_mlp/std":0,"train/tensor_param_model_layers_14_mlp_down_proj_weight/range":0.22265625,"train/tensor_param_model_layers_77_mlp_up_proj_weight/max":0.1357421875,"train/tensor_param_model_layers_33_mlp_up_proj_weight/min":-0.12353515625,"train/tensor_act_model_layers_2_residual_post_mlp/min":11.5,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_48_mlp_down_proj_weight/max":0.125,"train/tensor_param_model_layers_11_mlp_up_proj_weight/std":0.030029296875,"train/tensor_act_model_layers_49_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/range":0.2578125,"train/tensor_act_model_layers_10_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_18_mlp_up_proj_weight/min":-0.10205078125,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_78_mlp_gate_proj/max":4.5625,"train/tensor_act_model_layers_39_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_11_residual_post_mlp/norm":98169.8359375,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_20_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/norm":7.1875,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_9_mlp_up_proj_weight/max":0.10400390625,"train/tensor_act_model_layers_18_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_residual_pre_attn/max":280,"train/tensor_act_model_layers_4_mlp/max_abs":4,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_51_residual_pre_attn/max_abs":205,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_26_mlp_down_proj_weight/range":0.23486328125,"train/tensor_act_model_layers_40_residual_pre_attn/range":3,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_21_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_47_mlp_down_proj_weight/max_abs":0.1259765625,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_59_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_66_mlp_down_proj/max":17,"train/tensor_act_model_layers_65_residual_post_mlp/min":262,"train/tensor_param_model_layers_80_mlp_up_proj_weight/min":-0.1337890625,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/max_abs":0.12890625,"train/tensor_act_model_layers_45_mlp/range":0,"train/tensor_act_model_layers_26_mlp_gate_proj/max":5.03125,"train/tensor_act_model_layers_46_mlp_up_proj/min":-4.53125,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_96_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_84_mlp_down_proj/std":0.625,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_62_mlp_up_proj_weight/std":0.03759765625,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/std":0.032958984375,"train/tensor_act_model_layers_51_residual_post_mlp/min":206,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_45_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_75_mlp_up_proj/max":4.6875,"train/tensor_act_model_layers_10_mlp_up_proj/max_abs":4.0625,"train/tensor_param_model_layers_58_mlp_down_proj_weight/std":0.036865234375,"train/tensor_act_model_layers_7_mlp/norm":8192,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/range":0.2548828125,"train/tensor_grad_model_layers_77_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_50_mlp/max_abs":4,"train/tensor_act_model_layers_56_residual_post_mlp/max_abs":229,"train/tensor_act_model_layers_54_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp/std":0,"train/tensor_act_model_layers_38_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_23_residual_post_mlp/range":2.5,"train/tensor_act_model_layers_45_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_mlp_gate_proj/min":-1.78125,"train/tensor_act_model_layers_57_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_12_residual_post_attn/mean":48,"train/tensor_act_model_layers_87_residual_post_mlp/min":350,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_87_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_54_residual_post_mlp/max":221,"train/tensor_act_model_layers_79_mlp/mean":4,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_76_mlp_down_proj_weight/min":-0.1357421875,"train/tensor_act_model_layers_85_mlp_up_proj/max_abs":4.90625,"train/tensor_act_model_layers_15_residual_pre_attn/mean":60,"train/tensor_act_model_layers_12_mlp_up_proj/mean":-0.08349609375,"train/tensor_act_model_layers_59_residual_post_mlp/max_abs":241,"train/tensor_act_model_layers_26_residual_pre_attn/range":2.5,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_86_mlp/range":0,"train/tensor_act_model_layers_79_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_10_mlp_down_proj/mean":14.6875,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/std":0.031494140625,"train/tensor_grad_model_layers_60_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_50_mlp_gate_proj/min":-1.2578125,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_56_mlp/range":0,"train/tensor_param_model_layers_50_mlp_up_proj_weight/min":-0.125,"train/tensor_act_model_layers_42_residual_pre_attn/min":166,"train/tensor_param_model_layers_6_mlp_up_proj_weight/range":0.22802734375,"train/tensor_act_model_layers_71_residual_post_mlp/min":286,"train/tensor_act_model_layers_19_residual_post_attn/range":2.5,"train/tensor_act_model_layers_58_residual_pre_attn/max_abs":233,"train/tensor_act_model_layers_40_residual_post_mlp/norm":335724.6875,"train/tensor_act_model_layers_73_residual_pre_attn/range":2,"train/tensor_act_model_layers_23_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_12_mlp_up_proj_weight/mean":-0.000652313232421875,"train/tensor_param_model_layers_2_mlp_up_proj_weight/std":0.03173828125,"train/tensor_act_model_layers_18_mlp_gate_proj/norm":5109.884765625,"train/tensor_act_model_layers_42_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_40_residual_post_attn/min":158,"train/tensor_grad_model_layers_92_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_93_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_residual_post_mlp/range":2,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_35_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_15_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_51_mlp_down_proj/range":2.3125,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_40_mlp_up_proj/max_abs":4.8125,"train/tensor_act_model_layers_98_mlp_down_proj/range":6.0625,"train/tensor_act_model_layers_5_residual_pre_attn/mean":20,"train/tensor_act_model_layers_70_mlp_gate_proj/std":1.65625,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/min":-0.2158203125,"train/tensor_act_model_layers_98_mlp_gate_proj/norm":5310.04345703125,"train/tensor_act_model_layers_70_residual_post_attn/max_abs":280,"train/tensor_act_model_layers_11_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_residual_post_attn/max_abs":221,"train/tensor_param_model_layers_72_mlp_gate_proj_weight/mean":0.003265380859375,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/std":0.046875,"train/tensor_param_model_layers_38_mlp_up_proj_weight/min":-0.130859375,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_82_mlp/range":0,"train/tensor_act_model_layers_65_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/std":0.0311279296875,"train/tensor_act_model_layers_92_mlp/norm":8192,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_46_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_80_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_residual_post_attn/mean":276,"train/tensor_act_model_layers_45_residual_post_mlp/mean":184,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/range":0.24853515625,"train/tensor_act_model_layers_93_mlp/std":0,"train/tensor_param_model_layers_54_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_24_mlp/range":0,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_56_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/range":0.25439453125,"train/tensor_act_model_layers_58_mlp_down_proj/max":17.125,"train/tensor_act_model_layers_0_mlp/norm":8187.9267578125,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_38_mlp_gate_proj/min":-1.2734375,"train/tensor_act_model_layers_22_mlp/std":0,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/range":0.24072265625,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/std":0.03466796875,"train/tensor_act_model_layers_35_mlp_up_proj/std":1.859375,"train/tensor_act_model_layers_29_mlp/norm":8192,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_26_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_6_residual_post_mlp/max_abs":28.625,"train/tensor_act_model_layers_12_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_27_mlp_gate_proj/min":-1.28125,"train/tensor_act_model_layers_72_residual_post_attn/norm":589789.25,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/max":0.125,"train/tensor_act_model_layers_27_residual_post_mlp/mean":112,"train/tensor_act_model_layers_77_mlp/std":0,"train/tensor_act_model_layers_90_mlp_down_proj/std":0.73046875,"train/tensor_act_model_layers_39_mlp/min":4,"train/tensor_act_model_layers_20_mlp_up_proj/min":-4.90625,"train/tensor_act_model_layers_60_mlp_up_proj/mean":-0.16015625,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_90_residual_post_mlp/norm":745437.25,"train/tensor_act_model_layers_18_mlp/mean":4,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_70_mlp/mean":4,"train/tensor_act_model_layers_39_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/mean":0.00445556640625,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/range":0.25390625,"train/tensor_act_model_layers_0_mlp_up_proj/max":7.90625,"train/tensor_grad_model_layers_63_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_14_mlp_gate_proj/norm":5258.02099609375,"train/tensor_param_model_layers_55_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_62_mlp_down_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_49_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_47_mlp_gate_proj/max_abs":4.625,"train/tensor_act_model_layers_12_mlp_gate_proj/std":1.7421875,"train/tensor_act_model_layers_74_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_73_mlp_up_proj_weight/mean":0.000896453857421875,"train/tensor_act_model_layers_4_residual_pre_attn/norm":32742.162109375,"train/tensor_act_model_layers_39_residual_post_attn/min":154,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_26_residual_pre_attn/max_abs":105,"train/tensor_act_model_layers_80_residual_pre_attn/range":2,"train/tensor_param_model_layers_78_mlp_up_proj_weight/min":-0.130859375,"train/tensor_act_model_layers_39_residual_post_mlp/norm":327532.6875,"train/tensor_act_model_layers_28_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_75_mlp/max":4,"train/tensor_act_model_layers_17_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_22_mlp_up_proj_weight/std":0.03125,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/min":-0.1279296875,"train/tensor_act_model_layers_6_mlp_gate_proj/max_abs":4.71875,"train/tensor_param_model_layers_23_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_60_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/max":0.125,"train/tensor_act_model_layers_99_mlp/std":0,"train/tensor_act_model_layers_87_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_38_mlp_down_proj_weight/mean":0.0015106201171875,"train/tensor_grad_model_layers_2_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_47_mlp/min":4,"train/tensor_grad_model_layers_12_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_64_mlp_down_proj_weight/mean":-0.0013885498046875,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/mean":0.004241943359375,"train/tensor_act_model_layers_75_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_28_residual_post_mlp/mean":116,"train/tensor_act_model_layers_90_residual_post_attn/min":358,"train/tensor_act_model_layers_15_residual_post_attn/std":0.31640625,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_10_mlp_down_proj_weight/range":0.2119140625,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_97_mlp_down_proj_weight/mean":0.000102996826171875,"train/tensor_param_model_layers_96_mlp_up_proj_weight/range":0.3642578125,"train/tensor_act_model_layers_32_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_87_mlp_down_proj/max":18.25,"train/tensor_act_model_layers_57_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/min":-0.12158203125,"train/tensor_act_model_layers_3_residual_pre_attn/std":0.1298828125,"train/tensor_act_model_layers_32_mlp_up_proj/mean":-0.041259765625,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/mean":0.004730224609375,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_58_mlp_up_proj_weight/mean":0.000576019287109375,"train/tensor_param_model_layers_61_mlp_down_proj_weight/mean":-0.000858306884765625,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_11_mlp_down_proj_weight/norm":5.3125,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/std":0.0311279296875,"train/tensor_act_model_layers_86_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_post_attn/max_abs":276,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_62_mlp_gate_proj/mean":0.515625,"train/tensor_act_model_layers_22_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/std":0.05419921875,"train/tensor_act_model_layers_54_mlp/range":0,"train/tensor_act_model_layers_56_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_45_mlp_gate_proj/range":6.1328125,"train/tensor_act_model_layers_46_mlp_up_proj/max_abs":4.53125,"train/tensor_act_model_layers_49_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_61_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_90_mlp_gate_proj/range":6.0078125,"train/tensor_act_model_layers_73_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_60_residual_post_mlp/mean":244,"train/tensor_param_model_layers_16_mlp_up_proj_weight/std":0.0308837890625,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/range":0.2646484375,"train/tensor_act_model_layers_29_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_9_residual_post_attn/std":0.2265625,"train/tensor_param_model_layers_32_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_8_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/range":0.25048828125,"train/tensor_act_model_layers_20_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_59_mlp_gate_proj/std":1.6171875,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_26_mlp_up_proj/range":9.21875,"train/tensor_param_model_layers_29_mlp_up_proj_weight/std":0.031982421875,"train/tensor_act_model_layers_89_mlp_down_proj/min":14.0625,"train/tensor_act_model_layers_95_mlp_up_proj/std":1.7734375,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/mean":0.0050048828125,"train/tensor_act_model_layers_83_mlp_down_proj/mean":15.125,"train/tensor_act_model_layers_77_residual_post_mlp/max_abs":312,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_82_residual_post_attn/max_abs":328,"train/tensor_param_model_layers_32_mlp_down_proj_weight/range":0.248046875,"train/tensor_act_model_layers_78_mlp_up_proj/range":9.03125,"train/tensor_act_model_layers_28_mlp_up_proj/max_abs":4.8125,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_52_residual_post_attn/mean":208,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/range":0.419921875,"train/tensor_act_model_layers_45_mlp/min":4,"train/tensor_act_model_layers_3_residual_post_mlp/mean":16,"train/tensor_param_model_layers_62_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_7_mlp_down_proj/norm":30043.90625,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/std":0.040771484375,"train/tensor_act_model_layers_96_mlp/mean":4,"train/tensor_param_model_layers_70_mlp_up_proj_weight/std":0.038330078125,"train/tensor_act_model_layers_98_mlp/min":4,"train/tensor_param_model_layers_7_mlp_up_proj_weight/std":0.02978515625,"train/tensor_act_model_layers_29_mlp_up_proj/norm":5608.06689453125,"train/tensor_act_model_layers_93_mlp_gate_proj/norm":5153.29345703125,"train/tensor_act_model_layers_0_residual_post_attn/std":0.09130859375,"train/tensor_act_model_layers_3_mlp_up_proj/norm":5446.20751953125,"train/tensor_act_model_layers_77_mlp_down_proj/max_abs":19,"train/tensor_act_model_layers_40_mlp_gate_proj/min":-1.265625,"train/tensor_grad_model_layers_63_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_27_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/std":0.042724609375,"train/tensor_act_model_layers_52_mlp/max_abs":4,"train/tensor_act_model_layers_65_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/std":0.0517578125,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_74_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_mlp/range":0,"train/tensor_act_model_layers_37_mlp_gate_proj/std":1.6875,"train/tensor_act_model_layers_15_residual_post_attn/max":61,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/mean":0.00335693359375,"train/tensor_param_model_layers_60_mlp_up_proj_weight/mean":-0.001251220703125,"train/tensor_act_model_layers_47_residual_post_attn/mean":188,"train/tensor_param_model_layers_81_mlp_down_proj_weight/range":0.28125,"train/tensor_param_model_layers_9_mlp_down_proj_weight/norm":5.34375,"train/tensor_param_model_layers_89_mlp_up_proj_weight/max":0.15625,"train/tensor_act_model_layers_52_mlp_down_proj/min":14.25,"train/tensor_act_model_layers_27_mlp_gate_proj/max_abs":4.84375,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_92_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_43_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_69_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_24_residual_pre_attn/max":97,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/mean":0.003570556640625,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_residual_pre_attn/range":3,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_90_mlp/max_abs":4,"train/tensor_act_model_layers_95_mlp/std":0,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_mlp_gate_proj/max":4.4375,"train/tensor_act_model_layers_26_residual_post_mlp/max":109,"train/tensor_act_model_layers_10_mlp_down_proj/norm":30041.2578125,"train/tensor_act_model_layers_49_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_73_residual_pre_attn/max_abs":292,"train/tensor_act_model_layers_0_mlp_up_proj/min":-7.53125,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_28_mlp_down_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_92_residual_post_attn/max_abs":368,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/max":0.15625,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp_up_proj/range":9.1875,"train/tensor_act_model_layers_21_residual_post_attn/range":2.5,"train/tensor_act_model_layers_5_mlp_gate_proj/min":-1.328125,"train/tensor_act_model_layers_75_mlp_up_proj/norm":5336.88134765625,"train/tensor_act_model_layers_21_mlp_up_proj/norm":5608.6796875,"train/tensor_act_model_layers_78_mlp_gate_proj/range":5.8203125,"train/tensor_act_model_layers_81_residual_pre_attn/range":2,"train/tensor_act_model_layers_66_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_52_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_63_residual_post_mlp/min":254,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_99_mlp/max":4,"train/tensor_act_model_layers_90_residual_post_mlp/min":362,"train/tensor_param_model_layers_68_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_93_mlp/max_abs":4,"train/tensor_act_model_layers_84_mlp/min":4,"train/tensor_act_model_layers_89_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_79_mlp_up_proj_weight/std":0.041748046875,"train/tensor_param_model_layers_71_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_51_mlp_down_proj_weight/range":0.2548828125,"train/tensor_act_model_layers_49_mlp/range":0,"train/tensor_act_model_layers_81_mlp/max_abs":4,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_78_mlp_up_proj_weight/mean":0.00165557861328125,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/max":0.125,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_22_mlp_down_proj/max":15.5,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_71_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_71_residual_post_mlp/max_abs":288,"train/tensor_act_model_layers_92_mlp_down_proj/std":0.765625,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/range":0.2626953125,"train/tensor_act_model_layers_30_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_84_residual_pre_attn/range":2,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/range":0.25390625,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_48_mlp_down_proj_weight/std":0.03515625,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_47_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_43_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_52_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_43_mlp_up_proj_weight/mean":0.000209808349609375,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/max_abs":0.1298828125,"train/tensor_param_model_layers_81_mlp_up_proj_weight/range":0.28515625,"train/tensor_act_model_layers_58_residual_post_attn/std":0.416015625,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_75_residual_pre_attn/max":300,"train/tensor_act_model_layers_11_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_82_residual_post_mlp/min":330,"train/tensor_act_model_layers_50_mlp_up_proj/max_abs":4.21875,"train/tensor_act_model_layers_57_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_99_mlp_down_proj/max_abs":21.875,"train/tensor_param_model_layers_53_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_86_residual_pre_attn/norm":704477.25,"train/tensor_act_model_layers_64_mlp_down_proj/mean":14.875,"train/tensor_act_model_layers_19_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_11_mlp/max":4,"train/tensor_act_model_layers_12_mlp_gate_proj/min":-1.3203125,"train/tensor_act_model_layers_77_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_83_residual_post_mlp/max_abs":336,"train/tensor_act_model_layers_9_mlp_down_proj/min":14.0625,"train/tensor_act_model_layers_13_mlp_up_proj/min":-4.5625,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/std":0.034423828125,"train/tensor_param_model_layers_66_mlp_down_proj_weight/range":0.2568359375,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_17_mlp/max":4,"train/tensor_act_model_layers_15_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_82_mlp/max_abs":4,"train/tensor_act_model_layers_83_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_mlp/min":4,"train/tensor_param_model_layers_81_mlp_up_proj_weight/max_abs":0.1435546875,"train/tensor_act_model_layers_37_mlp_gate_proj/min":-1.2890625,"train/tensor_param_model_layers_50_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_mlp_down_proj/norm":30630.8671875,"train/tensor_act_model_layers_15_mlp_gate_proj/range":5.6640625,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_mlp/max":4,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_27_mlp_down_proj_weight/min":-0.125,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_36_mlp_up_proj_weight/min":-0.125,"train/tensor_act_model_layers_24_residual_post_mlp/std":0.359375,"train/tensor_param_model_layers_92_mlp_down_proj_weight/min":-0.15234375,"train/tensor_act_model_layers_8_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_97_residual_post_attn/norm":794589.25,"train/tensor_act_model_layers_52_mlp_gate_proj/norm":5180.44677734375,"train/tensor_act_model_layers_5_mlp_down_proj/std":0.421875,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/norm":51200,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/norm":6.375,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/max":0.1435546875,"train/tensor_grad_model_layers_30_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_37_mlp_up_proj_weight/norm":6.125,"train/tensor_act_model_layers_5_residual_pre_attn/max":20.625,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/max":0.1298828125,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/range":0.3017578125,"train/tensor_act_model_layers_6_mlp_down_proj/range":2,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/max_abs":0.1875,"train/tensor_act_model_layers_42_mlp/std":0,"train/tensor_act_model_layers_40_mlp_gate_proj/std":1.6953125,"train/tensor_act_model_layers_55_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_49_mlp_up_proj_weight/max_abs":0.1328125,"train/tensor_act_model_layers_85_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_37_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_93_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_59_mlp_gate_proj/min":-1.28125,"train/tensor_act_model_layers_34_residual_post_attn/range":3,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/range":0.24169921875,"train/tensor_act_model_layers_45_residual_post_mlp/max":185,"train/tensor_act_model_layers_69_residual_pre_attn/min":274,"train/tensor_act_model_layers_14_residual_post_mlp/max":61,"train/tensor_act_model_layers_4_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_56_mlp_gate_proj_weight/range":0.2578125,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_97_mlp_up_proj/max_abs":5.40625,"train/tensor_act_model_layers_91_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_mlp_down_proj/range":2,"train/tensor_param_model_layers_8_mlp_gate_proj_weight/max":0.10400390625,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/min":-1880,"train/tensor_param_model_layers_84_mlp_up_proj_weight/range":0.29296875,"train/tensor_act_model_layers_6_residual_pre_attn/mean":24,"train/tensor_act_model_layers_8_residual_pre_attn/norm":65435.83203125,"train/tensor_act_model_layers_20_residual_pre_attn/max_abs":81,"train/tensor_act_model_layers_0_mlp_up_proj/std":1.8984375,"train/tensor_act_model_layers_24_mlp_up_proj/norm":5579.4013671875,"train/tensor_act_model_layers_24_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_44_mlp_up_proj_weight/mean":2.8014183044433594e-05,"train/tensor_act_model_layers_30_mlp_up_proj/mean":-0.1474609375,"train/tensor_param_model_layers_1_mlp_down_proj_weight/max":0.125,"train/tensor_act_model_layers_32_mlp_down_proj/std":0.345703125,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_20_mlp_down_proj_weight/range":0.2490234375,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_69_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_41_mlp_up_proj/std":1.8828125,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/max":0.12060546875,"train/tensor_act_model_layers_35_mlp_down_proj/std":0.404296875,"train/tensor_act_model_layers_58_residual_post_attn/max":233,"train/tensor_act_model_layers_48_mlp/min":4,"train/tensor_act_model_layers_48_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_mlp/mean":4,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/max":0.11669921875,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_98_mlp_up_proj_weight/max":0.16796875,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_89_mlp_gate_proj/norm":5086.66455078125,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/mean":0,"train_samples_per_second":607.804,"train/tensor_act_model_layers_46_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/max":0.1728515625,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_25_mlp_down_proj/max_abs":16.375,"train/tensor_act_model_layers_71_mlp_gate_proj/std":1.6484375,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_95_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_73_mlp_up_proj/min":-4.78125,"train/tensor_act_model_layers_18_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_18_mlp_down_proj/max":15.5625,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_78_mlp_down_proj_weight/range":0.2783203125,"train/tensor_act_model_layers_95_residual_post_mlp/mean":384,"train/tensor_act_model_layers_70_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_86_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_80_residual_pre_attn/max_abs":320,"train/tensor_act_model_layers_33_mlp_up_proj/max":4.75,"train/tensor_act_model_layers_4_mlp_gate_proj/max_abs":3.953125,"train/tensor_act_model_layers_7_mlp_down_proj/max":16.25,"train/tensor_param_model_layers_73_mlp_up_proj_weight/max":0.1357421875,"train/tensor_param_model_layers_63_mlp_up_proj_weight/mean":-0.00032806396484375,"train/tensor_param_model_layers_90_mlp_up_proj_weight/std":0.0478515625,"train/tensor_act_model_layers_82_residual_post_mlp/max_abs":332,"train/tensor_act_model_layers_82_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_9_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_16_residual_post_mlp/mean":68,"train/tensor_param_model_layers_10_mlp_up_proj_weight/min":-0.09765625,"train/tensor_param_model_layers_74_mlp_up_proj_weight/std":0.039306640625,"train/tensor_act_model_layers_46_mlp_gate_proj/mean":0.5703125,"train/tensor_act_model_layers_74_residual_post_attn/mean":296,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_86_mlp/std":0,"train/tensor_act_model_layers_33_mlp_down_proj/max_abs":16.5,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_79_mlp_gate_proj_weight/mean":0.005218505859375,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_38_residual_post_mlp/norm":319340.71875,"train/tensor_act_model_layers_96_mlp_down_proj/range":6.25,"train/tensor_act_model_layers_96_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_28_residual_post_attn/norm":229113.265625,"train/tensor_param_model_layers_70_mlp_down_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_93_residual_post_attn/min":370,"train/tensor_act_model_layers_21_mlp_up_proj/min":-4.8125,"train/tensor_act_model_layers_55_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_42_mlp_gate_proj_weight/std":0.03466796875,"train/tensor_param_model_layers_53_mlp_down_proj_weight/min":-0.1298828125,"train/tensor_param_model_layers_25_mlp_up_proj_weight/max":0.1171875,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/mean":0.00457763671875,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_18_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_84_residual_pre_attn/max_abs":336,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/range":0.2509765625,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/norm":6.34375,"train/tensor_param_model_layers_9_mlp_up_proj_weight/norm":5.4375,"train/tensor_act_model_layers_22_residual_post_mlp/max_abs":93,"train/tensor_act_model_layers_81_mlp/std":0,"train/tensor_act_model_layers_97_mlp_down_proj/std":0.86328125,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_11_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_3_residual_post_mlp/norm":32742.162109375,"train/tensor_act_model_layers_88_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_36_residual_post_mlp/mean":148,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_49_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_63_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_30_residual_post_mlp/max_abs":125,"train/tensor_act_model_layers_93_residual_post_mlp/norm":770013.25,"train/tensor_act_model_layers_88_mlp_up_proj/max_abs":4.5625,"train/tensor_act_model_layers_52_residual_post_attn/max_abs":209,"train/tensor_act_model_layers_79_mlp_up_proj/std":1.796875,"train/tensor_act_model_layers_19_residual_post_mlp/mean":80,"train/tensor_act_model_layers_40_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_38_residual_post_attn/min":150,"train/tensor_param_model_layers_10_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_99_mlp_gate_proj/norm":5520.20947265625,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_99_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_residual_pre_attn/norm":417644.53125,"train/tensor_act_model_layers_42_mlp_down_proj/norm":30166.30078125,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/max_abs":0.1240234375,"train/tensor_param_model_layers_97_mlp_up_proj_weight/range":0.3828125,"train/tensor_act_model_layers_53_mlp/min":4,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/max":0.1875,"train/tensor_act_model_layers_35_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_65_residual_post_attn/min":258,"train/tensor_act_model_layers_54_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_95_mlp/max":4,"train/tensor_act_model_layers_58_residual_pre_attn/mean":232,"train/tensor_act_model_layers_97_residual_pre_attn/std":0.18359375,"train/tensor_grad_model_layers_56_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_79_mlp_gate_proj/mean":0.66796875,"train/tensor_act_model_layers_27_mlp_gate_proj/std":1.6640625,"train/tensor_param_model_layers_0_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_15_residual_pre_attn/min":59,"train/tensor_act_model_layers_2_residual_post_attn/range":0.96875,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_16_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_18_mlp_up_proj/min":-4.875,"train/tensor_act_model_layers_81_residual_pre_attn/norm":663517.3125,"train/tensor_act_model_layers_43_residual_post_attn/norm":352108.65625,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/max":0.130859375,"train/tensor_act_model_layers_97_residual_pre_attn/min":386,"train/tensor_param_model_layers_87_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_mlp_up_proj/max":4.3125,"train/tensor_act_model_layers_15_residual_pre_attn/std":0.3125,"train/tensor_param_model_layers_19_mlp_down_proj_weight/max_abs":0.12158203125,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_mlp_up_proj/max":4.65625,"train/tensor_act_model_layers_32_residual_post_mlp/min":130,"train/tensor_act_model_layers_65_residual_post_mlp/norm":540637.3125,"train/tensor_param_model_layers_90_mlp_up_proj_weight/max_abs":0.158203125,"train/tensor_act_model_layers_67_mlp_down_proj/mean":14.875,"train/tensor_act_model_layers_78_mlp_gate_proj/max_abs":4.5625,"train/tensor_act_model_layers_50_residual_post_attn/norm":409452.5625,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/max":0.1396484375,"train/tensor_act_model_layers_38_mlp_gate_proj/std":1.65625,"train/tensor_act_model_layers_36_mlp_gate_proj/mean":0.474609375,"train/tensor_act_model_layers_49_residual_post_attn/max_abs":197,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_61_residual_post_mlp/min":246,"train/tensor_act_model_layers_15_mlp/norm":8192,"train/tensor_grad_model_layers_15_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_81_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_79_mlp_down_proj_weight/norm":7.5,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/min":-0.1142578125,"train/tensor_act_model_layers_18_mlp_gate_proj/max_abs":4.71875,"train/tensor_act_model_layers_23_residual_post_mlp/min":94.5,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_70_mlp_gate_proj/min":-1.2734375,"train/tensor_act_model_layers_7_mlp_down_proj/range":2.1875,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_33_mlp_down_proj_weight/min":-0.126953125,"train/tensor_act_model_layers_24_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_0_mlp_down_proj/range":92.9375,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_42_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_55_mlp_down_proj/min":14.25,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_79_residual_pre_attn/max":316,"train/tensor_act_model_layers_22_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_75_mlp_down_proj/max":16,"train/tensor_act_model_layers_99_residual_post_mlp/mean":400,"train/tensor_param_model_layers_3_mlp_up_proj_weight/max":0.10693359375,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_50_mlp_gate_proj/std":1.6796875,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_8_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_99_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_76_residual_pre_attn/range":2,"train/tensor_param_model_layers_49_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_31_mlp_down_proj/max":16.125,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/min":-0.1689453125,"train/tensor_act_model_layers_66_mlp_gate_proj/range":5.703125,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_39_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_70_mlp_down_proj/mean":14.9375,"train/tensor_param_model_layers_94_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_67_residual_pre_attn/max_abs":268,"train/tensor_act_model_layers_53_mlp_up_proj/max_abs":4.9375,"train/tensor_act_model_layers_55_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_1_residual_post_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_99_residual_post_mlp/max":400,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/max_abs":0.208984375,"train/tensor_act_model_layers_41_mlp_gate_proj/range":5.8671875,"train/tensor_act_model_layers_96_mlp_up_proj/max":6.28125,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_9_mlp_gate_proj/norm":5144.61474609375,"train/tensor_act_model_layers_25_residual_post_mlp/max":105,"train/tensor_act_model_layers_67_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_mlp_up_proj/max_abs":4.8125,"train/tensor_grad_model_layers_1_mlp_gate_proj_weight/max":0.12109375,"train/tensor_param_model_layers_5_mlp_down_proj_weight/std":0.0294189453125,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_44_mlp_gate_proj_weight/min":-0.130859375,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_32_mlp_up_proj_weight/range":0.24462890625,"train/tensor_grad_model_layers_55_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_99_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/mean":0.00433349609375,"train/tensor_act_model_layers_97_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_88_mlp_up_proj/std":1.7734375,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_41_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_62_mlp_down_proj/max":16.875,"train/tensor_param_model_layers_43_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_param_model_layers_92_mlp_down_proj_weight/max":0.1513671875,"train/tensor_param_model_layers_5_mlp_down_proj_weight/norm":5.3125,"train/tensor_param_model_layers_40_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_59_residual_post_mlp/range":3,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_16_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_94_mlp_down_proj_weight/range":0.333984375,"train/tensor_param_model_layers_18_mlp_up_proj_weight/norm":5.625,"train/tensor_act_model_layers_39_residual_pre_attn/min":154,"train/tensor_act_model_layers_62_mlp/max_abs":4,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_20_mlp_gate_proj/max":4.96875,"train/tensor_act_model_layers_92_mlp_up_proj/max_abs":5.125,"train/tensor_act_model_layers_83_residual_post_mlp/mean":336,"train/tensor_act_model_layers_10_mlp_up_proj/mean":0.064453125,"train/tensor_act_model_layers_16_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_76_mlp_down_proj_weight/std":0.0400390625,"train/tensor_param_model_layers_27_mlp_down_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_8_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_7_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/range":0.24462890625,"train/tensor_act_model_layers_80_mlp_down_proj/std":0.59765625,"train/tensor_act_model_layers_54_mlp_down_proj/range":2.5,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_act_model_layers_60_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_70_mlp_down_proj/max_abs":17,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_56_mlp/max":4,"train/tensor_act_model_layers_68_mlp_down_proj/std":0.57421875,"train/tensor_act_model_layers_30_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_5_mlp/std":0,"train/tensor_act_model_layers_52_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_19_mlp_gate_proj/min":-1.28125,"train/tensor_param_model_layers_26_mlp_up_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_68_mlp_up_proj/max":3.796875,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_62_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_60_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_71_residual_post_attn/max":284,"train/tensor_act_model_layers_64_mlp_up_proj/max_abs":5.28125,"train/tensor_act_model_layers_60_residual_post_attn/norm":491292.78125,"train/tensor_act_model_layers_85_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/std":0.050537109375,"train/tensor_act_model_layers_66_mlp_gate_proj/max_abs":4.4375,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_97_mlp_up_proj/max":4.5625,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/norm":9.75,"train/tensor_act_model_layers_78_mlp_up_proj/std":1.7734375,"train/tensor_act_model_layers_63_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_89_mlp_up_proj/max":4.1875,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_82_mlp_down_proj/mean":15.1875,"train/tensor_act_model_layers_36_mlp/mean":4,"train/tensor_act_model_layers_16_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_23_mlp_down_proj_weight/range":0.24462890625,"train/tensor_act_model_layers_46_mlp_down_proj/std":0.42578125,"train/tensor_act_model_layers_38_mlp/std":0,"train/tensor_act_model_layers_44_residual_post_mlp/max_abs":181,"train/tensor_param_model_layers_27_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_7_mlp_up_proj/range":8.28125,"train/tensor_act_model_layers_63_mlp_up_proj/range":9.03125,"train/tensor_act_model_layers_22_residual_post_attn/max_abs":89,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_22_mlp_down_proj_weight/range":0.2216796875,"train/tensor_act_model_layers_47_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_1_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_mlp/max":4,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_act_model_layers_45_mlp_down_proj/max":16.5,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_31_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_40_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_91_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_58_mlp_down_proj_weight/min":-0.1259765625,"train/tensor_param_model_layers_53_mlp_down_proj_weight/norm":6.5625,"train/tensor_act_model_layers_31_mlp/norm":8192,"train/tensor_param_model_layers_98_mlp_up_proj_weight/mean":-0.0018768310546875,"train/tensor_act_model_layers_37_mlp_down_proj/max_abs":16.125,"train/tensor_param_model_layers_91_mlp_up_proj_weight/max":0.1669921875,"train/tensor_param_model_layers_43_mlp_up_proj_weight/max_abs":0.1259765625,"train/tensor_act_model_layers_38_mlp_gate_proj/norm":5158.802734375,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_8_mlp_gate_proj/max_abs":4.46875,"train/tensor_act_model_layers_98_mlp/max_abs":4,"train/tensor_act_model_layers_4_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_11_mlp_up_proj/range":7.765625,"train/tensor_act_model_layers_82_mlp_up_proj/range":9.0625,"train/tensor_param_model_layers_92_mlp_up_proj_weight/min":-0.150390625,"train/tensor_act_model_layers_57_residual_pre_attn/max":229,"train/tensor_param_model_layers_82_mlp_up_proj_weight/std":0.04443359375,"train/tensor_param_model_layers_1_mlp_up_proj_weight/max":0.1357421875,"train/tensor_act_model_layers_98_mlp_down_proj/min":13.9375,"train/tensor_act_model_layers_33_mlp_gate_proj/std":1.7421875,"train/tensor_act_model_layers_28_mlp/range":0,"train/tensor_act_model_layers_97_mlp_gate_proj/norm":4994.74462890625,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_68_mlp_down_proj_weight/norm":6.90625,"train/tensor_act_model_layers_32_mlp_down_proj/range":1.8125,"train/tensor_act_model_layers_94_mlp_down_proj/max":19,"train/tensor_act_model_layers_8_residual_post_mlp/range":1.75,"train/tensor_act_model_layers_42_residual_pre_attn/mean":168,"train/tensor_param_model_layers_86_mlp_up_proj_weight/range":0.28125,"train/tensor_param_model_layers_37_mlp_up_proj_weight/std":0.033935546875,"train/tensor_act_model_layers_15_residual_pre_attn/range":2,"train/tensor_param_model_layers_92_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_19_mlp_down_proj/range":2.25,"train/tensor_act_model_layers_71_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/norm":6.96875,"train/tensor_act_model_layers_10_mlp_gate_proj/max":4.09375,"train/tensor_act_model_layers_41_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_33_residual_post_mlp/mean":136,"train/tensor_grad_model_layers_31_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_35_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_92_residual_pre_attn/max_abs":368,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_44_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_90_mlp_gate_proj/mean":0.7890625,"train/tensor_act_model_layers_64_residual_post_mlp/min":258,"train/tensor_act_model_layers_75_residual_post_mlp/mean":304,"train/tensor_act_model_layers_86_residual_post_attn/max_abs":344,"train/tensor_act_model_layers_64_mlp_down_proj/norm":30462.94921875,"train/tensor_act_model_layers_54_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_8_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_77_mlp_down_proj_weight/range":0.2802734375,"train/tensor_act_model_layers_1_mlp_up_proj/mean":-0.2109375,"train/tensor_act_model_layers_44_mlp_down_proj/max_abs":16.125,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_36_mlp_gate_proj_weight/norm":6.1875,"train/tensor_act_model_layers_42_mlp_up_proj/std":1.8515625,"train/tensor_act_model_layers_0_residual_post_mlp/min":3.578125,"train/tensor_act_model_layers_36_mlp_up_proj/mean":-0.15625,"train/tensor_act_model_layers_44_residual_pre_attn/min":174,"train/tensor_act_model_layers_62_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_77_mlp_down_proj/max":19,"train/tensor_act_model_layers_44_mlp_gate_proj/min":-1.2578125,"train/tensor_act_model_layers_82_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_5_mlp_gate_proj/max":4.53125,"train/tensor_act_model_layers_14_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/mean":0.004119873046875,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_87_residual_pre_attn/mean":348,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_37_residual_post_mlp/min":150,"train/tensor_act_model_layers_42_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_77_residual_post_mlp/range":2,"train/tensor_act_model_layers_84_mlp_gate_proj/norm":5055.8916015625,"train/tensor_act_model_layers_13_residual_post_mlp/max_abs":57,"train/tensor_act_model_layers_34_mlp_down_proj/range":1.9375,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/norm":9.0625,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_22_residual_post_mlp/mean":92,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_57_mlp_up_proj_weight/range":0.2607421875,"train/tensor_param_model_layers_45_mlp_up_proj_weight/max_abs":0.12890625,"train/tensor_param_model_layers_39_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_18_residual_post_mlp/max_abs":77,"train/tensor_act_model_layers_24_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_77_residual_post_mlp/max":312,"train/tensor_act_model_layers_28_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_42_mlp_down_proj/mean":14.75,"train/tensor_param_model_layers_74_mlp_gate_proj_weight/min":-0.1376953125,"train/tensor_act_model_layers_42_residual_post_attn/std":0.31640625,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_81_mlp_gate_proj/norm":5005.9931640625,"train/tensor_param_model_layers_20_mlp_up_proj_weight/mean":0.00028228759765625,"train/tensor_act_model_layers_19_mlp_up_proj/max":4.03125,"train/tensor_act_model_layers_16_mlp_down_proj/norm":30108.021484375,"train/tensor_param_model_layers_98_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_66_mlp_down_proj_weight/norm":6.8125,"train/tensor_act_model_layers_46_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_41_mlp_up_proj_weight/range":0.2509765625,"train/tensor_act_model_layers_58_residual_post_mlp/min":234,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/min":-0.1357421875,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/min":-0.1611328125,"train/tensor_act_model_layers_67_mlp/norm":8192,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_66_residual_pre_attn/max":264,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_71_residual_pre_attn/mean":284,"train/tensor_act_model_layers_30_mlp_gate_proj/std":1.703125,"train/tensor_param_model_layers_22_mlp_up_proj_weight/mean":-0.00130462646484375,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_6_residual_post_attn/max_abs":24.625,"train/tensor_act_model_layers_48_residual_post_mlp/range":3,"train/tensor_act_model_layers_13_residual_post_attn/range":2,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_21_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_81_residual_post_attn/max_abs":324,"train/tensor_act_model_layers_10_mlp_up_proj/range":8.03125,"train/tensor_act_model_layers_87_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_44_residual_post_mlp/mean":180,"train/tensor_param_model_layers_70_mlp_down_proj_weight/range":0.255859375,"train/tensor_act_model_layers_13_mlp_down_proj/max":16.125,"train/tensor_act_model_layers_28_mlp_gate_proj/mean":0.515625,"train/tensor_act_model_layers_67_mlp_down_proj/norm":30469.083984375,"train/tensor_act_model_layers_42_mlp/min":4,"train/tensor_param_model_layers_69_mlp_gate_proj_weight/mean":0.004547119140625,"train/tensor_act_model_layers_87_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_76_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_50_mlp_up_proj_weight/range":0.2509765625,"train/tensor_act_model_layers_73_mlp_gate_proj/max_abs":4.40625,"train/tensor_act_model_layers_69_mlp_up_proj/norm":5234.2177734375,"train/tensor_act_model_layers_30_residual_post_mlp/max":125,"train/tensor_param_model_layers_36_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_96_mlp_gate_proj/mean":0.74609375,"train/tensor_param_model_layers_70_mlp_down_proj_weight/mean":0.000972747802734375,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_82_mlp_gate_proj/max":5.21875,"train/tensor_act_model_layers_11_mlp_gate_proj/max":3.859375,"train/tensor_act_model_layers_21_residual_pre_attn/max":85,"train/tensor_act_model_layers_88_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_83_mlp_up_proj_weight/min":-0.1396484375,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_56_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_73_mlp_gate_proj/min":-1.2578125,"train/tensor_param_model_layers_67_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_46_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/range":0.2490234375,"train/tensor_act_model_layers_63_mlp_gate_proj/max":4.875,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_33_mlp_down_proj_weight/max":0.12158203125,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_25_mlp_up_proj/max":4.1875,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_20_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_9_mlp_up_proj_weight/mean":-0.00171661376953125,"train/tensor_act_model_layers_6_residual_post_attn/range":1.5,"train/tensor_act_model_layers_48_mlp/std":0,"train/tensor_act_model_layers_36_mlp/norm":8192,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_58_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_mlp_down_proj/max_abs":16.75,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/max":0.12890625,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_99_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_16_mlp_down_proj/max":16.5,"train/tensor_act_model_layers_13_residual_post_attn/norm":106321.78125,"train/tensor_param_model_layers_35_mlp_up_proj_weight/std":0.032958984375,"train/tensor_act_model_layers_96_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_96_mlp/norm":8192,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_16_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_87_residual_post_attn/max":348,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_39_mlp/norm":8192,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/std":0.03271484375,"train/tensor_param_model_layers_51_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_mlp/range":0,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_70_residual_post_mlp/range":2,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/min":-0.1162109375,"train/tensor_act_model_layers_42_residual_post_attn/max":169,"train/tensor_act_model_layers_26_residual_post_mlp/mean":108,"train/tensor_act_model_layers_95_mlp/norm":8192,"train/tensor_act_model_layers_76_residual_pre_attn/max_abs":304,"train/tensor_act_model_layers_7_residual_post_mlp/min":31.125,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/min":-0.169921875,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_40_mlp_up_proj_weight/max":0.12451171875,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_43_residual_post_attn/min":170,"train/tensor_act_model_layers_86_residual_pre_attn/range":2,"train/tensor_param_model_layers_10_mlp_up_proj_weight/mean":0.000499725341796875,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_20_mlp_up_proj/range":9.25,"train/tensor_param_model_layers_79_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_21_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_91_mlp_up_proj_weight/max_abs":0.1669921875,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_55_residual_post_mlp/mean":224,"train/tensor_param_model_layers_65_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/max_abs":0.12158203125,"train/tensor_act_model_layers_34_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_80_mlp/max":4,"train_runtime":368.5397,"train/tensor_param_model_layers_40_mlp_down_proj_weight/mean":0.001007080078125,"train/tensor_param_model_layers_70_mlp_down_proj_weight/max":0.1298828125,"train/tensor_act_model_layers_59_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_53_mlp_down_proj_weight/max":0.125,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_59_mlp_up_proj/max":3.890625,"train/tensor_act_model_layers_95_residual_post_mlp/max":384,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/std":0.03857421875,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_57_residual_pre_attn/max_abs":229,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_16_mlp_up_proj_weight/norm":5.59375,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_43_residual_post_mlp/norm":360300.625,"train/tensor_act_model_layers_85_mlp_up_proj/max":4.90625,"train/tensor_param_model_layers_97_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_90_residual_post_attn/mean":360,"train/tensor_act_model_layers_69_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_24_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_mlp_gate_proj/min":-1.28125,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_30_mlp_down_proj_weight/range":0.2373046875,"train/tensor_param_model_layers_4_mlp_up_proj_weight/min":-0.10888671875,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_48_mlp_up_proj/std":1.828125,"train/tensor_act_model_layers_82_mlp_down_proj/norm":31106.0390625,"train/tensor_act_model_layers_54_residual_post_mlp/range":3,"train/tensor_act_model_layers_73_mlp/std":0,"train/tensor_act_model_layers_12_residual_pre_attn/max_abs":49,"train/tensor_act_model_layers_94_mlp_up_proj/max":5.25,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_66_residual_pre_attn/max_abs":264,"train/tensor_act_model_layers_8_mlp/range":0,"train/tensor_param_model_layers_84_mlp_down_proj_weight/range":0.3037109375,"train/tensor_act_model_layers_1_mlp_down_proj/min":13.8125,"train/tensor_act_model_layers_92_residual_pre_attn/range":2,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_58_mlp_down_proj_weight/max":0.1337890625,"train/tensor_act_model_layers_4_mlp_up_proj/mean":-0.003631591796875,"train/tensor_act_model_layers_48_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_81_mlp_down_proj/min":14.0625,"train/tensor_act_model_layers_92_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_5_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_16_residual_post_mlp/max_abs":69,"train/tensor_act_model_layers_15_mlp_down_proj/max":16.125,"train/tensor_act_model_layers_27_mlp/range":0,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_88_residual_pre_attn/range":2,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_21_mlp_up_proj/range":9.65625,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_71_mlp_up_proj/max_abs":4.65625,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_1_mlp/max":4,"train/tensor_act_model_layers_57_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_21_residual_post_attn/std":0.359375,"train/tensor_param_model_layers_19_mlp_gate_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_3_mlp_up_proj/max":4.21875,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_param_model_layers_82_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_46_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_45_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_59_mlp_up_proj_weight/std":0.037109375,"train/tensor_act_model_layers_55_residual_post_attn/mean":220,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_92_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_82_mlp_down_proj_weight/max_abs":0.1396484375,"train/tensor_act_model_layers_32_residual_post_attn/max_abs":129,"train/tensor_act_model_layers_59_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_92_mlp_gate_proj/min":-1.4921875,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_51_mlp_up_proj_weight/max_abs":0.1279296875,"train/tensor_act_model_layers_98_residual_pre_attn/range":2,"train/tensor_param_model_layers_35_mlp_up_proj_weight/max_abs":0.1220703125,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_10_mlp_up_proj_weight/norm":5.4375,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_74_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_73_mlp_up_proj/max":4.25,"train/tensor_act_model_layers_33_mlp_up_proj/min":-4.6875,"train/tensor_param_model_layers_80_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_79_mlp_down_proj/max_abs":16.5,"train/tensor_act_model_layers_88_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/range":0.24609375,"train/tensor_act_model_layers_60_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_67_mlp_gate_proj/range":5.9375,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_5_mlp_up_proj/mean":0.0966796875,"train/tensor_act_model_layers_28_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_59_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/range":0.24755859375,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_98_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_44_mlp_gate_proj/range":6.2890625,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_16_mlp_up_proj_weight/mean":-0.00060272216796875,"train/tensor_act_model_layers_88_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_48_mlp_up_proj_weight/std":0.035400390625,"train/tensor_param_model_layers_88_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_12_mlp_down_proj/std":0.40625,"train/tensor_act_model_layers_12_residual_post_mlp/std":0.267578125,"train/tensor_act_model_layers_8_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_70_mlp/range":0,"train/tensor_act_model_layers_5_mlp_down_proj/range":2.4375,"train/tensor_act_model_layers_53_residual_post_attn/max_abs":213,"train/tensor_act_model_layers_69_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_77_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_58_residual_pre_attn/min":230,"train/tensor_param_model_layers_14_mlp_up_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_9_mlp_down_proj/norm":30019.248046875,"train/tensor_param_model_layers_22_mlp_up_proj_weight/norm":5.65625,"train/tensor_act_model_layers_37_mlp_down_proj/std":0.41015625,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_20_mlp_up_proj_weight/norm":5.65625,"train/tensor_param_model_layers_66_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_91_mlp_gate_proj/range":6.3984375,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/mean":0.0038909912109375,"train/tensor_param_model_layers_4_mlp_down_proj_weight/norm":5.3125,"train/tensor_act_model_layers_21_residual_post_attn/max":85,"train/tensor_param_model_layers_76_mlp_up_proj_weight/range":0.259765625,"train/tensor_act_model_layers_30_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_62_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_70_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_67_mlp_up_proj_weight/std":0.038330078125,"train/tensor_act_model_layers_12_mlp_up_proj/std":1.890625,"train/tensor_act_model_layers_45_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/norm":10.25,"train/tensor_grad_model_layers_51_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_57_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_22_mlp_down_proj/max_abs":15.5,"train/tensor_param_model_layers_87_mlp_up_proj_weight/max":0.15234375,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/max":0.119140625,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/mean":0.0052490234375,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_64_mlp_up_proj/range":9.34375,"train/tensor_act_model_layers_58_residual_pre_attn/std":0.416015625,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_43_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_54_mlp_gate_proj/max":4.65625,"train/tensor_act_model_layers_90_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_15_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_32_mlp_down_proj_weight/min":-0.125,"train/tensor_act_model_layers_75_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_59_mlp_down_proj/std":0.443359375,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_51_mlp_down_proj_weight/max_abs":0.12890625,"train/tensor_param_model_layers_49_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_25_residual_post_attn/min":98.5,"train/tensor_act_model_layers_18_mlp_up_proj/max":3.421875,"train/tensor_act_model_layers_32_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/norm":0,"train/tensor_param_model_layers_8_mlp_up_proj_weight/max_abs":0.11962890625,"train/tensor_act_model_layers_43_mlp_gate_proj/norm":5035.10498046875,"train/tensor_act_model_layers_2_mlp_gate_proj/range":5.46875,"train/tensor_act_model_layers_96_residual_post_mlp/mean":388,"train/tensor_param_model_layers_56_mlp_down_proj_weight/min":-0.1376953125,"train/tensor_grad_model_layers_32_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_39_mlp_gate_proj/range":5.90625,"train/tensor_act_model_layers_40_mlp_gate_proj/norm":5062.08447265625,"train/tensor_param_model_layers_47_mlp_down_proj_weight/max":0.1259765625,"train/tensor_act_model_layers_78_mlp_down_proj/range":2.875,"train/tensor_param_model_layers_94_mlp_down_proj_weight/mean":0.000850677490234375,"train/tensor_act_model_layers_73_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_14_mlp_up_proj/min":-4.46875,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/max":0.1337890625,"train/tensor_act_model_layers_71_mlp_up_proj/min":-3.9375,"train/tensor_act_model_layers_12_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_69_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_25_residual_post_mlp/norm":212753.15625,"train/tensor_act_model_layers_69_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_83_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_35_mlp_gate_proj/min":-1.3125,"train/tensor_param_model_layers_71_mlp_down_proj_weight/range":0.2783203125,"train/tensor_act_model_layers_34_mlp_down_proj/max_abs":16.125,"train/tensor_act_model_layers_55_mlp_gate_proj/norm":4979.80615234375,"train/tensor_param_model_layers_4_mlp_down_proj_weight/std":0.029296875,"train/tensor_act_model_layers_85_residual_post_mlp/max":344,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_32_mlp_gate_proj/min":-1.28125,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_51_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_94_mlp_gate_proj/range":7.0390625,"train/tensor_act_model_layers_27_mlp_gate_proj/range":6.125,"train/tensor_param_model_layers_18_mlp_gate_proj_weight/mean":0.003631591796875,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_22_mlp/max":4,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_12_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_residual_post_mlp/mean":124,"train/tensor_param_model_layers_72_mlp_down_proj_weight/min":-0.1279296875,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_95_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_85_mlp/min":4,"train/tensor_act_model_layers_10_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_13_mlp_up_proj_weight/min":-0.11181640625,"train/tensor_act_model_layers_54_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/mean":0.0040283203125,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_66_mlp/norm":8192,"train/tensor_param_model_layers_29_mlp_up_proj_weight/range":0.2216796875,"train/tensor_act_model_layers_21_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_71_mlp_down_proj_weight/std":0.03857421875,"train/tensor_act_model_layers_23_residual_post_attn/norm":188177.328125,"train/tensor_act_model_layers_86_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_49_mlp_up_proj/std":1.8671875,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_75_mlp_down_proj/mean":14.875,"train/tensor_act_model_layers_25_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_32_mlp_down_proj/max_abs":15.9375,"train/tensor_act_model_layers_16_mlp_gate_proj/max_abs":5.0625,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_83_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_76_mlp_up_proj/mean":0.08203125,"train/tensor_act_model_layers_7_mlp_up_proj/norm":5571.44091796875,"train/tensor_act_model_layers_84_mlp_gate_proj/std":1.609375,"train/tensor_act_model_layers_3_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_98_mlp_up_proj_weight/norm":9.875,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_48_residual_post_mlp/mean":196,"train/tensor_act_model_layers_54_mlp_up_proj/range":8.46875,"train/tensor_act_model_layers_13_mlp_up_proj/max":4,"train/tensor_act_model_layers_91_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_9_mlp/range":0,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_54_mlp_down_proj/mean":14.875,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_43_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_88_mlp/min":4,"train/tensor_param_model_layers_86_mlp_down_proj_weight/mean":-0.000553131103515625,"train/tensor_act_model_layers_64_mlp_gate_proj/max":4.375,"train/tensor_act_model_layers_77_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_pre_attn/min":254,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/max_abs":0.1298828125,"train/tensor_act_model_layers_15_residual_pre_attn/norm":122679.5234375,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_47_residual_post_mlp/mean":192,"train/tensor_act_model_layers_0_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_39_mlp/std":0,"train/tensor_act_model_layers_94_mlp_up_proj/min":-4.3125,"train/tensor_act_model_layers_49_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_11_mlp_down_proj_weight/std":0.029296875,"train/tensor_act_model_layers_15_mlp_gate_proj/mean":0.47265625,"train/tensor_act_model_layers_13_mlp_gate_proj/range":5.9140625,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_37_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_13_mlp/mean":4,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_72_mlp_gate_proj/mean":0.41796875,"train/tensor_param_model_layers_90_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_37_residual_pre_attn/min":146,"train/tensor_act_model_layers_70_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_7_mlp_up_proj_weight/min":-0.103515625,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_97_mlp_gate_proj/min":-1.765625,"train/tensor_act_model_layers_69_mlp_down_proj/max_abs":17.125,"train/tensor_act_model_layers_20_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_39_residual_post_attn/max_abs":157,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_31_residual_pre_attn/max_abs":125,"train/tensor_act_model_layers_53_mlp/range":0,"train/tensor_act_model_layers_4_residual_post_mlp/norm":40939.65234375,"train/tensor_act_model_layers_64_residual_pre_attn/std":0.37109375,"train/tensor_param_model_layers_69_mlp_down_proj_weight/min":-0.1416015625,"train/tensor_param_model_layers_12_mlp_down_proj_weight/max":0.1201171875,"train/tensor_act_model_layers_15_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_9_mlp/mean":4,"train/tensor_act_model_layers_86_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_10_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_94_mlp/min":4,"train/tensor_act_model_layers_44_residual_post_mlp/min":178,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_58_mlp_up_proj_weight/max_abs":0.13671875,"train/tensor_act_model_layers_48_mlp_down_proj/range":2.75,"train/tensor_act_model_layers_32_residual_post_attn/mean":128,"train/tensor_param_model_layers_56_mlp_up_proj_weight/max_abs":0.130859375,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_37_mlp_gate_proj/max_abs":4.78125,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/norm":5.46875,"train/tensor_param_model_layers_62_mlp_down_proj_weight/std":0.036865234375,"train/tensor_act_model_layers_45_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_58_mlp_down_proj/max_abs":17.125,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_90_mlp_gate_proj/std":1.6484375,"train/tensor_act_model_layers_26_mlp_gate_proj/norm":5313.6259765625,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_96_residual_post_mlp/max":388,"train/tensor_act_model_layers_56_mlp_up_proj/std":1.8359375,"train/tensor_act_model_layers_29_mlp/range":0,"train/tensor_act_model_layers_45_mlp_up_proj/std":1.8828125,"train/tensor_param_model_layers_8_mlp_down_proj_weight/std":0.0291748046875,"train/tensor_act_model_layers_39_mlp_down_proj/norm":30161.16796875,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_residual_pre_attn/max":153,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_51_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_3_mlp_down_proj/max_abs":16.375,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/max":0.125,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_39_residual_post_mlp/min":158,"train/tensor_act_model_layers_68_residual_pre_attn/max_abs":272,"train/tensor_act_model_layers_95_residual_post_attn/norm":778205.25,"train/tensor_act_model_layers_77_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_99_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_18_mlp_up_proj_weight/mean":-0.00213623046875,"train/tensor_act_model_layers_16_residual_post_mlp/max":69,"train/tensor_act_model_layers_62_residual_post_mlp/std":0.416015625,"train/tensor_act_model_layers_94_residual_post_mlp/max":380,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_90_mlp_down_proj_weight/norm":8.5,"train/tensor_act_model_layers_35_mlp_up_proj/max_abs":4.9375,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_26_mlp_down_proj/range":2,"train/tensor_param_model_layers_67_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_75_mlp/mean":4,"train/tensor_act_model_layers_85_residual_pre_attn/range":2,"train/tensor_act_model_layers_12_mlp/std":0,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_35_residual_post_attn/min":138,"train/tensor_act_model_layers_41_mlp/range":0,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_0_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_pre_attn/range":2,"train/tensor_param_model_layers_80_mlp_down_proj_weight/max_abs":0.1416015625,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_6_mlp_up_proj_weight/norm":5.375,"train/tensor_param_model_layers_33_mlp_gate_proj_weight/mean":0.004119873046875,"train/tensor_act_model_layers_70_mlp_down_proj/std":0.5625,"train/tensor_act_model_layers_21_mlp_gate_proj/max":4.46875,"train/tensor_act_model_layers_69_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_12_mlp/max":4,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_94_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_46_mlp_down_proj/max_abs":16.5,"train/tensor_act_model_layers_98_mlp_up_proj/std":1.7890625,"train/tensor_param_model_layers_31_mlp_up_proj_weight/max_abs":0.1298828125,"train/tensor_param_model_layers_8_mlp_up_proj_weight/mean":-0.0004863739013671875,"train/tensor_param_model_layers_72_mlp_up_proj_weight/max_abs":0.1328125,"train/tensor_act_model_layers_83_mlp/max_abs":4,"train/tensor_param_model_layers_64_mlp_up_proj_weight/max":0.1318359375,"train/tensor_param_model_layers_1_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_73_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_33_mlp_down_proj/norm":30118.017578125,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_72_residual_post_mlp/range":2,"train/tensor_act_model_layers_70_mlp_up_proj/norm":5318.857421875,"train/tensor_act_model_layers_35_residual_post_mlp/norm":294764.78125,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_51_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_47_mlp_up_proj_weight/min":-0.13671875,"train/tensor_act_model_layers_21_residual_pre_attn/mean":84,"train/tensor_act_model_layers_15_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_77_mlp_up_proj/norm":5346.12353515625,"train/tensor_act_model_layers_38_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_56_residual_pre_attn/max_abs":225,"train/tensor_grad_model_layers_72_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_59_residual_post_attn/max":237,"train/tensor_act_model_layers_56_residual_pre_attn/max":225,"train/tensor_act_model_layers_75_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_98_mlp_gate_proj/mean":0.703125,"train/tensor_act_model_layers_20_residual_post_attn/norm":163601.53125,"train/tensor_act_model_layers_15_residual_post_mlp/min":63,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_5_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_26_residual_post_attn/range":2.5,"train/tensor_grad_model_layers_89_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_0_mlp/std":0.007171630859375,"train/tensor_act_model_layers_25_residual_post_attn/range":2.5,"train/tensor_act_model_layers_49_residual_pre_attn/norm":401260.59375,"train/tensor_grad_model_layers_1_mlp_up_proj_weight/norm":5.3125,"train/tensor_grad_model_layers_69_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_97_mlp_up_proj/mean":0.09716796875,"train/tensor_act_model_layers_53_mlp/mean":4,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_32_residual_pre_attn/std":0.34375,"train/tensor_act_model_layers_32_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_18_mlp_gate_proj/mean":0.46484375,"train/tensor_act_model_layers_1_residual_post_mlp/range":0.875,"train/tensor_act_model_layers_76_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_84_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_3_residual_post_mlp/std":0.1591796875,"train/tensor_act_model_layers_1_mlp/std":1.8715858459472656e-05,"train/tensor_act_model_layers_13_residual_post_attn/max_abs":53,"train/tensor_param_model_layers_86_mlp_up_proj_weight/max":0.1376953125,"train/tensor_act_model_layers_25_mlp_up_proj/min":-3.5625,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_32_residual_pre_attn/min":126.5,"train/tensor_act_model_layers_83_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_84_mlp_up_proj_weight/min":-0.14453125,"train/tensor_param_model_layers_43_mlp_up_proj_weight/std":0.0341796875,"train/tensor_param_model_layers_33_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_51_residual_post_mlp/max_abs":209,"train/tensor_act_model_layers_30_residual_post_attn/range":2.5,"train/tensor_act_model_layers_95_mlp_up_proj/max_abs":6,"train/tensor_act_model_layers_18_residual_post_attn/max":73,"train/tensor_act_model_layers_31_residual_pre_attn/range":2.5,"train/tensor_param_model_layers_27_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_80_mlp_gate_proj/min":-1.6171875,"train/tensor_act_model_layers_97_residual_post_mlp/min":390,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/range":0.24462890625,"train/tensor_act_model_layers_23_residual_pre_attn/min":90.5,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/std":0.0634765625,"train/tensor_act_model_layers_98_mlp_gate_proj/range":7.34375,"train/tensor_param_model_layers_94_mlp_up_proj_weight/std":0.05126953125,"train/tensor_act_model_layers_58_mlp/max":4,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/min":-0.146484375,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_34_residual_pre_attn/max":137,"train/tensor_param_model_layers_84_mlp_down_proj_weight/max_abs":0.158203125,"train/tensor_act_model_layers_19_mlp_up_proj/min":-5.0625,"train/tensor_param_model_layers_47_mlp_down_proj_weight/min":-0.11962890625,"train/tensor_act_model_layers_76_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_mlp_down_proj/max_abs":18.25,"train/tensor_act_model_layers_51_mlp_gate_proj/range":6.3125,"train/tensor_act_model_layers_13_mlp/min":4,"train/tensor_act_model_layers_35_residual_post_mlp/max_abs":145,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_61_residual_pre_attn/min":242,"train/tensor_act_model_layers_67_mlp/max":4,"train/tensor_param_model_layers_14_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_19_mlp_down_proj/std":0.412109375,"train/tensor_act_model_layers_49_mlp/max":4,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/max_abs":0.1259765625,"train/tensor_param_model_layers_32_mlp_down_proj_weight/max":0.123046875,"train/tensor_act_model_layers_61_residual_pre_attn/std":0.416015625,"train/tensor_act_model_layers_0_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_84_mlp_down_proj_weight/mean":0.0003490447998046875,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_12_mlp_down_proj/norm":30048.3515625,"train/tensor_act_model_layers_68_mlp_gate_proj/min":-1.2734375,"train/tensor_param_model_layers_56_mlp_down_proj_weight/max":0.1279296875,"train/tensor_param_model_layers_14_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_99_mlp_up_proj/max_abs":5.71875,"train/tensor_act_model_layers_64_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_residual_pre_attn/std":0.359375,"train/tensor_act_model_layers_85_mlp/mean":4,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_mlp_gate_proj/std":1.6484375,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_4_mlp_down_proj/max_abs":15.9375,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_35_mlp_down_proj_weight/range":0.23291015625,"train/tensor_param_model_layers_96_mlp_down_proj_weight/mean":0.00102996826171875,"train/tensor_act_model_layers_86_mlp_gate_proj/std":1.640625,"train/tensor_act_model_layers_46_mlp_gate_proj/min":-1.2734375,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/std":0.046875,"train/tensor_act_model_layers_89_mlp/norm":8192,"train/tensor_act_model_layers_41_residual_post_attn/range":3,"train/tensor_act_model_layers_75_residual_post_attn/mean":300,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_22_mlp_gate_proj/norm":5215.30517578125,"train/tensor_param_model_layers_29_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_residual_post_mlp/max":245,"train/tensor_act_model_layers_11_residual_post_mlp/max":49,"train/tensor_act_model_layers_4_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_18_residual_post_attn/std":0.3515625,"train/tensor_grad_model_layers_29_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_39_mlp_up_proj/range":8.34375,"train/tensor_param_model_layers_2_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/std":0.03271484375,"train/tensor_param_model_layers_71_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_59_residual_post_attn/std":0.416015625,"train/tensor_act_model_layers_59_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_residual_post_mlp/range":2,"train/tensor_act_model_layers_41_mlp/mean":4,"train/tensor_act_model_layers_75_mlp_up_proj/max_abs":4.6875,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_residual_post_mlp/range":3,"train/tensor_act_model_layers_59_mlp_up_proj/min":-5.03125,"train/tensor_act_model_layers_4_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_mlp_gate_proj/mean":0.58203125,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_76_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_61_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_93_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/max":0.130859375,"train/tensor_act_model_layers_16_residual_pre_attn/range":2,"train/tensor_param_model_layers_50_mlp_down_proj_weight/max_abs":0.130859375,"train/tensor_act_model_layers_64_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_63_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_72_mlp_up_proj/max_abs":5.125,"train/tensor_param_model_layers_49_mlp_down_proj_weight/std":0.03564453125,"train/tensor_param_model_layers_68_mlp_up_proj_weight/range":0.2626953125,"train/tensor_act_model_layers_8_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_61_residual_post_attn/norm":499484.78125,"train/tensor_act_model_layers_85_mlp_up_proj/min":-4.25,"train/tensor_param_model_layers_63_mlp_up_proj_weight/norm":6.8125,"train/tensor_act_model_layers_81_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_23_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_24_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_80_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_8_mlp_down_proj_weight/max_abs":0.111328125,"train/tensor_act_model_layers_93_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_29_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_99_mlp_down_proj_weight/mean":-0.0003376007080078125,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_58_mlp_gate_proj/max_abs":5.21875,"train/tensor_act_model_layers_99_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp/std":0,"train/tensor_act_model_layers_50_residual_pre_attn/max":201,"train/tensor_act_model_layers_90_mlp/min":4,"train/tensor_act_model_layers_63_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_40_mlp_up_proj_weight/range":0.24755859375,"train/tensor_act_model_layers_53_residual_post_mlp/mean":216,"train/tensor_act_model_layers_44_mlp_down_proj/std":0.4140625,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/std":0.038818359375,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_18_residual_post_attn/norm":147242.40625,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_58_residual_post_mlp/max_abs":237,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_10_residual_pre_attn/mean":40,"train/tensor_param_model_layers_63_mlp_down_proj_weight/norm":6.65625,"train/tensor_param_model_layers_44_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_act_model_layers_7_mlp_down_proj/max_abs":16.25,"train/tensor_param_model_layers_95_mlp_up_proj_weight/min":-0.15625,"train/tensor_act_model_layers_68_mlp_gate_proj/range":5.8984375,"train/tensor_act_model_layers_53_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_6_residual_pre_attn/max":24.625,"train/tensor_param_model_layers_68_mlp_down_proj_weight/max_abs":0.1328125,"train/tensor_param_model_layers_53_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_62_mlp_up_proj/max_abs":4.9375,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_24_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_59_residual_pre_attn/max_abs":237,"train/tensor_param_model_layers_28_mlp_down_proj_weight/mean":0.00014781951904296875,"train/tensor_param_model_layers_57_mlp_down_proj_weight/norm":6.625,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_70_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_8_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_48_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_39_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_53_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_10_mlp_gate_proj_weight/norm":5.46875,"train/tensor_grad_model_layers_6_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_75_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_78_residual_post_attn/mean":312,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_39_mlp_down_proj/mean":14.75,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_85_residual_post_attn/min":338,"train/tensor_act_model_layers_56_residual_post_attn/mean":224,"train/tensor_act_model_layers_49_mlp/max_abs":4,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_94_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_90_residual_post_attn/max_abs":360,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/mean":0.004913330078125,"train/tensor_param_model_layers_14_mlp_down_proj_weight/mean":0.00154876708984375,"train/tensor_act_model_layers_34_residual_post_mlp/range":3,"train/tensor_param_model_layers_74_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_10_residual_post_attn/std":0.224609375,"train/tensor_act_model_layers_71_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_17_residual_post_mlp/max":73,"train/tensor_act_model_layers_78_mlp/max_abs":4,"train/tensor_act_model_layers_85_residual_post_mlp/norm":704477.25,"train/tensor_act_model_layers_68_mlp_up_proj/min":-4.78125,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/norm":5.6875,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_28_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/mean":-21.625,"train/tensor_act_model_layers_28_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_61_mlp_up_proj/std":1.84375,"train/tensor_act_model_layers_73_mlp/min":4,"train/tensor_param_model_layers_64_mlp_down_proj_weight/max_abs":0.1328125,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_31_mlp_down_proj_weight/std":0.031982421875,"train/tensor_act_model_layers_3_mlp_up_proj/min":-4.125,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_37_mlp_up_proj_weight/mean":0.00058746337890625,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_46_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_50_mlp_up_proj/std":1.875,"train/tensor_grad_model_layers_20_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_95_mlp_down_proj/max":18.375,"train/tensor_act_model_layers_67_mlp_gate_proj/std":1.6796875,"train/tensor_act_model_layers_33_mlp_up_proj/max_abs":4.75,"train/tensor_act_model_layers_52_residual_post_attn/range":3,"train/tensor_act_model_layers_39_mlp_gate_proj/max_abs":4.59375,"train/tensor_act_model_layers_52_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_18_residual_pre_attn/mean":72,"train/tensor_act_model_layers_73_mlp_down_proj/std":0.640625,"train/tensor_act_model_layers_26_residual_post_mlp/max_abs":109,"train/tensor_act_model_layers_8_residual_post_attn/max":32.75,"train/tensor_act_model_layers_57_residual_post_mlp/mean":232,"train/tensor_act_model_layers_88_mlp/max_abs":4,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_33_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_mlp_up_proj/min":-4.4375,"train/tensor_act_model_layers_69_residual_post_mlp/min":278,"train/tensor_act_model_layers_96_mlp_down_proj/max":20.25,"train/tensor_act_model_layers_10_residual_post_mlp/range":1.75,"train/tensor_param_model_layers_0_mlp_down_proj_weight/max":0.125,"train/tensor_act_model_layers_53_residual_post_attn/mean":212,"train/tensor_act_model_layers_33_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_39_mlp_gate_proj/max":4.59375,"train/tensor_act_model_layers_21_residual_pre_attn/range":2.5,"train/tensor_param_model_layers_30_mlp_up_proj_weight/max":0.11669921875,"train/tensor_act_model_layers_20_mlp/norm":8192,"train/tensor_act_model_layers_72_mlp_down_proj/max":16.625,"train/tensor_act_model_layers_93_mlp_gate_proj/mean":0.6171875,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/std":0.037841796875,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_94_mlp/mean":4,"train/tensor_act_model_layers_38_residual_post_attn/mean":152,"train/tensor_param_model_layers_78_mlp_up_proj_weight/range":0.2626953125,"train/tensor_act_model_layers_59_residual_pre_attn/mean":236,"train/tensor_act_model_layers_62_mlp_up_proj/mean":0.09423828125,"train/tensor_act_model_layers_30_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_28_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_22_mlp_up_proj/norm":5391.85107421875,"train/tensor_act_model_layers_79_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_70_mlp/max_abs":4,"train/tensor_act_model_layers_19_residual_pre_attn/max":77,"train/tensor_act_model_layers_45_mlp_down_proj/range":2.25,"train/tensor_act_model_layers_25_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_73_mlp/max":4,"train/tensor_param_model_layers_44_mlp_up_proj_weight/norm":6.15625,"train/tensor_act_model_layers_90_mlp_down_proj/range":3.8125,"train/tensor_act_model_layers_75_mlp/norm":8192,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_38_residual_post_attn/range":3,"train/tensor_act_model_layers_10_residual_post_attn/max":40.75,"train/tensor_act_model_layers_90_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_71_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_residual_post_attn/max_abs":181,"train/tensor_act_model_layers_74_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_77_mlp_down_proj_weight/mean":-0.00081634521484375,"train/tensor_act_model_layers_44_residual_post_attn/norm":360300.625,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/max_abs":0.11669921875,"train/tensor_act_model_layers_49_mlp/norm":8192,"train/tensor_act_model_layers_71_mlp/norm":8192,"train/tensor_act_model_layers_76_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_74_mlp/mean":4,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/mean":-0.00029754638671875,"train/tensor_param_model_layers_2_mlp_gate_proj_weight/max_abs":0.125,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_33_mlp_down_proj_weight/norm":5.78125,"train/tensor_act_model_layers_90_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_73_residual_post_mlp/max_abs":296,"train/tensor_act_model_layers_36_mlp_down_proj/max":15.6875,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_84_mlp/max":4,"train/tensor_grad_model_layers_60_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_57_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_22_mlp_down_proj_weight/std":0.0311279296875,"train/tensor_act_model_layers_37_mlp_up_proj/std":1.8671875,"train/tensor_param_model_layers_78_mlp_up_proj_weight/norm":7.5,"train/tensor_act_model_layers_14_mlp_gate_proj/mean":0.4609375,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/norm":6.28125,"train/tensor_act_model_layers_96_residual_post_attn/norm":786397.25,"train/tensor_act_model_layers_81_residual_pre_attn/max":324,"train/tensor_param_model_layers_72_mlp_down_proj_weight/max":0.1298828125,"train/tensor_act_model_layers_75_mlp_down_proj/norm":30509.4453125,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_50_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_66_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_25_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_28_residual_pre_attn/max":113,"train/tensor_param_model_layers_51_mlp_up_proj_weight/std":0.035888671875,"train/tensor_act_model_layers_56_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_4_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_10_mlp/max":4,"train/tensor_act_model_layers_90_residual_pre_attn/min":358,"train/tensor_param_model_layers_80_mlp_down_proj_weight/std":0.042236328125,"train/tensor_act_model_layers_11_residual_post_attn/norm":89977.96875,"train/tensor_act_model_layers_57_mlp_up_proj/std":1.796875,"train/tensor_act_model_layers_30_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_89_mlp_up_proj/min":-4.8125,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_97_mlp_up_proj/min":-5.40625,"train/tensor_param_model_layers_6_mlp_up_proj_weight/mean":0.000152587890625,"train/tensor_act_model_layers_62_mlp_gate_proj/norm":4965.28076171875,"train/tensor_act_model_layers_77_residual_pre_attn/max_abs":308,"train/tensor_param_model_layers_88_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_62_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_43_mlp_up_proj_weight/norm":6.1875,"train/tensor_act_model_layers_54_mlp_down_proj/norm":30442.19921875,"train/tensor_act_model_layers_82_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_6_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_68_mlp_up_proj/norm":5174.5546875,"train/tensor_param_model_layers_12_mlp_down_proj_weight/min":-0.115234375,"train/tensor_param_model_layers_69_mlp_down_proj_weight/max":0.1279296875,"train/tensor_act_model_layers_47_residual_pre_attn/max":189,"train/tensor_act_model_layers_42_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_residual_post_attn/min":262,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_97_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_22_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_32_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_6_mlp/max_abs":4,"train/tensor_act_model_layers_44_mlp_up_proj/mean":0.0033721923828125,"train/tensor_act_model_layers_66_residual_post_mlp/norm":548829.3125,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_50_mlp_down_proj_weight/mean":0.0001392364501953125,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_69_mlp_down_proj/std":0.51171875,"train/tensor_param_model_layers_42_mlp_down_proj_weight/mean":0.0004673004150390625,"train/tensor_param_model_layers_8_mlp_up_proj_weight/norm":5.375,"train/tensor_act_model_layers_45_mlp_up_proj/mean":-0.1494140625,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_78_mlp_down_proj_weight/mean":0.00101470947265625,"train/tensor_act_model_layers_82_residual_pre_attn/min":326,"train/tensor_param_model_layers_57_mlp_up_proj_weight/min":-0.1328125,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/norm":6.6875,"train/tensor_param_model_layers_1_mlp_down_proj_weight/range":0.25,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_2_mlp_up_proj_weight/min":-0.125,"train/tensor_act_model_layers_67_mlp_down_proj/max_abs":17.625,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/mean":0.004791259765625,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_86_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/mean":0.003692626953125,"train/tensor_grad_model_layers_28_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_21_mlp_up_proj/max":4.84375,"train/tensor_param_model_layers_51_mlp_gate_proj_weight/max_abs":0.1357421875,"train/tensor_act_model_layers_73_residual_post_mlp/max":296,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_66_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_residual_post_mlp/max_abs":185,"train/tensor_param_model_layers_48_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_8_mlp_down_proj_weight/mean":0.00103759765625,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/min":-0.12890625,"train/tensor_act_model_layers_8_residual_post_mlp/mean":36,"train/tensor_act_model_layers_17_mlp/norm":8192,"train/tensor_act_model_layers_32_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_64_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_42_residual_pre_attn/max_abs":169,"train/tensor_grad_model_layers_64_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_50_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_17_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/norm":6.28125,"train/tensor_act_model_layers_67_mlp_gate_proj/min":-1.375,"train/tensor_grad_model_layers_66_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_48_residual_post_attn/norm":393068.59375,"train/tensor_act_model_layers_5_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_0_mlp_down_proj/min":6.5625,"train/tensor_act_model_layers_98_residual_post_attn/range":2,"train/tensor_act_model_layers_96_mlp_down_proj/max_abs":20.25,"train/tensor_grad_model_layers_99_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_71_residual_pre_attn/max_abs":284,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_8_residual_post_attn/norm":65420.06640625,"train/tensor_act_model_layers_57_mlp_gate_proj/range":6.421875,"train/tensor_param_model_layers_63_mlp_gate_proj_weight/range":0.2646484375,"train/tensor_act_model_layers_14_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_80_residual_post_attn/range":2,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/min":-868,"train/tensor_act_model_layers_48_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_27_residual_post_mlp/max":113,"train/tensor_act_model_layers_76_mlp_up_proj/range":9.34375,"train/tensor_act_model_layers_72_mlp_down_proj/max_abs":16.625,"train/tensor_act_model_layers_34_mlp/max_abs":4,"train/tensor_grad_model_layers_44_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_66_mlp_up_proj_weight/std":0.037841796875,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_57_mlp_gate_proj/max_abs":5.0625,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_60_mlp/min":4,"train/tensor_grad_model_layers_96_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_29_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_39_mlp_up_proj/mean":-0.3046875,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_10_mlp/mean":4,"train/tensor_act_model_layers_67_mlp_up_proj/norm":5323.75390625,"train/tensor_act_model_layers_51_mlp_up_proj/min":-4.90625,"train/tensor_param_model_layers_87_mlp_down_proj_weight/max_abs":0.1396484375,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_88_mlp_gate_proj/max":4.59375,"train/tensor_act_model_layers_48_mlp_up_proj/range":8.515625,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/range":0.2802734375,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_63_residual_post_attn/range":3,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_30_mlp_down_proj/std":0.404296875,"train/tensor_act_model_layers_15_mlp_down_proj/range":2,"train/tensor_act_model_layers_12_residual_post_mlp/max_abs":53,"train/tensor_act_model_layers_66_residual_post_mlp/max_abs":268,"train/tensor_grad_model_layers_29_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_83_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_81_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_60_mlp_up_proj_weight/norm":6.71875,"train/tensor_act_model_layers_50_residual_post_attn/max_abs":201,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/std":316,"train/tensor_act_model_layers_13_residual_post_mlp/range":2,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_24_residual_post_mlp/norm":204561.203125,"train/tensor_param_model_layers_99_mlp_up_proj_weight/min":-0.1591796875,"train/tensor_act_model_layers_17_mlp_up_proj/std":1.890625,"train/tensor_act_model_layers_76_mlp_up_proj/max":5,"train/tensor_act_model_layers_86_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_73_mlp_up_proj_weight/norm":7,"train/tensor_act_model_layers_47_mlp_down_proj/norm":30246.748046875,"train/tensor_param_model_layers_72_mlp_down_proj_weight/range":0.2578125,"train/tensor_param_model_layers_53_mlp_up_proj_weight/std":0.03662109375,"train/tensor_act_model_layers_20_residual_post_mlp/max_abs":85,"train/tensor_act_model_layers_4_residual_post_attn/max":16.625,"train/tensor_act_model_layers_1_residual_pre_attn/range":0.734375,"train/tensor_act_model_layers_62_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_72_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/mean":0.004058837890625,"train/tensor_param_model_layers_25_mlp_down_proj_weight/mean":-0.0001392364501953125,"train/tensor_act_model_layers_70_residual_post_attn/norm":573405.3125,"train/tensor_act_model_layers_60_mlp/norm":8192,"train/tensor_act_model_layers_47_residual_post_mlp/std":0.31640625,"train/tensor_param_model_layers_48_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/max":0.1328125,"train/tensor_param_model_layers_44_mlp_up_proj_weight/std":0.0341796875,"train/tensor_param_model_layers_96_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_40_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_24_residual_post_attn/min":94.5,"train/tensor_act_model_layers_30_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_4_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_71_residual_pre_attn/norm":581597.3125,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_54_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_96_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_53_residual_post_attn/range":3,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/max":0.12451171875,"train/tensor_param_model_layers_97_mlp_up_proj_weight/min":-0.1875,"train/tensor_act_model_layers_20_mlp_down_proj/norm":29992.9140625,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_99_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_28_residual_post_mlp/std":0.388671875,"train/tensor_act_model_layers_9_mlp_up_proj/max_abs":4.5625,"train/tensor_param_model_layers_49_mlp_down_proj_weight/max":0.12451171875,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_26_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_41_mlp_up_proj/range":8.34375,"train/tensor_act_model_layers_26_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_70_mlp_gate_proj_weight/std":0.0400390625,"train/tensor_act_model_layers_59_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_36_mlp_up_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_87_mlp_gate_proj/max_abs":5.09375,"train/tensor_act_model_layers_94_residual_post_attn/range":2,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/max":0.15625,"train/tensor_param_model_layers_72_mlp_up_proj_weight/max":0.1298828125,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_77_mlp/range":0,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_76_residual_pre_attn/max":304,"train/tensor_param_model_layers_38_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_56_mlp_up_proj_weight/max":0.130859375,"train/tensor_act_model_layers_99_residual_post_attn/range":2,"train/tensor_act_model_layers_81_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_mlp_down_proj/mean":14.8125,"train/tensor_param_model_layers_58_mlp_up_proj_weight/norm":6.75,"train/tensor_param_model_layers_64_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/max":9.5625,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_20_mlp_up_proj_weight/max":0.11962890625,"train/tensor_act_model_layers_88_mlp_up_proj/norm":5137.822265625,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_71_residual_pre_attn/min":282,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/max":0.125,"train/tensor_act_model_layers_30_residual_pre_attn/max_abs":121,"train/tensor_act_model_layers_94_mlp_down_proj/max_abs":19,"train/tensor_param_model_layers_55_mlp_up_proj_weight/min":-0.12890625,"train/tensor_act_model_layers_23_residual_post_attn/min":90.5,"train/tensor_grad_model_layers_8_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_30_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/max_abs":0.1240234375,"train/tensor_act_model_layers_42_residual_post_mlp/min":170,"train/tensor_grad_model_layers_38_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_36_mlp_gate_proj/norm":4991.59375,"train/tensor_act_model_layers_68_mlp_gate_proj/std":1.6015625,"train/tensor_act_model_layers_72_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_2_mlp_down_proj_weight/min":-0.11083984375,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_22_residual_pre_attn/max_abs":89,"train/tensor_act_model_layers_0_residual_post_attn/min":-0.33203125,"train/tensor_grad_model_layers_73_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_75_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_43_mlp_gate_proj/min":-1.2578125,"train/tensor_act_model_layers_92_mlp/max_abs":4,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_residual_pre_attn/norm":343916.65625,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/norm":6.5625,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_79_mlp_gate_proj/norm":5233.91748046875,"train/tensor_act_model_layers_87_residual_pre_attn/norm":712669.25,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_71_mlp/std":0,"train/tensor_act_model_layers_98_residual_post_attn/max_abs":392,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_34_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_41_mlp_gate_proj/max":4.53125,"train/tensor_act_model_layers_50_mlp/mean":4,"train/tensor_act_model_layers_23_mlp_gate_proj/max_abs":4.75,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_residual_pre_attn/range":3,"train/tensor_act_model_layers_59_residual_post_attn/norm":483100.78125,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/range":0.2451171875,"train/tensor_act_model_layers_78_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_88_mlp_down_proj/max_abs":18.125,"train/tensor_act_model_layers_71_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_1_mlp_down_proj/range":6.8125,"train/tensor_act_model_layers_84_residual_post_mlp/max_abs":340,"train/tensor_act_model_layers_27_mlp/max":4,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/min":-0.10888671875,"train/tensor_grad_model_layers_4_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_52_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_23_residual_post_mlp/norm":196369.25,"train/tensor_act_model_layers_10_residual_post_attn/max_abs":40.75,"train/tensor_param_model_layers_12_mlp_up_proj_weight/norm":5.5,"train/tensor_act_model_layers_55_mlp_up_proj/mean":-0.0302734375,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_65_mlp_down_proj_weight/max_abs":0.1279296875,"train/tensor_act_model_layers_9_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_97_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_mlp_gate_proj/max_abs":4.40625,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_19_residual_post_attn/max_abs":77,"train/tensor_param_model_layers_70_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_91_mlp_up_proj/mean":0.078125,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_62_mlp_down_proj_weight/min":-0.125,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_31_mlp_down_proj_weight/max":0.123046875,"train/tensor_act_model_layers_0_residual_post_attn/range":0.63671875,"train/tensor_act_model_layers_82_residual_pre_attn/max_abs":328,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_80_mlp_gate_proj/std":1.59375,"train/tensor_act_model_layers_9_residual_post_attn/min":35,"train/tensor_act_model_layers_80_residual_post_mlp/max_abs":324,"train/tensor_act_model_layers_78_residual_post_mlp/max":316,"train/tensor_act_model_layers_56_residual_post_mlp/std":0.408203125,"train/tensor_grad_model_layers_22_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/std":0.0294189453125,"train/tensor_act_model_layers_32_residual_post_mlp/mean":132,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_93_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_60_residual_post_mlp/range":3,"train/tensor_act_model_layers_57_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_89_mlp_up_proj_weight/max_abs":0.15625,"train/tensor_grad_model_layers_99_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_61_mlp_gate_proj/max_abs":4.46875,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_52_mlp_down_proj_weight/max_abs":0.1328125,"train/tensor_act_model_layers_46_mlp_gate_proj/norm":5061.31201171875,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_20_mlp_up_proj/mean":0.035888671875,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_11_mlp_down_proj_weight/min":-0.0986328125,"train/tensor_act_model_layers_12_mlp/min":4,"train/tensor_act_model_layers_82_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_51_mlp_up_proj/norm":5366.2958984375,"train/tensor_act_model_layers_63_mlp_up_proj/mean":-0.042236328125,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_48_mlp_gate_proj/max":5.3125,"train/tensor_act_model_layers_88_mlp_up_proj/range":9.09375,"train/tensor_act_model_layers_35_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_34_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_46_mlp/norm":8192,"train/tensor_grad_model_layers_40_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_54_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_76_mlp_up_proj_weight/std":0.04052734375,"train/tensor_act_model_layers_17_residual_pre_attn/range":2,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_62_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_72_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_50_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_94_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_52_mlp_up_proj_weight/min":-0.1484375,"train/tensor_act_model_layers_25_mlp/max_abs":4,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_7_mlp_gate_proj/norm":5076.39013671875,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/range":0.4111328125,"train/tensor_param_model_layers_69_mlp_down_proj_weight/std":0.037841796875,"train/tensor_act_model_layers_96_mlp_gate_proj/range":7.3671875,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_89_mlp_gate_proj/mean":0.671875,"train/tensor_act_model_layers_76_mlp_gate_proj/max":5.46875,"train/tensor_act_model_layers_90_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_58_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/max_abs":1448,"train/tensor_act_model_layers_34_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_16_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_63_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_8_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_99_mlp_gate_proj/mean":0.796875,"train/tensor_act_model_layers_27_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_37_residual_pre_attn/range":3,"train/tensor_act_model_layers_51_residual_pre_attn/std":0.31640625,"train/tensor_param_model_layers_78_mlp_up_proj_weight/max_abs":0.1318359375,"train/tensor_act_model_layers_65_residual_pre_attn/min":258,"train/tensor_act_model_layers_87_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_33_mlp_down_proj/std":0.443359375,"train/tensor_param_model_layers_39_mlp_down_proj_weight/norm":6.09375,"train/tensor_act_model_layers_9_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_80_mlp_down_proj_weight/max":0.1396484375,"train/tensor_grad_model_layers_98_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_13_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_37_mlp/max_abs":4,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_58_mlp_down_proj/norm":30673.953125,"train/tensor_param_model_layers_71_mlp_down_proj_weight/min":-0.1435546875,"train/tensor_act_model_layers_46_mlp_gate_proj/std":1.6484375,"train/tensor_act_model_layers_86_mlp_gate_proj/min":-1.2734375,"train/tensor_act_model_layers_87_residual_pre_attn/min":346,"train/tensor_param_model_layers_45_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_69_mlp/min":4,"train/tensor_act_model_layers_24_mlp_down_proj/max":16.5,"train/tensor_grad_model_layers_78_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_68_residual_pre_attn/range":2,"train/tensor_act_model_layers_91_mlp_up_proj/range":9.625,"train/tensor_act_model_layers_31_mlp/std":0,"train/tensor_act_model_layers_27_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_3_mlp_down_proj_weight/range":0.21484375,"train/tensor_act_model_layers_67_residual_post_mlp/mean":272,"train/tensor_act_model_layers_77_residual_post_attn/norm":630749.3125,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_65_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_40_mlp_up_proj/std":1.875,"train/tensor_act_model_layers_7_mlp_gate_proj/std":1.7109375,"train/tensor_act_model_layers_51_residual_pre_attn/min":202,"train/tensor_act_model_layers_34_mlp_gate_proj/range":6,"train/tensor_param_model_layers_75_mlp_down_proj_weight/mean":0.00069427490234375,"train/tensor_act_model_layers_55_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_47_mlp_down_proj/mean":14.75,"train/tensor_param_model_layers_46_mlp_up_proj_weight/max":0.126953125,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_24_mlp_down_proj_weight/norm":5.6875,"train/tensor_act_model_layers_8_residual_post_attn/max_abs":32.75,"train/tensor_act_model_layers_9_residual_post_attn/mean":36,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_88_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_97_mlp_gate_proj/std":1.6015625,"train/tensor_act_model_layers_60_mlp_up_proj/std":1.8515625,"train/tensor_act_model_layers_97_mlp_gate_proj/mean":0.63671875,"train/tensor_act_model_layers_96_mlp_gate_proj/max_abs":5.53125,"train/tensor_act_model_layers_22_residual_post_mlp/max":93,"train/tensor_param_model_layers_24_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_22_mlp_down_proj_weight/max":0.11572265625,"train/tensor_act_model_layers_33_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_61_mlp/range":0,"train/tensor_act_model_layers_8_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/min":-0.12890625,"train/tensor_act_model_layers_16_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_38_mlp_gate_proj/mean":0.6640625,"train/tensor_param_model_layers_72_mlp_down_proj_weight/std":0.03857421875,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/range":0.3720703125,"train/tensor_param_model_layers_74_mlp_up_proj_weight/norm":7.125,"train/tensor_param_model_layers_76_mlp_down_proj_weight/mean":0.00144195556640625,"train/tensor_param_model_layers_68_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_3_mlp_up_proj/max_abs":4.21875,"train/tensor_act_model_layers_8_mlp_down_proj/min":14.0625,"train/tensor_act_model_layers_19_mlp_gate_proj/range":5.8125,"train/tensor_act_model_layers_64_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_82_mlp/mean":4,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_23_mlp_up_proj_weight/min":-0.1142578125,"train/tensor_act_model_layers_94_residual_post_attn/mean":376,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_30_mlp_down_proj/norm":30068.234375,"train/tensor_param_model_layers_20_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_40_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_66_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_32_mlp_gate_proj/range":5.875,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/max_abs":0.1376953125,"train/tensor_act_model_layers_81_residual_pre_attn/mean":324,"train/tensor_act_model_layers_36_mlp/min":4,"train/tensor_act_model_layers_97_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_24_mlp/max":4,"train/tensor_act_model_layers_20_mlp_down_proj/mean":14.625,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/min":-0.14453125,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_89_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_84_mlp_gate_proj/mean":0.67578125,"train/tensor_act_model_layers_28_residual_post_mlp/range":2.5,"train/tensor_param_model_layers_55_mlp_up_proj_weight/max_abs":0.12890625,"train/tensor_act_model_layers_18_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_0_mlp/mean":4,"train/tensor_param_model_layers_69_mlp_down_proj_weight/range":0.26953125,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/mean":0.005584716796875,"train/tensor_grad_model_layers_81_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_residual_pre_attn/norm":474908.8125,"train/tensor_grad_model_layers_73_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_93_mlp_up_proj_weight/std":0.050537109375,"train/tensor_act_model_layers_51_residual_pre_attn/mean":204,"train/tensor_act_model_layers_83_residual_pre_attn/min":330,"train/tensor_act_model_layers_42_residual_pre_attn/range":3,"train/tensor_param_model_layers_54_mlp_up_proj_weight/std":0.03662109375,"train/tensor_act_model_layers_76_mlp/mean":4,"train/tensor_act_model_layers_16_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_84_mlp_gate_proj/min":-1.28125,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_28_mlp_gate_proj_weight/min":-0.1298828125,"train/tensor_act_model_layers_14_mlp/norm":8192,"train/tensor_param_model_layers_20_mlp_up_proj_weight/range":0.24462890625,"train/tensor_act_model_layers_54_residual_post_attn/min":214,"train/tensor_act_model_layers_90_mlp_gate_proj/norm":5291.59423828125,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_83_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_0_mlp/max":4,"train/tensor_act_model_layers_18_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_54_residual_post_mlp/norm":450348.8125,"train/tensor_act_model_layers_96_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_mlp_down_proj/std":0.50390625,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_46_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_57_mlp_down_proj/norm":30358.966796875,"train/global_step":3500,"train/tensor_grad_model_layers_36_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_67_residual_post_mlp/max_abs":272,"train/tensor_param_model_layers_49_mlp_down_proj_weight/min":-0.1298828125,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_26_mlp_down_proj_weight/mean":-0.00012063980102539062,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/std":0.04248046875,"train/tensor_act_model_layers_1_mlp/mean":4,"train/tensor_act_model_layers_86_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_99_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_47_mlp_down_proj_weight/std":0.034423828125,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_17_residual_post_attn/max_abs":69,"train/tensor_act_model_layers_13_residual_pre_attn/max_abs":53,"train/tensor_act_model_layers_56_mlp_down_proj/mean":14.8125,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_9_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_43_residual_post_mlp/range":3,"train/tensor_act_model_layers_82_residual_post_mlp/norm":679901.25,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_98_mlp_up_proj_weight/range":0.3662109375,"train/tensor_act_model_layers_82_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_57_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_50_residual_post_attn/min":198,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_50_mlp_down_proj/std":0.40234375,"train/tensor_param_model_layers_65_mlp_up_proj_weight/mean":0.000217437744140625,"train/tensor_param_model_layers_4_mlp_up_proj_weight/std":0.0299072265625,"train/tensor_act_model_layers_1_residual_post_mlp/min":7.5,"train/tensor_act_model_layers_50_residual_post_mlp/norm":417644.53125,"train/tensor_act_model_layers_61_mlp_up_proj/range":8.609375,"train/tensor_act_model_layers_49_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_78_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_42_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_37_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_54_mlp_gate_proj/range":6.015625,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/norm":8.9375,"train/tensor_act_model_layers_71_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_92_residual_post_mlp/max_abs":372,"train/tensor_act_model_layers_82_residual_post_attn/min":326,"train/tensor_act_model_layers_41_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_50_residual_post_mlp/range":3,"train/tensor_act_model_layers_8_residual_post_mlp/norm":73657.7421875,"train/tensor_param_model_layers_64_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/range":0,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_84_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_40_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_38_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_97_residual_post_attn/min":386,"train/tensor_act_model_layers_4_residual_post_mlp/max":20.625,"train/tensor_param_model_layers_33_mlp_down_proj_weight/max_abs":0.126953125,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_98_residual_post_mlp/max":396,"train/tensor_param_model_layers_34_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_38_mlp_up_proj_weight/max_abs":0.130859375,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_70_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_69_mlp_up_proj_weight/max":0.1279296875,"train/tensor_act_model_layers_64_mlp_down_proj/max_abs":16.625,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_44_mlp/max":4,"train/tensor_act_model_layers_38_mlp_down_proj/std":0.42578125,"train/tensor_act_model_layers_71_mlp/max_abs":4,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_37_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_86_residual_post_attn/max":344,"train/tensor_act_model_layers_48_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_62_residual_pre_attn/max":249,"train/tensor_param_model_layers_90_mlp_up_proj_weight/mean":-0.0006256103515625,"train/tensor_param_model_layers_23_mlp_down_proj_weight/norm":5.625,"train/tensor_act_model_layers_57_mlp_down_proj/max":16.125,"train/tensor_act_model_layers_21_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_2_mlp_up_proj/mean":0.047607421875,"train/tensor_grad_model_layers_71_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_86_residual_post_mlp/min":346,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_75_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_41_mlp_down_proj/range":2.0625,"train/tensor_act_model_layers_80_mlp/min":4,"train/tensor_grad_model_layers_65_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_1_mlp_down_proj_weight/mean":-0.002227783203125,"train/tensor_act_model_layers_90_mlp_gate_proj/min":-1.2578125,"train/tensor_param_model_layers_7_mlp_up_proj_weight/max_abs":0.103515625,"train/tensor_param_model_layers_31_mlp_down_proj_weight/norm":5.78125,"train/tensor_act_model_layers_97_residual_post_mlp/norm":802781.25,"train/tensor_param_model_layers_34_mlp_up_proj_weight/max":0.1220703125,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_98_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_mlp_down_proj/std":0.314453125,"train/tensor_act_model_layers_66_mlp_down_proj/range":2.8125,"train/tensor_act_model_layers_44_residual_pre_attn/max_abs":177,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_55_mlp_down_proj_weight/norm":6.6875,"train/tensor_act_model_layers_3_residual_pre_attn/norm":24558.56640625,"train/tensor_act_model_layers_54_residual_post_attn/norm":442156.84375,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_4_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_84_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_77_mlp_gate_proj_weight/max_abs":0.146484375,"train/tensor_act_model_layers_6_residual_post_attn/mean":24,"train/tensor_grad_model_layers_41_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_85_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_12_mlp_down_proj_weight/std":0.0301513671875,"train/tensor_grad_model_layers_6_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_46_mlp_up_proj/max":4.03125,"train/tensor_act_model_layers_53_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_74_mlp_up_proj_weight/max":0,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_12_residual_pre_attn/max":49,"train/tensor_act_model_layers_18_residual_post_mlp/min":74.5,"train/tensor_act_model_layers_88_residual_pre_attn/max":352,"train/tensor_param_model_layers_13_mlp_gate_proj_weight/norm":5.5,"train/tensor_act_model_layers_59_mlp_up_proj/norm":5182.17822265625,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_1_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp_up_proj/min":-4.1875,"train/tensor_act_model_layers_60_mlp_down_proj/min":14.25,"train/tensor_act_model_layers_45_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_81_residual_post_attn/max":324,"train/tensor_act_model_layers_30_mlp/max":4,"train/tensor_param_model_layers_85_mlp_gate_proj_weight/mean":0.00494384765625,"train/tensor_act_model_layers_92_mlp_gate_proj/std":1.6484375,"train/tensor_param_model_layers_45_mlp_down_proj_weight/norm":6.25,"train/tensor_param_model_layers_9_mlp_down_proj_weight/max_abs":0.11865234375,"train/tensor_act_model_layers_51_residual_post_attn/max":205,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_84_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_residual_post_attn/max_abs":304,"train/tensor_act_model_layers_31_mlp_down_proj/max_abs":16.125,"train/tensor_param_model_layers_11_mlp_down_proj_weight/max":0.111328125,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_41_mlp/min":4,"train/tensor_act_model_layers_39_mlp_up_proj/min":-4.15625,"train/tensor_act_model_layers_34_mlp_down_proj/norm":30057.802734375,"train/tensor_param_model_layers_71_mlp_down_proj_weight/max":0.134765625,"train/tensor_param_model_layers_76_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_80_mlp_down_proj_weight/mean":-0.000308990478515625,"train/tensor_act_model_layers_48_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_28_mlp/max_abs":4,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_22_mlp_down_proj_weight/min":-0.10595703125,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_68_residual_post_attn/min":270,"train/tensor_act_model_layers_61_mlp_gate_proj/max":4.46875,"train/tensor_act_model_layers_14_mlp/mean":4,"train/tensor_param_model_layers_7_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_1_residual_post_attn/min":3.484375,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_23_mlp/mean":4,"train/tensor_act_model_layers_68_residual_post_mlp/norm":565213.3125,"train/tensor_act_model_layers_65_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_24_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_16_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_20_residual_post_mlp/max":85,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/std":0.0301513671875,"train/tensor_act_model_layers_58_mlp_down_proj/mean":14.9375,"train/tensor_param_model_layers_76_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_40_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_49_mlp_up_proj_weight/mean":-0.00081634521484375,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_18_residual_pre_attn/max_abs":73,"train/tensor_param_model_layers_38_mlp_down_proj_weight/range":0.24609375,"train/tensor_grad_model_layers_80_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_33_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_80_mlp_down_proj_weight/min":-0.1416015625,"train/tensor_act_model_layers_9_residual_pre_attn/max":36.75,"train/tensor_param_model_layers_92_mlp_down_proj_weight/mean":-9.822845458984375e-05,"train/tensor_act_model_layers_36_mlp_up_proj/max_abs":4.4375,"train/tensor_act_model_layers_51_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_12_residual_post_attn/max_abs":49,"train/tensor_act_model_layers_63_residual_pre_attn/min":250,"train/tensor_param_model_layers_21_mlp_up_proj_weight/norm":5.6875,"train/tensor_act_model_layers_81_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_92_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_28_mlp_gate_proj/norm":5229.5458984375,"train/tensor_act_model_layers_48_mlp_up_proj/norm":5307.703125,"train/tensor_param_model_layers_47_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_32_mlp/norm":8192,"train/tensor_act_model_layers_51_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_35_mlp_gate_proj/std":1.640625,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/range":3808,"train/tensor_act_model_layers_97_residual_post_mlp/range":2,"train/tensor_grad_model_layers_98_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_78_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_49_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_60_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_57_residual_post_mlp/max_abs":233,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_81_mlp_gate_proj/mean":0.640625,"train/tensor_act_model_layers_77_mlp_gate_proj/max_abs":4.78125,"train/tensor_param_model_layers_93_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_58_residual_post_attn/range":3,"train/tensor_act_model_layers_32_mlp_gate_proj/std":1.7109375,"train/tensor_grad_model_layers_53_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_53_mlp_gate_proj_weight/min":-0.126953125,"train/tensor_grad_model_layers_47_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_66_mlp_up_proj/mean":-0.042724609375,"train/tensor_grad_model_layers_16_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_86_residual_pre_attn/mean":344,"train/tensor_param_model_layers_76_mlp_up_proj_weight/mean":0.000640869140625,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_98_mlp/range":0,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_63_residual_pre_attn/std":0.416015625,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_20_mlp_up_proj/max":4.34375,"train/tensor_act_model_layers_42_mlp/mean":4,"train/tensor_act_model_layers_72_mlp/range":0,"train/tensor_act_model_layers_16_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_89_residual_post_attn/min":354,"train/tensor_act_model_layers_76_residual_post_mlp/max_abs":308,"train/tensor_act_model_layers_5_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_residual_post_attn/range":2.5,"train/tensor_act_model_layers_99_residual_pre_attn/range":2,"train/tensor_act_model_layers_99_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_91_residual_pre_attn/min":362,"train/tensor_param_model_layers_87_mlp_down_proj_weight/range":0.27734375,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_49_mlp_up_proj/mean":-0.10498046875,"train/tensor_act_model_layers_94_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_54_mlp_up_proj/norm":5410.81494140625,"train/tensor_act_model_layers_85_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_40_mlp_down_proj/std":0.4296875,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_35_mlp/range":0,"train/tensor_act_model_layers_71_mlp_up_proj/mean":0.0225830078125,"train/tensor_act_model_layers_4_residual_post_mlp/mean":20,"train/tensor_act_model_layers_3_mlp_up_proj/std":1.8828125,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_30_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/mean":0.0045166015625,"train/tensor_grad_model_layers_94_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_34_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_69_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_residual_post_mlp/mean":60,"train/tensor_act_model_layers_77_residual_post_mlp/mean":312,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_6_mlp/mean":4,"train/tensor_act_model_layers_55_mlp_up_proj/std":1.8671875,"train/tensor_param_model_layers_31_mlp_down_proj_weight/min":-0.1328125,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/range":0.263671875,"train/tensor_param_model_layers_62_mlp_down_proj_weight/norm":6.6875,"train/tensor_act_model_layers_89_residual_post_mlp/norm":737245.25,"train/tensor_param_model_layers_62_mlp_down_proj_weight/max":0.126953125,"train/tensor_param_model_layers_37_mlp_down_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_96_mlp_gate_proj/norm":5156.0205078125,"train/tensor_act_model_layers_91_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_3_mlp/max":4,"train/tensor_grad_model_layers_33_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_3_mlp_down_proj_weight/mean":-0.0001277923583984375,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_21_mlp_down_proj_weight/max_abs":0.119140625,"train/tensor_act_model_layers_56_mlp_down_proj/norm":30387.314453125,"train/tensor_act_model_layers_14_residual_pre_attn/range":2,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_91_mlp_gate_proj/max":5.0625,"train/tensor_act_model_layers_36_mlp_up_proj/max":4.4375,"train/tensor_param_model_layers_45_mlp_up_proj_weight/max":0.12890625,"train/tensor_act_model_layers_83_residual_pre_attn/max_abs":332,"train/tensor_param_model_layers_32_mlp_down_proj_weight/std":0.031982421875,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_55_mlp_down_proj_weight/range":0.265625,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_70_residual_post_mlp/mean":284,"train/tensor_grad_model_layers_7_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_52_mlp_gate_proj/mean":0.6875,"train/tensor_act_model_layers_47_mlp/max":4,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_27_mlp_down_proj_weight/norm":5.71875,"train/tensor_param_model_layers_56_mlp_down_proj_weight/std":0.036376953125,"train/tensor_act_model_layers_52_residual_post_mlp/max":213,"train/tensor_act_model_layers_19_residual_post_mlp/std":0.357421875,"train/tensor_act_model_layers_70_mlp_up_proj/std":1.8359375,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_71_mlp_up_proj_weight/max":0.1357421875,"train/tensor_act_model_layers_2_residual_post_mlp/mean":12,"train/tensor_param_model_layers_93_mlp_up_proj_weight/norm":9.125,"train/tensor_param_model_layers_65_mlp_up_proj_weight/range":0.2685546875,"train/tensor_param_model_layers_11_mlp_gate_proj_weight/norm":5.46875,"train/tensor_param_model_layers_56_mlp_up_proj_weight/range":0.2568359375,"train/tensor_act_model_layers_45_residual_post_mlp/min":182,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_58_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_54_residual_pre_attn/norm":442156.84375,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_58_residual_post_attn/min":230,"train/tensor_act_model_layers_51_residual_pre_attn/range":3,"train/tensor_act_model_layers_38_mlp_down_proj/max":16,"train/tensor_act_model_layers_65_residual_pre_attn/mean":260,"train/tensor_act_model_layers_77_mlp_up_proj/max_abs":4.71875,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/max":0.19921875,"train/tensor_act_model_layers_58_mlp_gate_proj/min":-1.25,"train/tensor_act_model_layers_43_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_99_mlp_gate_proj/max":5.84375,"train/tensor_act_model_layers_93_residual_pre_attn/max_abs":372,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/norm":5.75,"train/tensor_grad_model_layers_73_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_17_residual_pre_attn/std":0.34375,"train/tensor_param_model_layers_98_mlp_up_proj_weight/max_abs":0.1982421875,"train/tensor_act_model_layers_29_mlp_gate_proj/max_abs":5.0625,"train/tensor_act_model_layers_87_mlp_gate_proj/mean":0.48828125,"train/tensor_act_model_layers_45_mlp_down_proj/mean":14.8125,"train/tensor_act_model_layers_75_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_70_mlp_up_proj_weight/mean":0.000415802001953125,"train/tensor_act_model_layers_89_residual_pre_attn/max_abs":356,"train/tensor_act_model_layers_16_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_68_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_76_mlp_down_proj/range":2.5625,"train/tensor_act_model_layers_77_mlp/max":4,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_16_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_47_mlp_up_proj/max_abs":4.28125,"train/tensor_param_model_layers_61_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_80_mlp_up_proj/norm":5110.10888671875,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_1_mlp_up_proj_weight/range":0.25439453125,"train/tensor_param_model_layers_47_mlp_gate_proj_weight/norm":6.53125,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_13_mlp/std":0,"train/tensor_act_model_layers_16_mlp_gate_proj/range":6.328125,"train/tensor_act_model_layers_75_mlp_gate_proj/std":1.640625,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_3_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_85_mlp_up_proj_weight/max":0.146484375,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_17_residual_pre_attn/min":67,"train/tensor_param_model_layers_35_mlp_up_proj_weight/min":-0.12109375,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/std":0.05078125,"train/tensor_act_model_layers_31_mlp_gate_proj/std":1.7578125,"train/tensor_act_model_layers_54_residual_post_attn/mean":216,"train/tensor_grad_model_layers_0_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_19_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_52_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_65_mlp_up_proj/range":9.21875,"train/tensor_act_model_layers_44_mlp_down_proj/norm":30109.029296875,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/std":0.058349609375,"train/tensor_grad_model_layers_58_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_83_mlp_down_proj/min":14,"train/tensor_grad_model_layers_17_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_92_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_mlp_down_proj/mean":14.9375,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_5_residual_pre_attn/min":19.25,"train/tensor_param_model_layers_29_mlp_gate_proj_weight/norm":5.9375,"train/tensor_act_model_layers_50_mlp_down_proj/max":16.25,"train/tensor_act_model_layers_58_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_98_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_45_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_83_mlp/mean":4,"train/tensor_grad_model_layers_94_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_18_mlp_gate_proj/std":1.703125,"train/tensor_act_model_layers_69_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_15_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_21_mlp/max_abs":4,"train/tensor_param_model_layers_80_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_74_mlp_gate_proj/max":4.6875,"train/tensor_act_model_layers_74_mlp_gate_proj/range":5.96875,"train_steps_per_second":9.497,"train/tensor_param_model_layers_49_mlp_up_proj_weight/max":0.1328125,"train/tensor_param_model_layers_64_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_71_mlp_down_proj/max":17,"train/tensor_act_model_layers_70_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_97_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_82_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_66_mlp_gate_proj/min":-1.265625,"train/tensor_act_model_layers_25_mlp_gate_proj/max":4.59375,"train/tensor_act_model_layers_97_residual_pre_attn/norm":794589.25,"train/tensor_param_model_layers_83_mlp_up_proj_weight/std":0.044921875,"train/tensor_act_model_layers_82_mlp_up_proj/min":-4.8125,"train/tensor_act_model_layers_82_residual_post_mlp/max":332,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_86_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_96_residual_post_mlp/max_abs":388,"train/tensor_act_model_layers_44_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_residual_post_mlp/max":28.625,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/max_abs":0.1318359375,"train/tensor_grad_model_layers_53_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_55_mlp_down_proj/max_abs":16.375,"train/tensor_act_model_layers_32_mlp_up_proj/min":-4.4375,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/max_abs":0,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/min":-0.12255859375,"train/tensor_act_model_layers_9_mlp_up_proj/std":1.8671875,"train/tensor_act_model_layers_47_mlp_up_proj/range":8.140625,"train/tensor_act_model_layers_82_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_0_mlp_up_proj_weight/norm":11.5,"train/tensor_act_model_layers_38_residual_pre_attn/mean":152,"train/tensor_act_model_layers_21_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_10_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_66_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_82_residual_pre_attn/norm":671709.25,"train/tensor_param_model_layers_47_mlp_down_proj_weight/mean":-0.001190185546875,"train/tensor_act_model_layers_29_mlp_gate_proj/min":-1.3125,"train/tensor_act_model_layers_55_mlp/std":0,"train/tensor_act_model_layers_88_mlp_gate_proj/mean":0.61328125,"train/tensor_param_model_layers_0_mlp_down_proj_weight/std":0.06201171875,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_17_mlp/max_abs":4,"train/tensor_act_model_layers_35_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_3_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/min":-0.11865234375,"train/tensor_grad_model_layers_61_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_90_residual_post_mlp/max":364,"train/tensor_param_model_layers_17_mlp_up_proj_weight/range":0.228515625,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_79_mlp_down_proj_weight/std":0.04150390625,"train/tensor_act_model_layers_40_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_67_mlp_down_proj_weight/range":0.2578125,"train/tensor_grad_model_layers_57_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_16_residual_post_attn/max_abs":65,"train/tensor_grad_model_layers_98_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_60_mlp_down_proj_weight/mean":-0.0024261474609375,"train/tensor_act_model_layers_13_mlp/range":0,"train/tensor_act_model_layers_1_residual_pre_attn/mean":4,"train/tensor_param_model_layers_55_mlp_gate_proj_weight/mean":0.003448486328125,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_73_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_50_residual_pre_attn/norm":409452.5625,"train/tensor_act_model_layers_11_mlp_down_proj/range":1.6875,"train/tensor_act_model_layers_72_residual_post_attn/max_abs":288,"train/tensor_act_model_layers_32_mlp_down_proj/mean":14.625,"train/tensor_act_model_layers_93_residual_pre_attn/mean":372,"train/tensor_param_model_layers_3_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_61_residual_post_mlp/max":249,"train/tensor_act_model_layers_55_residual_post_mlp/std":0.408203125,"train/tensor_act_model_layers_29_mlp_down_proj/max_abs":16.625,"train/tensor_act_model_layers_4_mlp/norm":8192,"train/tensor_act_model_layers_23_residual_pre_attn/max_abs":93,"train/tensor_act_model_layers_92_mlp/std":0,"train/tensor_act_model_layers_38_residual_post_attn/std":0.31640625,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/norm":7.0625,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/max_abs":0.1953125,"train/tensor_act_model_layers_96_mlp/max_abs":4,"train/tensor_act_model_layers_41_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_18_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_11_mlp_down_proj/max_abs":15.875,"train/tensor_act_model_layers_86_mlp_down_proj/norm":30781.13671875,"train/tensor_act_model_layers_10_residual_post_attn/range":1.75,"train/tensor_act_model_layers_6_residual_post_attn/norm":49105.953125,"train/tensor_act_model_layers_77_mlp_down_proj/norm":30954.673828125,"train/tensor_grad_model_layers_46_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_89_residual_pre_attn/range":2,"train/tensor_param_model_layers_4_mlp_up_proj_weight/max_abs":0.10888671875,"train/tensor_param_model_layers_62_mlp_up_proj_weight/mean":0.000736236572265625,"train/tensor_act_model_layers_23_residual_post_mlp/max":97,"train/tensor_grad_model_layers_9_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_11_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_31_mlp_up_proj_weight/max":0.1298828125,"train/tensor_grad_model_layers_62_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_83_mlp_down_proj_weight/mean":-0.0003108978271484375,"train/tensor_act_model_layers_70_mlp_up_proj/max_abs":4.375,"train/tensor_param_model_layers_93_mlp_down_proj_weight/min":-0.1630859375,"train/tensor_act_model_layers_9_mlp/min":4,"train/tensor_grad_model_layers_85_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_2_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_55_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_52_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_79_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_26_mlp_up_proj/max":5.0625,"train/tensor_act_model_layers_40_mlp_down_proj/max":16.375,"train/tensor_act_model_layers_23_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_46_mlp_down_proj_weight/norm":6.25,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/max":0.12255859375,"train/tensor_act_model_layers_93_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_68_residual_pre_attn/norm":557021.3125,"train/tensor_act_model_layers_47_mlp_gate_proj/max":4.625,"train/tensor_act_model_layers_44_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_59_mlp_gate_proj/norm":4922.2783203125,"train/tensor_grad_model_layers_68_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_88_mlp_down_proj_weight/max":0.1416015625,"train/tensor_param_model_layers_92_mlp_up_proj_weight/std":0.04931640625,"train/tensor_param_model_layers_46_mlp_up_proj_weight/norm":6.28125,"train/tensor_param_model_layers_1_mlp_up_proj_weight/max_abs":0.1357421875,"train/tensor_act_model_layers_32_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_37_mlp_up_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_82_mlp_up_proj_weight/min":-0.14453125,"train/tensor_act_model_layers_74_residual_post_mlp/range":2,"train/tensor_param_model_layers_44_mlp_down_proj_weight/max":0.1279296875,"train/tensor_param_model_layers_39_mlp_down_proj_weight/range":0.25390625,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_16_mlp_up_proj/max":4.1875,"train/tensor_act_model_layers_82_mlp_up_proj/mean":0.006744384765625,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_60_mlp_down_proj/norm":30340.06640625,"train/tensor_param_model_layers_85_mlp_down_proj_weight/norm":8.125,"train/tensor_grad_model_layers_91_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_65_residual_post_attn/max":260,"train/tensor_act_model_layers_86_mlp_up_proj/mean":0.09814453125,"train/tensor_act_model_layers_25_mlp_up_proj/range":7.75,"train/tensor_grad_model_layers_13_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_55_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_94_residual_post_attn/norm":770013.25,"train/tensor_act_model_layers_74_mlp_up_proj/range":9.125,"train/tensor_act_model_layers_79_mlp/max_abs":4,"train/tensor_param_model_layers_76_mlp_down_proj_weight/range":0.26953125,"train/tensor_param_model_layers_54_mlp_down_proj_weight/min":-0.1259765625,"train/tensor_act_model_layers_69_residual_post_mlp/mean":280,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/max":0.166015625,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/mean":0.005279541015625,"train/tensor_act_model_layers_76_mlp_down_proj/max_abs":16.75,"train/tensor_act_model_layers_22_mlp_up_proj/max_abs":4.125,"train/tensor_act_model_layers_32_residual_post_mlp/max":133,"train/tensor_act_model_layers_89_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_34_mlp_up_proj/min":-3.953125,"train/tensor_param_model_layers_77_mlp_down_proj_weight/min":-0.1357421875,"train/tensor_act_model_layers_61_residual_post_mlp/mean":248,"train/tensor_param_model_layers_32_mlp_up_proj_weight/mean":-0.0003261566162109375,"train/tensor_act_model_layers_87_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_13_mlp_gate_proj/max_abs":4.625,"train/tensor_act_model_layers_76_mlp/norm":8192,"train/tensor_act_model_layers_65_mlp_gate_proj/max":4.96875,"train/tensor_grad_model_layers_29_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_0_mlp_gate_proj/min":-6.6875,"train/tensor_act_model_layers_30_mlp_up_proj/max_abs":4.53125,"train/tensor_act_model_layers_83_mlp_up_proj/std":1.828125,"train/tensor_act_model_layers_95_residual_pre_attn/max_abs":380,"train/tensor_act_model_layers_38_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_57_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_33_mlp_up_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_8_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_mlp_down_proj/min":14.1875,"train/tensor_grad_model_layers_56_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_20_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_70_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_68_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_93_residual_post_attn/max":372,"train/tensor_act_model_layers_70_residual_post_attn/range":2,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_60_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_19_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_89_residual_pre_attn/norm":729053.25,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/mean":0,"train/tensor_param_model_layers_40_mlp_up_proj_weight/norm":6.125,"train/tensor_grad_model_layers_70_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_45_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_70_mlp_up_proj_weight/max_abs":0.13671875,"train/tensor_act_model_layers_6_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_residual_post_attn/max":153,"train/tensor_grad_model_layers_51_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_31_mlp_down_proj/frac_near_user_limit":1,"train/tensor_act_model_layers_61_residual_post_attn/range":3,"train/tensor_act_model_layers_66_mlp/range":0,"train/tensor_act_model_layers_15_residual_post_mlp/max_abs":65,"train/tensor_act_model_layers_22_mlp_gate_proj/std":1.703125,"train/tensor_act_model_layers_80_mlp_down_proj/norm":30883.10546875,"train/tensor_act_model_layers_87_mlp_gate_proj/min":-1.3125,"train/tensor_act_model_layers_87_mlp_up_proj/mean":0.25,"train/tensor_param_model_layers_15_mlp_down_proj_weight/norm":5.4375,"train/tensor_grad_model_layers_39_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_82_residual_pre_attn/range":2,"train/tensor_grad_model_layers_17_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_48_mlp_down_proj_weight/mean":0.00038909912109375,"train/tensor_act_model_layers_2_mlp_up_proj/std":1.9921875,"train/tensor_act_model_layers_11_residual_pre_attn/min":43,"train/tensor_param_model_layers_37_mlp_down_proj_weight/min":-0.12451171875,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_53_mlp_up_proj/mean":-0.0703125,"train/tensor_param_model_layers_89_mlp_gate_proj_weight/max_abs":0.1611328125,"train/tensor_act_model_layers_63_mlp/norm":8192,"train/tensor_act_model_layers_62_mlp_up_proj/min":-4.75,"train/tensor_act_model_layers_62_residual_post_attn/range":3,"train/tensor_param_model_layers_0_mlp_gate_proj_weight/max":0.125,"train/tensor_param_model_layers_13_mlp_up_proj_weight/norm":5.46875,"train/tensor_param_model_layers_14_mlp_up_proj_weight/range":0.23486328125,"train/tensor_act_model_layers_59_residual_post_mlp/std":0.416015625,"train/tensor_act_model_layers_99_mlp_up_proj/range":11.28125,"train/tensor_act_model_layers_60_residual_pre_attn/min":238,"train/tensor_act_model_layers_74_mlp_gate_proj/mean":0.625,"train/tensor_param_model_layers_32_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_7_mlp_up_proj/std":1.921875,"train/tensor_act_model_layers_19_mlp/mean":4,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_68_mlp_gate_proj/mean":0.5625,"train/tensor_act_model_layers_92_mlp_up_proj/max":4.875,"train/tensor_param_model_layers_92_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/min":-0.1318359375,"train/tensor_param_model_layers_75_mlp_gate_proj_weight/mean":0.004730224609375,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_44_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_11_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_24_mlp/std":0,"train/tensor_act_model_layers_5_mlp_gate_proj/norm":5106.2294921875,"train/tensor_grad_model_layers_65_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_29_residual_post_mlp/std":0.388671875,"train/tensor_act_model_layers_53_mlp_down_proj/min":14.25,"train/tensor_grad_model_layers_49_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_11_residual_pre_attn/std":0.224609375,"train/tensor_act_model_layers_17_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_39_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_21_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_mlp_gate_proj/max":4.28125,"train/tensor_act_model_layers_60_residual_pre_attn/std":0.416015625,"train/tensor_act_model_layers_21_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_96_mlp/std":0,"train/tensor_act_model_layers_65_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_residual_pre_attn/range":3,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_49_mlp_up_proj/range":8.8125,"train/tensor_act_model_layers_13_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_4_mlp_gate_proj/norm":5014.06884765625,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_16_mlp_gate_proj/std":1.734375,"train/tensor_param_model_layers_50_mlp_gate_proj_weight/norm":6.75,"train/tensor_param_model_layers_8_mlp_down_proj_weight/max":0.10205078125,"train/tensor_act_model_layers_80_residual_pre_attn/max":320,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_20_mlp_up_proj/max_abs":4.90625,"train/tensor_act_model_layers_64_residual_post_mlp/max":260,"train/tensor_act_model_layers_92_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_9_mlp_gate_proj_weight/std":0.030029296875,"train/tensor_grad_model_layers_84_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_83_mlp_up_proj_weight/range":0.2802734375,"train/tensor_act_model_layers_62_mlp/min":4,"train/tensor_act_model_layers_28_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_89_mlp_up_proj/range":9,"train/tensor_param_model_layers_40_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_24_residual_post_mlp/max_abs":101,"train/tensor_act_model_layers_26_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_87_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_8_residual_post_attn/mean":32,"train/tensor_param_model_layers_18_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_30_mlp_down_proj_weight/norm":5.78125,"train/tensor_grad_model_layers_37_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_37_residual_post_mlp/max":153,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_78_mlp_down_proj/min":14.125,"train/tensor_grad_model_layers_5_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_59_residual_post_attn/mean":236,"train/tensor_grad_model_layers_93_mlp_down_proj_weight/norm":0,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/max_abs":0,"train/tensor_act_model_layers_38_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_44_mlp_up_proj/max":5.0625,"train/tensor_param_model_layers_86_mlp_gate_proj_weight/norm":8.8125,"train/tensor_param_model_layers_59_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_6_mlp_up_proj/mean":0.0196533203125,"train/tensor_act_model_layers_72_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_5_mlp/norm":8192,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_56_mlp_down_proj/frac_near_user_limit":1,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_4_residual_pre_attn/max_abs":16.625,"train/tensor_act_model_layers_1_residual_post_attn/max_abs":4.40625,"train/tensor_act_model_layers_35_mlp_down_proj/norm":30122.537109375,"train/tensor_act_model_layers_79_mlp_gate_proj/max":4.59375,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/range":0.22216796875,"train/tensor_grad_model_layers_91_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_81_mlp_up_proj/std":1.8203125,"train/tensor_act_model_layers_8_mlp_down_proj/norm":29916.330078125,"train/tensor_grad_model_layers_79_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_57_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_26_mlp_up_proj_weight/mean":0.00049591064453125,"train/tensor_act_model_layers_57_mlp/range":0,"train/tensor_grad_model_layers_10_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/std":0.03125,"train/tensor_act_model_layers_58_mlp_up_proj/std":1.8515625,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/max":0.1220703125,"train/tensor_act_model_layers_52_mlp_up_proj/range":9.5,"train/tensor_param_model_layers_21_mlp_down_proj_weight/mean":0.000156402587890625,"train/tensor_act_model_layers_73_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_53_mlp_down_proj_weight/range":0.2548828125,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_32_mlp_down_proj/norm":29934.494140625,"train/tensor_act_model_layers_28_mlp/norm":8192,"train/tensor_act_model_layers_47_mlp_up_proj/mean":-0.36328125,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/max":0.1240234375,"train/tensor_param_model_layers_37_mlp_gate_proj_weight/range":0.25390625,"train/tensor_act_model_layers_56_mlp_gate_proj/range":5.4609375,"train/tensor_act_model_layers_92_residual_pre_attn/max":368,"train/tensor_grad_model_layers_42_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_43_residual_pre_attn/mean":172,"train/tensor_param_model_layers_23_mlp_up_proj_weight/range":0.23779296875,"train/tensor_act_model_layers_91_mlp/mean":4,"train/tensor_act_model_layers_38_mlp_up_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_13_mlp_gate_proj/std":1.71875,"train/tensor_param_model_layers_45_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_77_residual_post_attn/mean":308,"train/tensor_act_model_layers_18_mlp_gate_proj/range":6.0078125,"train/tensor_act_model_layers_59_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_67_mlp_up_proj/min":-4.625,"train/tensor_act_model_layers_76_mlp_gate_proj/norm":4815.46826171875,"train/tensor_act_model_layers_7_residual_pre_attn/std":0.1953125,"train/tensor_param_model_layers_53_mlp_up_proj_weight/norm":6.65625,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_30_mlp_down_proj/max":16.25,"train/tensor_act_model_layers_3_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_75_mlp_down_proj_weight/max_abs":0.14453125,"train/tensor_act_model_layers_26_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/min":-0.220703125,"train/tensor_act_model_layers_71_residual_post_mlp/norm":589789.25,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/std":0.03173828125,"train/tensor_act_model_layers_71_residual_post_attn/min":282,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_56_mlp_down_proj_weight/mean":-0.00201416015625,"train/tensor_grad_model_layers_92_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_50_mlp_down_proj_weight/norm":0,"train/tensor_param_model_layers_21_mlp_down_proj_weight/min":-0.119140625,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_0_residual_pre_attn/mean":0.00106048583984375,"train/tensor_act_model_layers_19_mlp_gate_proj/max_abs":4.53125,"train/tensor_grad_model_layers_35_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_16_residual_pre_attn/std":0.298828125,"train/tensor_act_model_layers_45_mlp_down_proj/norm":30380.31640625,"train/tensor_act_model_layers_96_residual_pre_attn/min":382,"train/tensor_param_model_layers_10_mlp_up_proj_weight/std":0.0301513671875,"train/tensor_param_model_layers_74_mlp_up_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_42_mlp_up_proj_weight/norm":6.1875,"train/tensor_act_model_layers_81_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_82_mlp_up_proj_weight/std":0,"train/tensor_param_model_layers_41_mlp_up_proj_weight/max":0.12890625,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_77_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_90_residual_post_mlp/max_abs":364,"train/tensor_grad_model_layers_31_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_99_residual_post_mlp/min":398,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_0_mlp_up_proj_weight/mean":1.8715858459472656e-05,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_76_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_94_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_29_mlp_gate_proj/mean":0.57421875,"train/tensor_param_model_layers_95_mlp_up_proj_weight/norm":9.375,"train/tensor_act_model_layers_58_residual_post_attn/max_abs":233,"train/tensor_grad_model_layers_97_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_25_residual_post_mlp/min":102.5,"train/tensor_param_model_layers_90_mlp_up_proj_weight/min":-0.1494140625,"train/tensor_act_model_layers_24_residual_pre_attn/norm":196369.25,"train/tensor_act_model_layers_2_mlp/max":4,"train/tensor_act_model_layers_45_mlp/max_abs":4,"train/tensor_grad_model_layers_77_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_94_residual_pre_attn/min":374,"train/tensor_act_model_layers_33_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_69_mlp_gate_proj_weight/mean":0,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/range":0,"train/tensor_param_model_layers_89_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_0_mlp/min":3.703125,"train/tensor_act_model_layers_99_residual_post_mlp/max_abs":400,"train/tensor_act_model_layers_17_mlp_up_proj/max_abs":4.5625,"train/tensor_grad_model_layers_94_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_1_residual_post_mlp/std":0.11376953125,"train/tensor_act_model_layers_71_mlp_down_proj/norm":30565.375,"train/tensor_act_model_layers_96_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_61_mlp_up_proj_weight/range":0.26171875,"train/tensor_act_model_layers_16_residual_pre_attn/min":63,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_51_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_44_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_9_residual_pre_attn/mean":36,"train/tensor_act_model_layers_60_residual_pre_attn/max_abs":241,"train/tensor_param_model_layers_81_mlp_up_proj_weight/max":0.1435546875,"train/tensor_act_model_layers_14_mlp/max_abs":4,"train/tensor_act_model_layers_21_residual_pre_attn/max_abs":85,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_24_mlp_down_proj/std":0.38671875,"train/tensor_param_model_layers_66_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_7_residual_pre_attn/norm":57297.72265625,"train/tensor_grad_model_layers_89_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_43_residual_post_attn/max":173,"train/tensor_act_model_layers_38_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_45_mlp_gate_proj/std":1.6796875,"train/tensor_act_model_layers_11_mlp/std":0,"train/tensor_act_model_layers_61_mlp/max":4,"train/tensor_grad_model_layers_22_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_79_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_94_mlp_gate_proj/mean":0.671875,"train/tensor_act_model_layers_89_mlp_up_proj/max_abs":4.8125,"train/tensor_act_model_layers_95_residual_post_mlp/min":382,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_58_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_6_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_92_residual_post_mlp/mean":372,"train/tensor_act_model_layers_35_mlp/min":4,"train/tensor_act_model_layers_33_mlp_down_proj/max":16.5,"train/tensor_act_model_layers_42_mlp_gate_proj/min":-1.34375,"train/tensor_act_model_layers_80_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_grad_model_layers_71_mlp_up_proj_weight/mean":0,"train/tensor_param_model_layers_91_mlp_down_proj_weight/norm":8.625,"train/tensor_act_model_layers_62_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_7_mlp_gate_proj_weight/std":0.029541015625,"train/tensor_grad_model_layers_90_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_80_mlp_down_proj/mean":15.0625,"train/tensor_param_model_layers_45_mlp_up_proj_weight/norm":6.3125,"train/tensor_param_model_layers_49_mlp_down_proj_weight/range":0.25439453125,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/norm":5.78125,"train/tensor_grad_model_layers_54_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_42_residual_post_mlp/norm":352108.65625,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_99_mlp_gate_proj/max_abs":5.84375,"train/tensor_grad_model_layers_24_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_5_mlp_down_proj/max_abs":16.5,"train/tensor_act_model_layers_87_residual_post_attn/min":346,"train/tensor_act_model_layers_40_residual_post_mlp/std":0.31640625,"train/tensor_act_model_layers_76_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_6_mlp_up_proj_weight/std":0.0296630859375,"train/tensor_act_model_layers_94_mlp_gate_proj/std":1.5859375,"train/tensor_act_model_layers_73_residual_pre_attn/min":290,"train/tensor_act_model_layers_77_mlp_gate_proj/std":1.6640625,"train/tensor_param_model_layers_34_mlp_up_proj_weight/max_abs":0.1220703125,"train/tensor_act_model_layers_48_residual_post_attn/mean":192,"train/tensor_grad_model_layers_7_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_27_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_16_mlp_down_proj_weight/max_abs":0.12158203125,"train/tensor_act_model_layers_73_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_23_residual_pre_attn/norm":188177.328125,"train/tensor_grad_model_layers_14_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_37_mlp_down_proj_weight/max":0.125,"train/tensor_grad_model_layers_81_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_38_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_param_model_layers_36_mlp_up_proj_weight/std":0.033203125,"train/tensor_act_model_layers_92_mlp_up_proj/norm":5340.46533203125,"train/tensor_act_model_layers_2_residual_post_attn/max":8.4375,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/max":0.1318359375,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_38_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_22_mlp_up_proj/std":1.8515625,"train/tensor_param_model_layers_4_mlp_down_proj_weight/max":0.107421875,"train/tensor_param_model_layers_34_mlp_gate_proj_weight/mean":0.004425048828125,"train/tensor_grad_model_layers_87_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_2_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_22_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_act_model_layers_53_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_42_residual_post_attn/min":166,"train/tensor_act_model_layers_52_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_residual_post_mlp/norm":720861.25,"train/tensor_act_model_layers_19_mlp_up_proj/norm":5418.77783203125,"train/tensor_act_model_layers_50_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_mlp_gate_proj/min":-1.3359375,"train/tensor_grad_model_layers_47_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_91_mlp_up_proj/std":1.8359375,"train/tensor_act_model_layers_90_residual_post_mlp/range":2,"train/tensor_act_model_layers_30_residual_post_mlp/min":122.5,"train/tensor_act_model_layers_35_residual_post_attn/range":3,"train/tensor_param_model_layers_93_mlp_down_proj_weight/max":0.1630859375,"train/tensor_act_model_layers_71_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/min":-0.125,"train/tensor_act_model_layers_85_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_67_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_52_mlp_gate_proj/max":4.75,"train/tensor_param_model_layers_93_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_68_mlp_gate_proj_weight/mean":0.00439453125,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/max_abs":0.123046875,"train/tensor_act_model_layers_78_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_25_mlp_up_proj/norm":5515.9169921875,"train/tensor_param_model_layers_74_mlp_down_proj_weight/range":0.259765625,"train/tensor_act_model_layers_94_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_86_mlp_down_proj/mean":15,"train/tensor_act_model_layers_77_mlp_down_proj/mean":15.125,"train/tensor_act_model_layers_79_mlp_up_proj/range":8.625,"train/tensor_grad_model_layers_47_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_93_mlp/mean":4,"train/tensor_act_model_layers_85_residual_pre_attn/min":338,"train/tensor_grad_model_layers_11_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_27_mlp/min":4,"train/tensor_act_model_layers_13_residual_post_mlp/mean":56,"train/tensor_param_model_layers_5_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_54_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_36_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_4_residual_pre_attn/range":1.25,"train/tensor_param_model_layers_71_mlp_gate_proj_weight/mean":0.0037384033203125,"train/tensor_act_model_layers_43_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_31_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_2_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_34_mlp_up_proj_weight/max_abs":0,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_62_mlp_gate_proj/std":1.6328125,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_86_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/min":-0.1630859375,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_3_mlp_gate_proj/max_abs":4.25,"train/tensor_act_model_layers_77_residual_post_mlp/min":310,"train/tensor_param_model_layers_54_mlp_down_proj_weight/max":0.125,"train/tensor_param_model_layers_11_mlp_down_proj_weight/range":0.2099609375,"train/tensor_act_model_layers_31_residual_post_mlp/std":0.34375,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_77_mlp_down_proj_weight/max":0.14453125,"train/tensor_act_model_layers_0_residual_pre_attn/std":0.0908203125,"train/tensor_act_model_layers_19_mlp_gate_proj/mean":0.435546875,"train/tensor_act_model_layers_17_residual_post_mlp/max_abs":73,"train/tensor_param_model_layers_10_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_5_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_70_residual_post_attn/mean":280,"train/tensor_act_model_layers_36_mlp/max":4,"train/tensor_grad_model_layers_30_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_12_mlp_up_proj_weight/max_abs":0.125,"train/tensor_act_model_layers_60_mlp_down_proj/mean":14.8125,"train/tensor_param_model_layers_79_mlp_up_proj_weight/min":-0.12890625,"train/tensor_act_model_layers_46_mlp_up_proj/mean":-0.007293701171875,"train/tensor_act_model_layers_89_mlp_up_proj/mean":0.0419921875,"train/tensor_act_model_layers_58_residual_post_mlp/max":237,"train/tensor_act_model_layers_90_mlp/mean":4,"train/tensor_act_model_layers_22_mlp/mean":4,"train/tensor_param_model_layers_44_mlp_down_proj_weight/min":-0.125,"train/tensor_act_model_layers_23_mlp_up_proj/norm":5539.23974609375,"train/tensor_param_model_layers_15_mlp_up_proj_weight/max":0.107421875,"train/tensor_grad_model_layers_3_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_67_mlp_gate_proj/max_abs":4.5625,"train/tensor_act_model_layers_46_residual_post_mlp/max_abs":189,"train/tensor_act_model_layers_51_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_34_mlp_up_proj/mean":0.040283203125,"train/tensor_act_model_layers_36_mlp/std":0,"train/tensor_param_model_layers_88_mlp_down_proj_weight/min":-0.1484375,"train/tensor_grad_model_layers_19_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_58_mlp/norm":8192,"train/tensor_param_model_layers_68_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_21_mlp_down_proj/max_abs":16.5,"train/tensor_act_model_layers_61_mlp_gate_proj/norm":4927.2265625,"train/tensor_act_model_layers_18_mlp/range":0,"train/tensor_grad_model_layers_11_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_78_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_29_residual_pre_attn/max_abs":117,"train/tensor_act_model_layers_47_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_1_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_72_residual_pre_attn/max":288,"train/tensor_grad_model_layers_21_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_95_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_44_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_49_mlp_gate_proj/max":4.4375,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_51_mlp/mean":4,"train/tensor_act_model_layers_57_mlp_down_proj/std":0.42578125,"train/tensor_act_model_layers_1_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_10_residual_pre_attn/max":40.75,"train/tensor_act_model_layers_89_mlp_gate_proj/max":4.53125,"train/tensor_param_model_layers_44_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_82_mlp_down_proj_weight/mean":-0.000823974609375,"train/tensor_act_model_layers_14_mlp_up_proj/range":8.78125,"train/tensor_act_model_layers_31_mlp_up_proj/min":-4.53125,"train/tensor_act_model_layers_14_residual_post_mlp/min":59,"train/tensor_grad_model_layers_48_mlp_up_proj_weight/max":0,"train/tensor_param_model_layers_6_mlp_up_proj_weight/max_abs":0.12158203125,"train/tensor_act_model_layers_9_residual_post_attn/norm":73657.8359375,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_10_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_86_mlp_up_proj/min":-4.96875,"train/tensor_param_model_layers_48_mlp_up_proj_weight/range":0.251953125,"train/tensor_act_model_layers_58_residual_post_mlp/norm":483100.78125,"train/tensor_param_model_layers_20_mlp_down_proj_weight/norm":5.625,"train/tensor_act_model_layers_9_residual_post_attn/range":1.75,"train/tensor_param_model_layers_41_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_94_residual_post_mlp/max_abs":380,"train/tensor_act_model_layers_23_mlp_gate_proj/norm":5191.8046875,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/mean":0.002532958984375,"train/tensor_param_model_layers_61_mlp_gate_proj_weight/max_abs":0.130859375,"train/tensor_act_model_layers_27_mlp_down_proj/max":15.875,"train/tensor_param_model_layers_21_mlp_down_proj_weight/max":0.115234375,"train/tensor_grad_model_layers_26_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_9_mlp_up_proj/min":-4.25,"train/tensor_act_model_layers_17_residual_post_attn/range":2,"train/tensor_act_model_layers_74_mlp/std":0,"train/tensor_act_model_layers_86_mlp/frac_near_dtype_limit":0,"train/tensor_param_model_layers_79_mlp_down_proj_weight/min":-0.134765625,"train/tensor_act_model_layers_40_residual_pre_attn/max":161,"train/tensor_grad_model_layers_35_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_73_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_94_mlp_gate_proj/min":-1.6953125,"train/tensor_act_model_layers_63_mlp/max":4,"train/tensor_act_model_layers_20_residual_pre_attn/min":78.5,"train/tensor_param_model_layers_24_mlp_up_proj_weight/max_abs":0.11181640625,"train/tensor_grad_model_layers_22_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_44_mlp_down_proj/max":16.125,"train/tensor_act_model_layers_77_mlp_gate_proj/min":-1.28125,"train/tensor_act_model_layers_1_mlp/norm":8192,"train/tensor_grad_model_layers_41_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_70_mlp_up_proj_weight/min":-0.1337890625,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/norm":8.5625,"train/tensor_grad_model_layers_25_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_27_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_0_mlp_gate_proj/norm":5388.9453125,"train/tensor_act_model_layers_5_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_56_residual_post_mlp/mean":228,"train/tensor_act_model_layers_49_residual_post_mlp/norm":409452.5625,"train/tensor_act_model_layers_42_mlp_gate_proj/std":1.671875,"train/tensor_param_model_layers_7_mlp_down_proj_weight/norm":5.3125,"train/tensor_act_model_layers_81_mlp/max":4,"train/tensor_act_model_layers_10_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_73_mlp_gate_proj_weight/mean":0.0027923583984375,"train/tensor_act_model_layers_29_mlp_down_proj/range":2.5,"train/tensor_param_model_layers_14_mlp_gate_proj_weight/mean":0.00360107421875,"train/tensor_param_model_layers_29_mlp_up_proj_weight/min":-0.10888671875,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_72_mlp_up_proj/min":-4.75,"train/tensor_act_model_layers_55_mlp_up_proj/range":9.4375,"train/tensor_act_model_layers_81_mlp_gate_proj/max":4.4375,"train/tensor_param_model_layers_16_mlp_up_proj_weight/min":-0.111328125,"train/tensor_act_model_layers_56_residual_post_attn/norm":458540.8125,"train/tensor_param_model_layers_83_mlp_gate_proj_weight/max_abs":0.15625,"train/tensor_act_model_layers_45_mlp_gate_proj/min":-1.2890625,"train/tensor_param_model_layers_27_mlp_gate_proj_weight/max_abs":0.123046875,"train/tensor_act_model_layers_49_residual_post_attn/min":194,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_76_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_21_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_90_mlp_gate_proj_weight/mean":0.00616455078125,"train/tensor_act_model_layers_70_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_78_residual_post_attn/norm":638941.25,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_19_mlp_down_proj/max":16.375,"train/tensor_grad_model_layers_24_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_23_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_4_residual_pre_attn/max":16.625,"train/tensor_act_model_layers_64_mlp_up_proj/min":-5.28125,"train/tensor_grad_model_layers_67_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_33_mlp/range":0,"train/tensor_act_model_layers_91_residual_post_mlp/max":368,"train/tensor_param_model_layers_5_mlp_up_proj_weight/std":0.02978515625,"train/tensor_act_model_layers_2_mlp_down_proj/max":17.375,"train/tensor_act_model_layers_72_residual_post_mlp/std":0.18359375,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_62_mlp_gate_proj/max_abs":4.6875,"train/tensor_act_model_layers_49_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_46_residual_post_mlp/range":3,"train/tensor_grad_model_layers_43_mlp_up_proj_weight/min":0,"train/tensor_act_model_layers_22_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/range":0,"train/tensor_param_model_layers_58_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_53_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_87_residual_pre_attn/range":2,"train/tensor_act_model_layers_78_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_63_mlp_down_proj/range":1.9375,"train/tensor_act_model_layers_78_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_92_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_24_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_31_residual_pre_attn/max":125,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_63_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_77_mlp/min":4,"train/tensor_act_model_layers_44_residual_post_attn/max_abs":177,"train/tensor_act_model_layers_37_residual_post_attn/std":0.31640625,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/min":-0.154296875,"train/tensor_act_model_layers_31_mlp/mean":4,"train/tensor_act_model_layers_10_residual_post_mlp/std":0.224609375,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_15_mlp/range":0,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_20_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/min":-0.2197265625,"train/tensor_act_model_layers_93_mlp_gate_proj/max":5.46875,"train/tensor_act_model_layers_30_mlp_up_proj/max":3.9375,"train/tensor_grad_model_layers_76_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_26_residual_post_mlp/min":106.5,"train/tensor_act_model_layers_25_residual_post_mlp/frac_near_user_limit":1,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/mean":0.003021240234375,"train/tensor_grad_model_layers_19_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_58_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_59_mlp/std":0,"train/tensor_act_model_layers_34_mlp/std":0,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_11_residual_post_attn/mean":44,"train/tensor_act_model_layers_25_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_74_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_48_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_41_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_3_mlp_down_proj/max":16.375,"train/tensor_param_model_layers_22_mlp_up_proj_weight/min":-0.10888671875,"train/tensor_act_model_layers_63_residual_post_mlp/max":256,"train/tensor_param_model_layers_78_mlp_up_proj_weight/std":0.04150390625,"train/tensor_act_model_layers_84_residual_post_mlp/mean":340,"train/tensor_param_model_layers_8_mlp_up_proj_weight/std":0.0296630859375,"train/tensor_act_model_layers_3_residual_post_mlp/max_abs":16.625,"train/tensor_grad_model_layers_53_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_62_mlp_gate_proj/min":-1.2578125,"train/tensor_act_model_layers_71_residual_post_mlp/range":2,"train/tensor_param_model_layers_1_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_65_mlp_up_proj/max_abs":4.9375,"train/tensor_act_model_layers_96_residual_post_attn/mean":384,"train/tensor_act_model_layers_0_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_mlp_gate_proj/norm":5027.48388671875,"train/tensor_grad_model_layers_74_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_72_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_27_mlp_up_proj/min":-4.15625,"train/tensor_param_model_layers_3_mlp_gate_proj_weight/max":0.125,"train/tensor_act_model_layers_3_residual_post_attn/range":1.25,"train/tensor_act_model_layers_20_residual_post_mlp/mean":84,"train/tensor_param_model_layers_52_mlp_down_proj_weight/norm":6.53125,"train/tensor_param_model_layers_14_mlp_down_proj_weight/max_abs":0.1142578125,"train/tensor_act_model_layers_54_mlp_gate_proj/max_abs":4.65625,"train/tensor_act_model_layers_41_mlp_down_proj/mean":14.75,"train/tensor_act_model_layers_73_residual_pre_attn/mean":292,"train/tensor_act_model_layers_45_residual_post_mlp/range":3,"train/tensor_act_model_layers_14_residual_post_attn/max_abs":57,"train/tensor_grad_model_layers_96_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_67_residual_pre_attn/max":268,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/max":0.12060546875,"train/tensor_param_model_layers_61_mlp_up_proj_weight/norm":6.75,"train/tensor_grad_model_layers_61_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_60_mlp_gate_proj_weight/max_abs":0.1328125,"train/tensor_act_model_layers_44_residual_post_attn/mean":176,"train/tensor_act_model_layers_72_residual_pre_attn/mean":288,"train/tensor_act_model_layers_24_residual_post_mlp/range":2.5,"train/tensor_grad_model_layers_86_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_68_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_83_residual_post_mlp/max":336,"train/tensor_param_model_layers_37_mlp_down_proj_weight/std":0.033203125,"train/tensor_act_model_layers_27_mlp_gate_proj/mean":0.609375,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/std":0,"train/tensor_param_model_layers_52_mlp_gate_proj_weight/norm":6.875,"train/tensor_act_model_layers_38_mlp_up_proj/max_abs":4.53125,"train/tensor_param_model_layers_60_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_33_residual_post_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_89_mlp_up_proj_weight/mean":0.00032806396484375,"train/tensor_act_model_layers_8_residual_pre_attn/std":0.1962890625,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_69_mlp_down_proj/max":17.125,"train/tensor_act_model_layers_84_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_25_mlp_gate_proj/norm":5203.63818359375,"train/tensor_act_model_layers_92_residual_post_mlp/std":0.18359375,"train/tensor_param_model_layers_62_mlp_gate_proj_weight/std":0.0390625,"train/tensor_act_model_layers_77_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_95_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_4_residual_post_mlp/range":1.375,"train/tensor_act_model_layers_69_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_mlp_down_proj/mean":15.1875,"train/tensor_act_model_layers_2_mlp_up_proj/min":-3.921875,"train/tensor_act_model_layers_56_mlp_gate_proj/max_abs":4.1875,"train/tensor_act_model_layers_85_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_73_residual_post_attn/std":0.18359375,"train/tensor_grad_model_layers_13_mlp_gate_proj_weight/std":0,"train/tensor_act_model_layers_54_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_7_mlp_up_proj/min":-3.9375,"train/tensor_act_model_layers_0_mlp_down_proj/std":14.6875,"train/tensor_act_model_layers_84_mlp_up_proj/max":4.4375,"train/tensor_param_model_layers_38_mlp_up_proj_weight/std":0.03369140625,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/min":0,"train/tensor_grad_model_layers_60_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_38_mlp/min":4,"train/tensor_act_model_layers_65_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_20_mlp_down_proj_weight/max_abs":0.1259765625,"train/tensor_act_model_layers_72_residual_post_attn/std":0.18359375,"train/tensor_param_model_layers_65_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_85_mlp_up_proj_weight/range":0.2939453125,"train/tensor_grad_model_layers_75_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_24_residual_post_attn/range":2.5,"train/tensor_grad_model_layers_42_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_93_mlp_up_proj/range":10.15625,"train/tensor_param_model_layers_99_mlp_down_proj_weight/max_abs":0.1796875,"train/tensor_param_model_layers_69_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_22_residual_pre_attn/min":86.5,"train/tensor_param_model_layers_39_mlp_gate_proj_weight/max_abs":0.126953125,"train/tensor_param_model_layers_81_mlp_gate_proj_weight/max_abs":0.146484375,"train/tensor_param_model_layers_95_mlp_gate_proj_weight/mean":0.00531005859375,"train/tensor_param_model_layers_25_mlp_up_proj_weight/norm":5.71875,"train/tensor_act_model_layers_9_residual_post_mlp/max_abs":40.75,"train/tensor_act_model_layers_33_residual_post_attn/max_abs":133,"train/tensor_act_model_layers_61_residual_post_mlp/max_abs":249,"train/tensor_act_model_layers_28_residual_pre_attn/max_abs":113,"train/tensor_act_model_layers_48_residual_post_mlp/min":194,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/min":0,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/norm":0,"train/tensor_act_model_layers_96_mlp/min":4,"train/tensor_param_model_layers_23_mlp_up_proj_weight/max_abs":0.12353515625,"train/tensor_param_model_layers_31_mlp_gate_proj_weight/mean":0.0034027099609375,"train/tensor_act_model_layers_28_mlp_up_proj/min":-4.5,"train/tensor_act_model_layers_10_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_33_mlp/min":4,"train/tensor_act_model_layers_28_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_64_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_mlp_gate_proj/max":4.46875,"train/tensor_act_model_layers_16_mlp_gate_proj/min":-1.265625,"train/tensor_param_model_layers_88_mlp_up_proj_weight/range":0.3046875,"train/tensor_grad_model_layers_80_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_25_mlp_gate_proj_weight/range":0.2431640625,"train/tensor_act_model_layers_40_mlp_up_proj/mean":0.06640625,"train/tensor_act_model_layers_6_mlp_down_proj/max":16.125,"train/tensor_act_model_layers_76_residual_post_attn/range":2,"train/tensor_param_model_layers_58_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_85_residual_pre_attn/norm":696285.25,"train/tensor_act_model_layers_20_residual_post_attn/mean":80,"train/tensor_act_model_layers_36_mlp_down_proj/std":0.345703125,"train/tensor_param_model_layers_50_mlp_up_proj_weight/std":0.03564453125,"train/tensor_param_model_layers_22_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_22_residual_pre_attn/std":0.359375,"train/tensor_act_model_layers_80_residual_post_mlp/std":0.18359375,"train/tensor_act_model_layers_90_mlp_up_proj/max":4.25,"train/tensor_act_model_layers_37_residual_pre_attn/mean":148,"train/tensor_grad_model_layers_72_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_35_mlp_down_proj_weight/max":0,"train/tensor_param_model_layers_39_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_67_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_58_mlp_down_proj/frac_near_user_limit":1,"train/tensor_param_model_layers_18_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_95_mlp/range":0,"train/tensor_act_model_layers_88_residual_post_mlp/max_abs":356,"train/tensor_act_model_layers_53_mlp_gate_proj/mean":0.515625,"train/tensor_act_model_layers_67_mlp/max_abs":4,"train/tensor_act_model_layers_73_residual_post_attn/max":292,"train/tensor_param_model_layers_86_mlp_up_proj_weight/max_abs":0.1435546875,"train/tensor_param_model_layers_84_mlp_gate_proj_weight/range":0.3525390625,"train/tensor_act_model_layers_72_mlp_down_proj/std":0.474609375,"train/tensor_act_model_layers_18_residual_post_attn/mean":72,"train/tensor_act_model_layers_72_residual_post_attn/max":288,"train/tensor_act_model_layers_23_residual_post_attn/range":2.5,"train/tensor_act_model_layers_22_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_98_residual_pre_attn/min":390,"train/tensor_param_model_layers_41_mlp_down_proj_weight/range":0.24560546875,"train/tensor_act_model_layers_92_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_81_mlp_up_proj/range":9.34375,"train/tensor_act_model_layers_55_mlp_gate_proj/max":4.6875,"train/tensor_act_model_layers_25_residual_post_mlp/max_abs":105,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_79_mlp/min":4,"train/tensor_grad_model_layers_59_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_34_mlp_down_proj_weight/norm":5.84375,"train/tensor_param_model_layers_5_mlp_gate_proj_weight/min":-0.12109375,"train/tensor_grad_model_layers_10_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_80_mlp_up_proj_weight/std":0.04248046875,"train/tensor_act_model_layers_95_mlp_down_proj/std":0.87890625,"train/tensor_act_model_layers_53_residual_post_mlp/min":214,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_62_mlp_up_proj_weight/min":-0.1357421875,"train/tensor_act_model_layers_42_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_98_residual_pre_attn/std":0.18359375,"train/tensor_param_model_layers_54_mlp_down_proj_weight/std":0.036376953125,"train/tensor_grad_model_layers_48_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_78_mlp_gate_proj_weight/max_abs":0.1396484375,"train/tensor_act_model_layers_73_mlp_down_proj/max":17.5,"train/tensor_act_model_layers_6_residual_post_attn/std":0.1953125,"train/tensor_act_model_layers_64_residual_post_mlp/mean":260,"train/tensor_act_model_layers_96_mlp_down_proj/std":1.140625,"train/tensor_grad_model_layers_67_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_8_mlp_up_proj/mean":-0.062255859375,"train/tensor_grad_model_layers_33_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_22_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_77_mlp_down_proj_weight/std":0.0400390625,"train/tensor_act_model_layers_83_mlp/range":0,"train/tensor_act_model_layers_10_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_66_residual_pre_attn/norm":540637.3125,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_67_mlp_gate_proj_weight/range":0.267578125,"train/tensor_act_model_layers_6_mlp_down_proj/norm":29973.16796875,"train/tensor_act_model_layers_23_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_2_mlp_gate_proj/max_abs":4.09375,"train/tensor_grad_model_layers_58_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_15_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_96_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_59_mlp_down_proj_weight/mean":-0.00156402587890625,"train/tensor_act_model_layers_46_residual_post_attn/max_abs":185,"train/tensor_act_model_layers_61_residual_post_attn/mean":244,"train/tensor_act_model_layers_9_residual_post_mlp/max":40.75,"train/tensor_grad_model_layers_0_mlp_down_proj_weight/frac_near_user_limit":1,"train/tensor_act_model_layers_73_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_36_mlp_up_proj_weight/norm":6.03125,"train/tensor_grad_model_layers_32_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_64_mlp/range":0,"train/tensor_param_model_layers_14_mlp_up_proj_weight/std":0.0302734375,"train/tensor_param_model_layers_58_mlp_up_proj_weight/min":-0.126953125,"train/tensor_param_model_layers_33_mlp_up_proj_weight/std":0.0322265625,"train/tensor_param_model_layers_95_mlp_down_proj_weight/mean":0.000530242919921875,"train/tensor_param_model_layers_11_mlp_up_proj_weight/min":-0.1220703125,"train/tensor_param_model_layers_18_mlp_up_proj_weight/std":0.031005859375,"train/tensor_param_model_layers_42_mlp_up_proj_weight/max":0.1259765625,"train/tensor_param_model_layers_2_mlp_up_proj_weight/range":0.23828125,"train/tensor_act_model_layers_83_residual_pre_attn/range":2,"train/tensor_act_model_layers_75_residual_post_attn/norm":614365.25,"train/tensor_param_model_layers_93_mlp_down_proj_weight/std":0.049560546875,"train/tensor_grad_model_layers_79_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_83_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_29_residual_pre_attn/range":2.5,"train/tensor_act_model_layers_13_mlp_gate_proj/min":-1.2890625,"train/tensor_act_model_layers_83_mlp_gate_proj/max_abs":4.4375,"train/tensor_param_model_layers_77_mlp_up_proj_weight/max_abs":0.15234375,"train/tensor_act_model_layers_65_mlp_up_proj/norm":5215.640625,"train/tensor_act_model_layers_81_mlp_gate_proj/range":5.78125,"train/tensor_act_model_layers_60_mlp/max":4,"train/tensor_param_model_layers_35_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_84_mlp_gate_proj/range":6.28125,"train/tensor_act_model_layers_25_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_40_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_85_residual_post_attn/range":2,"train/tensor_act_model_layers_20_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_37_mlp_down_proj/mean":14.6875,"train/tensor_act_model_layers_85_mlp_gate_proj/max":4.6875,"train/tensor_act_model_layers_46_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_mlp/mean":4,"train/tensor_act_model_layers_78_residual_pre_attn/min":310,"train/tensor_param_model_layers_1_mlp_up_proj_weight/norm":6.4375,"train/tensor_act_model_layers_12_residual_post_mlp/range":2,"train/tensor_act_model_layers_23_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_39_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_59_mlp_up_proj_weight/mean":-0.00110626220703125,"train/tensor_act_model_layers_90_mlp_up_proj/norm":5100.22265625,"train/tensor_act_model_layers_41_mlp_up_proj/norm":5450.01416015625,"train/tensor_param_model_layers_98_mlp_gate_proj_weight/max":0.19140625,"train/tensor_act_model_layers_50_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_75_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_91_mlp_gate_proj_weight/max":0.1513671875,"train/tensor_grad_model_layers_21_mlp_down_proj_weight/std":0,"train/tensor_act_model_layers_44_mlp/min":4,"train/tensor_act_model_layers_93_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_12_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_95_mlp_gate_proj_weight/range":0,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/mean":0,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_11_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_91_mlp_gate_proj/mean":0.71484375,"train/tensor_grad_model_layers_99_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_14_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_12_mlp_gate_proj/range":6.1015625,"train/tensor_act_model_layers_67_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_6_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_2_mlp_up_proj/max":4.25,"train/tensor_act_model_layers_59_mlp_down_proj/max":17.25,"train/tensor_grad_model_layers_11_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_21_mlp_up_proj_weight/min":-0.12353515625,"train/tensor_act_model_layers_2_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_78_residual_pre_attn/std":0.18359375,"train/tensor_act_model_layers_86_mlp_gate_proj/range":5.9609375,"train/tensor_act_model_layers_7_mlp/std":0,"train/tensor_grad_model_layers_25_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_74_mlp/max_abs":4,"train/tensor_act_model_layers_98_residual_pre_attn/frac_near_user_limit":1,"train/tensor_param_model_layers_94_mlp_up_proj_weight/range":0.31640625,"train/tensor_param_model_layers_79_mlp_up_proj_weight/mean":-0.000640869140625,"train/tensor_act_model_layers_7_mlp_up_proj/max":4.34375,"train/tensor_grad_model_layers_54_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_14_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_29_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_54_residual_post_mlp/min":218,"train/tensor_grad_model_layers_95_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_85_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_74_residual_post_mlp/max_abs":300,"train/tensor_grad_model_layers_16_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_20_mlp_down_proj/min":14,"train/tensor_param_model_layers_36_mlp_up_proj_weight/range":0.251953125,"train/tensor_act_model_layers_2_mlp_down_proj/norm":30683.291015625,"train/tensor_param_model_layers_24_mlp_up_proj_weight/max":0.1103515625,"train/tensor_act_model_layers_23_residual_post_attn/mean":92,"train/tensor_param_model_layers_54_mlp_gate_proj_weight/max_abs":0.12890625,"train/tensor_grad_model_layers_38_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_3_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_34_mlp_up_proj/range":8.546875,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/std":0.0576171875,"train/tensor_act_model_layers_71_residual_post_mlp/max":288,"train/tensor_grad_model_layers_51_mlp_gate_proj_weight/max":0,"train/tensor_param_model_layers_82_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_55_residual_pre_attn/max":221,"train/tensor_act_model_layers_3_residual_post_mlp/min":15.375,"train/tensor_act_model_layers_41_mlp_down_proj/norm":30235.431640625,"train/tensor_act_model_layers_54_residual_pre_attn/max_abs":217,"train/tensor_param_model_layers_16_mlp_down_proj_weight/mean":-0.0011138916015625,"train/tensor_param_model_layers_23_mlp_gate_proj_weight/min":-0.1240234375,"train/tensor_act_model_layers_37_mlp_down_proj/max":16.125,"train/tensor_param_model_layers_75_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_93_mlp_up_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_12_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_75_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_96_mlp_gate_proj_weight/norm":0,"train/tensor_grad_model_layers_65_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_11_residual_post_attn/std":0.26171875,"train/tensor_param_model_layers_75_mlp_up_proj_weight/std":0.0400390625,"train/tensor_act_model_layers_40_mlp_down_proj/max_abs":16.375,"train/tensor_act_model_layers_29_mlp_up_proj/max":5.0625,"train/tensor_act_model_layers_40_residual_post_attn/max":161,"train/tensor_param_model_layers_32_mlp_up_proj_weight/std":0.032470703125,"train/tensor_act_model_layers_79_residual_post_mlp/range":2,"train/tensor_act_model_layers_43_mlp/min":4,"train/tensor_act_model_layers_93_mlp_up_proj/max_abs":5.40625,"train/tensor_grad_model_layers_44_mlp_gate_proj_weight/max_abs":0,"train/tensor_param_model_layers_98_mlp_down_proj_weight/norm":9.5,"train/tensor_param_model_layers_21_mlp_down_proj_weight/std":0.031005859375,"train/tensor_act_model_layers_69_mlp_up_proj/max_abs":4.1875,"train/tensor_act_model_layers_54_mlp_up_proj/std":1.8671875,"train/tensor_act_model_layers_4_mlp_up_proj/max_abs":3.9375,"train/tensor_param_model_layers_0_mlp_up_proj_weight/max_abs":0.125,"train/tensor_param_model_layers_79_mlp_down_proj_weight/max":0.1455078125,"train/tensor_act_model_layers_90_residual_pre_attn/mean":360,"train/tensor_grad_model_layers_0_mlp_up_proj_weight/norm":57088,"train/tensor_act_model_layers_16_mlp_down_proj/max_abs":16.5,"train/tensor_param_model_layers_0_mlp_up_proj_weight/std":0.0634765625,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/max_abs":9.5625,"train/tensor_act_model_layers_36_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_67_residual_post_attn/range":2,"train/tensor_act_model_layers_13_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_29_residual_pre_attn/max":117,"train/tensor_act_model_layers_2_residual_post_mlp/norm":24558.56640625,"train/tensor_act_model_layers_28_mlp_up_proj/range":9.3125,"train/tensor_act_model_layers_73_mlp_gate_proj/mean":0.357421875,"train/tensor_param_model_layers_39_mlp_up_proj_weight/std":0.03369140625,"train/tensor_grad_model_layers_43_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_13_mlp_gate_proj/mean":0.306640625,"train/tensor_param_model_layers_24_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_41_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_19_mlp_down_proj_weight/norm":5.53125,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_65_residual_post_mlp/max":264,"train/tensor_param_model_layers_39_mlp_up_proj_weight/max_abs":0.1259765625,"train/tensor_act_model_layers_26_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_43_mlp_up_proj/max_abs":4.90625,"train/tensor_act_model_layers_25_mlp/max":4,"train/tensor_grad_model_layers_97_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_12_mlp_down_proj/max":16,"train/tensor_act_model_layers_50_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_89_mlp_down_proj_weight/mean":0.00115203857421875,"train/tensor_param_model_layers_63_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_51_residual_post_mlp/max":209,"train/tensor_grad_model_layers_27_mlp_down_proj_weight/range":0,"train/tensor_act_model_layers_90_residual_post_attn/range":2,"train/tensor_act_model_layers_39_residual_pre_attn/mean":156,"train/tensor_act_model_layers_49_mlp_gate_proj/mean":0.51953125,"train/tensor_act_model_layers_53_mlp_up_proj/max":4.9375,"train/tensor_act_model_layers_63_mlp_up_proj/max":4.9375,"train/tensor_grad_model_layers_25_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_10_mlp_down_proj/min":14.125,"train/tensor_param_model_layers_64_mlp_gate_proj_weight/max_abs":0.1328125,"train/tensor_act_model_layers_59_residual_pre_attn/max":237,"train/tensor_act_model_layers_38_mlp_up_proj/min":-4.53125,"train/tensor_grad_model_layers_1_mlp_down_proj_weight/range":17.875,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_82_mlp_down_proj_weight/std":0,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/range":0,"train/tensor_act_model_layers_25_residual_pre_attn/norm":204561.203125,"train/tensor_act_model_layers_23_mlp_up_proj/frac_near_user_limit":0,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/min":-0.12109375,"train/tensor_param_model_layers_2_mlp_down_proj_weight/max_abs":0.11083984375,"train/tensor_act_model_layers_98_residual_post_mlp/mean":396,"train/tensor_act_model_layers_69_residual_post_mlp/range":2,"train/tensor_grad_model_layers_92_mlp_gate_proj_weight/mean":0,"train/tensor_act_model_layers_87_residual_post_mlp/max":352,"train/tensor_act_model_layers_66_residual_pre_attn/range":2,"train/tensor_param_model_layers_9_mlp_up_proj_weight/range":0.22509765625,"train/tensor_act_model_layers_55_mlp_up_proj/max_abs":4.9375,"train/tensor_act_model_layers_40_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_param_model_layers_41_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_40_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_15_mlp_down_proj/std":0.373046875,"train/tensor_act_model_layers_98_mlp_gate_proj/max":5.75,"train/tensor_grad_model_layers_32_mlp_up_proj_weight/range":0,"train/tensor_grad_model_layers_18_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_14_mlp_gate_proj_weight/min":0,"train/tensor_act_model_layers_6_mlp_up_proj/range":7.953125,"train/tensor_act_model_layers_62_residual_post_mlp/mean":252,"train/tensor_grad_model_layers_27_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_13_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_22_mlp_gate_proj/min":-1.3515625,"train/tensor_act_model_layers_87_mlp_gate_proj/std":1.6171875,"train/tensor_act_model_layers_41_residual_pre_attn/range":3,"train/tensor_grad_model_layers_31_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_55_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_11_residual_post_attn/max":45,"train/tensor_act_model_layers_40_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_27_mlp_up_proj/mean":0.060791015625,"train/tensor_act_model_layers_38_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_76_mlp_down_proj_weight/max":0,"train/tensor_grad_model_layers_50_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_25_mlp_up_proj_weight/range":0.2421875,"train/tensor_param_model_layers_80_mlp_gate_proj_weight/max_abs":0.1416015625,"train/tensor_act_model_layers_95_residual_post_attn/mean":380,"train/tensor_grad_model_layers_74_mlp_gate_proj_weight/range":0,"train/tensor_param_model_layers_6_mlp_gate_proj_weight/max":0.10546875,"train/tensor_param_model_layers_57_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_25_mlp_gate_proj/range":5.8984375,"train/tensor_param_model_layers_67_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_4_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_1_mlp_gate_proj/mean":0.52734375,"train/tensor_param_model_layers_29_mlp_down_proj_weight/mean":-0.0007171630859375,"train/tensor_grad_model_layers_91_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_49_mlp/min":4,"train/tensor_act_model_layers_8_mlp_down_proj/max_abs":15.875,"train/tensor_act_model_layers_32_residual_post_attn/norm":261778.859375,"train/tensor_param_model_layers_77_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_49_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_47_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_64_residual_pre_attn/max_abs":256,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/std":0.049072265625,"train/tensor_act_model_layers_62_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp/min":4,"train/tensor_act_model_layers_37_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_75_mlp_gate_proj/min":-1.3515625,"train/tensor_param_model_layers_77_mlp_down_proj_weight/max_abs":0.14453125,"train/tensor_grad_model_layers_90_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_98_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_15_residual_post_mlp/norm":130804.1796875,"train/tensor_act_model_layers_29_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_61_mlp/min":4,"train/tensor_act_model_layers_65_residual_pre_attn/range":2,"train/tensor_act_model_layers_31_mlp_gate_proj/min":-1.296875,"train/tensor_grad_model_layers_7_mlp_gate_proj_weight/max":0,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/max_abs":0,"train/tensor_act_model_layers_2_residual_pre_attn/std":0.11376953125,"train/tensor_grad_model_layers_28_mlp_down_proj_weight/mean":0,"train/tensor_act_model_layers_63_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_7_residual_post_mlp/max_abs":32.75,"train/tensor_grad_model_layers_4_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_40_residual_pre_attn/mean":160,"train/tensor_act_model_layers_27_residual_pre_attn/min":106.5,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_8_mlp/max_abs":4,"train/tensor_param_model_layers_88_mlp_gate_proj_weight/max_abs":0.166015625,"train/tensor_act_model_layers_11_mlp/max_abs":4,"train/tensor_act_model_layers_2_mlp_up_proj/range":8.171875,"train/tensor_act_model_layers_75_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_19_mlp_up_proj_weight/mean":-0.0003662109375,"train/tensor_param_model_layers_44_mlp_down_proj_weight/std":0.033935546875,"train/tensor_act_model_layers_74_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_5_residual_post_attn/max_abs":20.625,"train/tensor_grad_model_layers_12_mlp_gate_proj_weight/max_abs":0,"train/tensor_grad_model_layers_80_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_27_residual_pre_attn/std":0.359375,"train/tensor_act_model_layers_91_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_86_residual_post_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_9_mlp_up_proj/max":4.5625,"train/tensor_param_model_layers_88_mlp_down_proj_weight/std":0.0458984375,"train/tensor_act_model_layers_37_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_2_mlp_up_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_78_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_59_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_40_mlp_gate_proj_weight/max":0,"train/tensor_act_model_layers_15_residual_post_attn/min":59,"train/tensor_act_model_layers_70_mlp_down_proj/max":17,"train/tensor_act_model_layers_68_residual_pre_attn/max":272,"train/tensor_act_model_layers_11_mlp_down_proj/min":14.1875,"train/tensor_act_model_layers_79_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_98_mlp_up_proj/range":10.375,"train/tensor_grad_model_layers_88_mlp_gate_proj_weight/norm":0,"train/tensor_act_model_layers_8_mlp_down_proj/std":0.38671875,"train/tensor_act_model_layers_85_mlp/max_abs":4,"train/tensor_act_model_layers_82_mlp_down_proj/min":14.125,"train/tensor_act_model_layers_95_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_act_model_layers_78_mlp/mean":4,"train/tensor_act_model_layers_43_residual_post_mlp/frac_near_dtype_limit":0,"train/tensor_act_model_layers_52_residual_pre_attn/min":206,"train/tensor_act_model_layers_13_mlp/max":4,"train/tensor_grad_model_layers_85_mlp_down_proj_weight/max_abs":0,"train/tensor_grad_model_layers_87_mlp_up_proj_weight/mean":0,"train/tensor_act_model_layers_0_mlp/max_abs":4,"train/tensor_act_model_layers_55_residual_post_attn/norm":450348.8125,"train/tensor_grad_model_layers_64_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_94_mlp/std":0,"train/tensor_act_model_layers_90_residual_post_attn/std":0.18359375,"train/tensor_act_model_layers_67_mlp_up_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_58_mlp_gate_proj/std":1.6484375,"train/tensor_param_model_layers_12_mlp_gate_proj_weight/std":0.0306396484375,"train/tensor_act_model_layers_31_residual_post_attn/max_abs":125,"train/tensor_act_model_layers_43_residual_pre_attn/max":173,"train/tensor_grad_model_layers_66_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_35_mlp_up_proj_weight/norm":5.9375,"train/tensor_act_model_layers_81_mlp/norm":8192,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_59_mlp/max":4,"train/tensor_act_model_layers_12_residual_post_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_55_mlp_down_proj_weight/min":0,"train/tensor_act_model_layers_12_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_26_mlp_up_proj_weight/std":0.03173828125,"train/tensor_act_model_layers_54_mlp/mean":4,"train/tensor_param_model_layers_41_mlp_gate_proj_weight/range":0.24951171875,"train/tensor_act_model_layers_0_mlp_gate_proj/max":7.46875,"train/tensor_param_model_layers_35_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_83_mlp_up_proj_weight/mean":-0.000316619873046875,"train/tensor_act_model_layers_38_mlp_up_proj/norm":5422.810546875,"train/tensor_act_model_layers_3_mlp_up_proj/range":8.34375,"train/tensor_act_model_layers_44_mlp/frac_near_user_limit":0,"train/tensor_act_model_layers_39_residual_post_attn/std":0.31640625,"train/tensor_act_model_layers_8_mlp_down_proj/range":1.8125,"train/tensor_act_model_layers_71_mlp/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_20_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_52_residual_pre_attn/std":0.31640625,"train/tensor_act_model_layers_28_mlp_gate_proj/max":4.46875,"train/tensor_act_model_layers_37_mlp_down_proj/norm":30096.869140625,"train/tensor_param_model_layers_37_mlp_down_proj_weight/range":0.24951171875,"train/tensor_act_model_layers_61_residual_post_attn/frac_near_dtype_limit":0,"train/tensor_grad_model_layers_23_mlp_gate_proj_weight/min":0,"train/tensor_param_model_layers_35_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_grad_model_layers_9_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_param_model_layers_30_mlp_gate_proj_weight/mean":0.003936767578125,"train/tensor_act_model_layers_0_mlp_up_proj/norm":5488.49755859375,"train/tensor_act_model_layers_9_residual_pre_attn/frac_near_dtype_limit":0,"train/tensor_param_model_layers_81_mlp_down_proj_weight/max_abs":0.1474609375,"train/tensor_act_model_layers_13_mlp/max_abs":4,"train/tensor_param_model_layers_75_mlp_down_proj_weight/range":0.2734375,"train/tensor_grad_model_layers_15_mlp_down_proj_weight/max_abs":0,"train/tensor_param_model_layers_10_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_99_mlp_up_proj/max":5.71875,"train/tensor_param_model_layers_33_mlp_up_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_47_residual_post_mlp/range":3,"train/tensor_param_model_layers_89_mlp_down_proj_weight/range":0.2841796875,"train/tensor_act_model_layers_86_residual_post_attn/range":2,"train/tensor_act_model_layers_76_mlp/max_abs":4,"train/tensor_act_model_layers_3_mlp_gate_proj/norm":5030.93505859375,"train/tensor_act_model_layers_34_mlp_gate_proj/max":4.71875,"train/tensor_grad_model_layers_72_mlp_down_proj_weight/max":0,"train/tensor_act_model_layers_41_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_29_mlp_gate_proj/std":1.703125,"train/tensor_act_model_layers_84_mlp/range":0,"train/tensor_grad_model_layers_9_mlp_down_proj_weight/min":0,"train/tensor_grad_model_layers_42_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_35_residual_pre_attn/norm":286572.78125,"train/tensor_act_model_layers_50_residual_pre_attn/frac_near_user_limit":1,"train/tensor_act_model_layers_84_mlp/mean":4,"train/tensor_act_model_layers_26_residual_pre_attn/frac_near_user_limit":1,"train/tensor_grad_model_layers_90_mlp_down_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_74_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_52_residual_post_mlp/range":3,"train/tensor_param_model_layers_43_mlp_gate_proj_weight/frac_near_user_limit":0,"train/tensor_act_model_layers_72_residual_post_attn/range":2,"train/tensor_grad_model_layers_16_mlp_up_proj_weight/max":0,"train/tensor_act_model_layers_90_mlp_gate_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_95_mlp_down_proj/norm":31119.416015625,"train/tensor_act_model_layers_11_residual_pre_attn/mean":44,"train/tensor_act_model_layers_93_mlp/max":4,"train/tensor_act_model_layers_13_residual_post_mlp/frac_near_user_limit":1,"train/tensor_act_model_layers_18_residual_post_attn/range":2.5,"train/tensor_param_model_layers_37_mlp_up_proj_weight/range":0.255859375,"train/tensor_grad_model_layers_23_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_66_mlp_gate_proj/std":1.640625,"train/tensor_param_model_layers_49_mlp_gate_proj_weight/max_abs":0.12890625,"train/tensor_act_model_layers_64_residual_post_mlp/frac_near_dtype_limit":0,"_timestamp":1.7867433049037457e+09,"train/tensor_grad_model_layers_44_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_8_mlp_down_proj_weight/frac_near_dtype_limit":0,"train/tensor_param_model_layers_63_mlp_up_proj_weight/max":0.12890625,"train/tensor_act_model_layers_85_residual_pre_attn/max_abs":340,"train/tensor_act_model_layers_80_residual_post_mlp/mean":324,"train/tensor_act_model_layers_47_residual_post_mlp/frac_near_user_limit":1,"train/tensor_param_model_layers_12_mlp_up_proj_weight/max":0.1220703125,"train/tensor_param_model_layers_35_mlp_up_proj_weight/range":0.2431640625,"train/tensor_act_model_layers_39_residual_post_attn/max":157,"train/tensor_act_model_layers_27_mlp/max_abs":4,"train/tensor_grad_model_layers_61_mlp_down_proj_weight/norm":0,"train/tensor_act_model_layers_10_residual_pre_attn/max_abs":40.75,"train/tensor_grad_model_layers_88_mlp_up_proj_weight/std":0,"train/tensor_act_model_layers_77_mlp_up_proj/min":-4.4375,"train/tensor_param_model_layers_44_mlp_down_proj_weight/range":0.2529296875,"train/tensor_param_model_layers_99_mlp_gate_proj_weight/max_abs":0.220703125,"train/tensor_grad_model_layers_69_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_7_mlp/frac_near_user_limit":0,"train/tensor_grad_model_layers_81_mlp_gate_proj_weight/norm":0,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/range":0.2373046875,"train/tensor_act_model_layers_25_mlp_gate_proj/max_abs":4.59375,"train/tensor_act_model_layers_88_mlp_gate_proj/frac_near_user_limit":0,"train/tensor_act_model_layers_90_residual_post_attn/max":360,"train/tensor_act_model_layers_0_residual_post_mlp/frac_near_user_limit":0,"train/tensor_param_model_layers_15_mlp_gate_proj_weight/frac_near_dtype_limit":0,"train/tensor_act_model_layers_24_mlp_gate_proj/mean":0.484375,"train/tensor_act_model_layers_18_mlp_gate_proj/min":-1.2890625,"train/tensor_act_model_layers_18_mlp_up_proj/norm":5512.4951171875,"train/tensor_param_model_layers_21_mlp_gate_proj_weight/max_abs":0.12109375,"train/tensor_act_model_layers_85_mlp_up_proj/range":9.15625,"train/tensor_act_model_layers_50_mlp_up_proj/mean":0.0177001953125,"train/tensor_param_model_layers_16_mlp_gate_proj_weight/std":0.0311279296875,"train/tensor_act_model_layers_17_mlp_gate_proj/mean":0.65234375,"train/tensor_act_model_layers_20_residual_pre_attn/std":0.357421875,"train/tensor_grad_model_layers_3_mlp_gate_proj_weight/std":0,"train/tensor_grad_model_layers_68_mlp_down_proj_weight/mean":0,"train/tensor_grad_model_layers_38_mlp_gate_proj_weight/mean":0,"train/tensor_param_model_layers_46_mlp_gate_proj_weight/max_abs":0.134765625,"train/tensor_grad_model_layers_34_mlp_down_proj_weight/std":0,"train/tensor_param_model_layers_56_mlp_down_proj_weight/range":0.265625,"train/tensor_grad_model_layers_36_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_22_mlp_gate_proj/max_abs":4.46875,"train/tensor_grad_model_layers_33_mlp_up_proj_weight/norm":0,"train/tensor_grad_model_layers_19_mlp_up_proj_weight/std":0,"train/tensor_grad_model_layers_18_mlp_down_proj_weight/min":0,"train/tensor_param_model_layers_20_mlp_gate_proj_weight/max":0.123046875,"train/tensor_act_model_layers_8_mlp/min":4,"train/tensor_act_model_layers_53_residual_post_attn/norm":433964.84375,"train/tensor_grad_model_layers_5_mlp_up_proj_weight/max_abs":0,"train/tensor_act_model_layers_83_mlp_gate_proj/range":5.75,"train/tensor_grad_model_layers_8_mlp_up_proj_weight/min":0,"train/tensor_param_model_layers_1_mlp_up_proj_weight/mean":-0.00164031982421875,"train/tensor_act_model_layers_37_residual_post_attn/max":149,"train/tensor_grad_model_layers_26_mlp_up_proj_weight/range":0,"train/tensor_act_model_layers_31_residual_post_mlp/mean":128,"train/tensor_act_model_layers_11_mlp_down_proj/frac_near_dtype_limit":0,"train/tensor_act_model_layers_38_residual_post_attn/norm":311148.71875,"train/tensor_param_model_layers_17_mlp_gate_proj_weight/std":0.031005859375} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-core.log b/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-core.log index 59acf96f4c751e74e9217cb996ef31c503e3590c..814ed69533d389f53d1c3ade0f0bdffbe6f84b3c 100644 --- a/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-core.log +++ b/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-core.log @@ -75,3 +75,11 @@ {"time":"2026-08-14T21:28:35.344777056Z","level":"INFO","msg":"handleInformInit: received","streamId":"gqskne0x","id":"3(@)"} {"time":"2026-08-14T21:28:35.603024157Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"gqskne0x","id":"3(@)"} {"time":"2026-08-14T21:28:41.078893683Z","level":"INFO","msg":"connection: cancelling request","id":"3(@)","requestId":"yxbo55z35lse"} +{"time":"2026-08-14T21:35:05.007052341Z","level":"INFO","msg":"connection: cancelling request","id":"3(@)","requestId":"yxbo55z35lse"} +{"time":"2026-08-14T21:35:06.745707472Z","level":"INFO","msg":"connection: cancelling request","id":"3(@)","requestId":"yxbo55z35lse"} +{"time":"2026-08-14T21:35:06.767299957Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"gqskne0x","id":"3(@)"} +{"time":"2026-08-14T21:35:06.773662133Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"gqskne0x","id":"3(@)"} +{"time":"2026-08-14T21:35:08.806158062Z","level":"INFO","msg":"connection: closing","id":"3(@)"} +{"time":"2026-08-14T21:35:08.806241813Z","level":"INFO","msg":"connection: closed successfully","id":"3(@)"} +{"time":"2026-08-14T21:35:08.806161779Z","level":"INFO","msg":"processOutgoingData: finished","id":"3(@)"} +{"time":"2026-08-14T21:35:08.806251477Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"3(@)"} diff --git a/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-internal.log index 70b55dc30b3bb82a035e6dee6e50a99f3b0291fd..e066fb401a45adc6022b441e0efee8135c5319df 100644 --- a/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-internal.log +++ b/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug-internal.log @@ -45,3 +45,49 @@ {"time":"2026-08-14T21:29:52.226871571Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} {"time":"2026-08-14T21:30:06.110393089Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":2,"events_offset":9,"events_lines":2,"console_offset":15,"console_lines":10} {"time":"2026-08-14T21:30:07.570174219Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:30:21.109617614Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":3,"events_offset":11,"events_lines":2,"console_offset":21,"console_lines":1} +{"time":"2026-08-14T21:30:22.080315159Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:30:36.115961679Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":3,"events_offset":13,"events_lines":2,"console_offset":25,"console_lines":11} +{"time":"2026-08-14T21:30:37.217389873Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:30:51.100718712Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":2,"events_offset":15,"events_lines":2,"console_offset":31,"console_lines":1} +{"time":"2026-08-14T21:30:51.948020311Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:31:06.117731609Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":3,"events_offset":17,"events_lines":2,"console_offset":36,"console_lines":10} +{"time":"2026-08-14T21:31:07.223434557Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:31:21.099230907Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":2,"events_offset":19,"events_lines":2,"console_offset":41,"console_lines":1} +{"time":"2026-08-14T21:31:21.759032284Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:31:36.114372704Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":3,"events_offset":21,"events_lines":2,"console_offset":46,"console_lines":10} +{"time":"2026-08-14T21:31:37.153517827Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:31:51.098719406Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":2,"events_offset":23,"events_lines":2,"console_offset":51,"console_lines":1} +{"time":"2026-08-14T21:31:52.120465709Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:06.108268856Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":2,"events_offset":25,"events_lines":2,"console_offset":56,"console_lines":9} +{"time":"2026-08-14T21:32:07.025091977Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:21.114007643Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":2,"events_offset":27,"events_lines":2,"console_offset":61,"console_lines":1} +{"time":"2026-08-14T21:32:22.043468825Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:36.110666778Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":2,"events_offset":29,"events_lines":2,"console_offset":65,"console_lines":9} +{"time":"2026-08-14T21:32:36.96385382Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:51.106950797Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":2,"events_offset":31,"events_lines":2,"console_offset":71,"console_lines":1} +{"time":"2026-08-14T21:32:52.043139409Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:06.101277485Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":71,"console_lines":1} +{"time":"2026-08-14T21:33:07.00621301Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:21.115155817Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":2,"events_offset":35,"events_lines":2,"console_offset":74,"console_lines":10} +{"time":"2026-08-14T21:33:22.004559582Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:36.113278989Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":2,"events_offset":37,"events_lines":2,"console_offset":81,"console_lines":1} +{"time":"2026-08-14T21:33:37.036173708Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:51.103713137Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":2,"events_offset":39,"events_lines":2,"console_offset":81,"console_lines":1} +{"time":"2026-08-14T21:33:51.923559316Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:06.111807693Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":84,"console_lines":10} +{"time":"2026-08-14T21:34:07.118394557Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:21.110439909Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":2,"events_offset":43,"events_lines":2,"console_offset":91,"console_lines":1} +{"time":"2026-08-14T21:34:22.046475792Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:36.099853771Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":91,"console_lines":1} +{"time":"2026-08-14T21:34:36.908754879Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:51.101796618Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":50,"history_lines":2,"events_offset":47,"events_lines":2,"console_offset":91,"console_lines":1} +{"time":"2026-08-14T21:34:51.852581483Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:35:05.824230118Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T21:35:05.833974131Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":52,"history_lines":1,"events_offset":49,"events_lines":1,"console_offset":94,"console_lines":17,"uploaded_len":3,"complete":true,"exit_code":0} +{"time":"2026-08-14T21:35:06.721587751Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:35:06.722695731Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T21:35:06.767338139Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T21:35:06.767370539Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T21:35:06.767476774Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T21:35:06.767482411Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug.log b/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug.log index b43ac43447df0773671fa75307878dd4d6210cf9..a09b4a2539fe21828dafc2d59ed17b57c0b9f524 100644 --- a/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug.log +++ b/zain/Activation/wandb/run-20260814_212835-gqskne0x/logs/debug.log @@ -18,3 +18,8 @@ config: {'_wandb': {}} 2026-08-14 21:28:36,078 INFO MainThread:936813 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 100, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'glu', 'activation': 'silu-waleed10', 'waleed_beta': 4, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/glu-silu-waleed10-100L_trash_run', 'per_device_train_batch_size': 64, 'num_train_epochs': 1, 'max_steps': 3500, 'learning_rate': 9e-05, 'lr_scheduler_type': 'constant_with_warmup', 'lr_scheduler_kwargs': None, 'warmup_steps': 50, 'optim': 'adamw_torch_fused', 'optim_args': None, 'weight_decay': 0.0, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 1, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-glu-silu-waleed10-100L-16.9M-20260814-212834', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 2498, 'eval_delay': 0, 'per_device_eval_batch_size': 1024, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 50, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/6L-glu-silu-waleed10-100L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} 2026-08-14 21:28:36,083 INFO MainThread:936813 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 16934016 - > 2026-08-14 21:28:36,083 INFO MainThread:936813 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 16934016 None +2026-08-14 21:35:05,006 INFO MainThread:936813 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research7/gqskne0x +2026-08-14 21:35:05,006 INFO MainThread:936813 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 21:35:05,006 INFO MainThread:936813 [wandb_run.py:_restore():2570] restore +2026-08-14 21:35:05,006 INFO MainThread:936813 [wandb_run.py:_restore():2576] restore done +2026-08-14 21:35:06,766 INFO MainThread:936813 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_212835-gqskne0x/run-gqskne0x.wandb b/zain/Activation/wandb/run-20260814_212835-gqskne0x/run-gqskne0x.wandb index 0383343d7a849462ff40ea01d7bcea2173ff1a55..c5657adba424feddb4e01bbcf7b394cd21059206 100644 --- a/zain/Activation/wandb/run-20260814_212835-gqskne0x/run-gqskne0x.wandb +++ b/zain/Activation/wandb/run-20260814_212835-gqskne0x/run-gqskne0x.wandb @@ -1,3 +1,3 @@ version https://git-lfs.github.com/spec/v1 -oid sha256:642685d115792bd6ab89f27f6b3b3941e2759dd1c0b094ae917478acfdb1e3ac -size 14843904 +oid sha256:59d9ff1dae5584c3328fb08df040f9affb6b10bd792c5ca40be4072c95d32f90 +size 51423131 diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/config.yaml b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..3f41366a7db4579f3f74604e5ce40193cf03a790 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/config.yaml @@ -0,0 +1,439 @@ +_name_or_path: + value: "" +_wandb: + value: + cli_version: 0.28.1 + e: + avd7wjdsk6ji6qbb7ojy8wwedk5izmcm: + args: + - --config + - /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml + - --variants + - glu-waleed10 + - glu-silu-waleed10 + - mlp-waleed10 + - mlp-silu-waleed10 + - glu-waleed + - glu-situglu_low + - glu-waleedglu_low + codePath: sweep.py + codePathLocal: sweep.py + cpu_count: 112 + cpu_count_logical: 224 + cudaVersion: "12.4" + disk: + /: + total: "1560765693952" + used: "709060804608" + email: deepnevro@gmail.com + executable: /mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python + git: + commit: 3fb47c6943d6927eafc39bd399a3d3a520bae74a + remote: https://github.com/w-ahmad1a10/Activation.git + gpu: NVIDIA H100 80GB HBM3 + gpu_count: 8 + gpu_nvidia: + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-39c684a5-fde6-83d7-1663-0859795881ae + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-132944c4-b689-2b5f-89a4-d730401677ab + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bc6c3e3c-9b90-09ca-c034-774961847c54 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea + host: deeplens-k3s-node1 + memory: + total: "2164089937920" + os: Linux-5.15.0-126-generic-x86_64-with-glibc2.35 + program: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py + python: CPython 3.11.15 + root: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation + startedAt: "2026-08-14T21:31:53.799629Z" + writerId: avd7wjdsk6ji6qbb7ojy8wwedk5izmcm + m: + - "1": train/global_step + "6": + - 3 + "7": [] + - "2": '*' + "5": 1 + "6": + - 1 + "7": [] + python_version: 3.11.15 + t: + "1": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "2": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "3": + - 2 + - 7 + - 13 + - 19 + - 62 + - 66 + "4": 3.11.15 + "5": 0.28.1 + "6": 5.16.0.dev0 + "9": + "1": transformers_trainer + "12": 0.28.1 + "13": linux-x86_64 +accelerator_config: + value: + dispatch_batches: null + even_batches: true + gradient_accumulation_kwargs: null + non_blocking: false + split_batches: false + use_seedable_sampler: true +activation: + value: waleed10 +adam_beta1: + value: 0.9 +adam_beta2: + value: 0.999 +adam_epsilon: + value: 1e-08 +architectures: + value: null +attention_bias: + value: false +attention_dropout: + value: 0 +auto_find_batch_size: + value: false +average_tokens_across_devices: + value: true +batch_eval_metrics: + value: false +bf16: + value: true +bf16_full_eval: + value: false +bos_token_id: + value: 1 +chunk_size_feed_forward: + value: 0 +data_seed: + value: 42 +dataloader_drop_last: + value: false +dataloader_in_order: + value: true +dataloader_multiprocessing_context: + value: null +dataloader_num_workers: + value: 0 +dataloader_persistent_workers: + value: false +dataloader_pin_memory: + value: true +dataloader_prefetch_factor: + value: null +ddp_backend: + value: null +ddp_broadcast_buffers: + value: null +ddp_bucket_cap_mb: + value: null +ddp_find_unused_parameters: + value: null +ddp_static_graph: + value: null +ddp_timeout: + value: 1800 +debug: + value: [] +deepspeed: + value: null +disable_tqdm: + value: false +do_eval: + value: true +do_predict: + value: false +do_train: + value: false +dtype: + value: null +enable_jit_checkpoint: + value: false +eos_token_id: + value: 2 +eval_accumulation_steps: + value: null +eval_delay: + value: 0 +eval_do_concat_batches: + value: true +eval_on_start: + value: false +eval_steps: + value: 60000 +eval_strategy: + value: steps +eval_use_gather_object: + value: false +fp16: + value: false +fp16_full_eval: + value: false +fsdp: + value: null +fsdp_config: + value: null +full_determinism: + value: false +gradient_accumulation_steps: + value: 4 +gradient_checkpointing: + value: false +gradient_checkpointing_kwargs: + value: null +greater_is_better: + value: null +head_dim: + value: 32 +hidden_act: + value: silu +hidden_size: + value: 128 +hub_always_push: + value: false +hub_model_id: + value: w-ahmad/A-glu-waleed10-21L +hub_private_repo: + value: null +hub_revision: + value: null +hub_strategy: + value: every_save +hub_token: + value: +id2label: + value: + "0": LABEL_0 + "1": LABEL_1 +ignore_data_skip: + value: false +include_for_metrics: + value: [] +include_num_input_tokens_seen: + value: "no" +initializer_range: + value: 0.02 +intermediate_size: + value: 256 +is_encoder_decoder: + value: false +label_names: + value: null +label_smoothing_factor: + value: 0 +label2id: + value: + LABEL_0: 0 + LABEL_1: 1 +learning_rate: + value: 0.001 +length_column_name: + value: length +liger_kernel_config: + value: null +load_best_model_at_end: + value: false +local_rank: + value: -1 +log_level: + value: passive +log_level_replica: + value: warning +log_on_each_node: + value: true +logging_first_step: + value: false +logging_nan_inf_filter: + value: true +logging_steps: + value: 20 +logging_strategy: + value: steps +lr_scheduler_kwargs: + value: null +lr_scheduler_type: + value: constant +max_grad_norm: + value: 1 +max_position_embeddings: + value: 512 +max_steps: + value: 1000 +metric_for_best_model: + value: null +mlp_bias: + value: false +mlp_type: + value: glu +model/num_parameters: + value: 3970432 +model_type: + value: tiny_llama +neftune_noise_alpha: + value: null +num_attention_heads: + value: 4 +num_hidden_layers: + value: 21 +num_key_value_heads: + value: 4 +num_train_epochs: + value: 1 +optim: + value: adamw_torch +optim_args: + value: null +optim_target_modules: + value: null +output_attentions: + value: false +output_dir: + value: out/glu-waleed10-21L_run +output_hidden_states: + value: false +pad_token_id: + value: 0 +parallelism_config: + value: null +per_device_eval_batch_size: + value: 128 +per_device_train_batch_size: + value: 128 +prediction_loss_only: + value: false +pretraining_tp: + value: 1 +problem_type: + value: null +project: + value: huggingface +push_to_hub: + value: false +remove_unused_columns: + value: false +report_to: + value: + - wandb +restore_callback_states_from_checkpoint: + value: false +resume_from_checkpoint: + value: null +return_dict: + value: true +rms_norm_eps: + value: 1e-06 +rope_parameters: + value: + rope_theta: 10000 + rope_type: default +run_name: + value: LM-glu-waleed10-21L-4.0M-20260814-213152 +save_on_each_node: + value: false +save_only_model: + value: false +save_steps: + value: 100 +save_strategy: + value: steps +save_total_limit: + value: null +seed: + value: 42 +skip_memory_metrics: + value: true +tf32: + value: null +tie_word_embeddings: + value: true +tokenizer_name: + value: w-ahmad/tiny-stories-tokenizer +torch_compile: + value: false +torch_compile_backend: + value: null +torch_compile_mode: + value: null +torch_empty_cache_steps: + value: null +trackio_bucket_id: + value: null +trackio_space_id: + value: null +trackio_static_space_id: + value: null +train_sampling_strategy: + value: random +transformers_version: + value: 5.16.0.dev0 +use_cache: + value: false +use_cpu: + value: false +use_liger_kernel: + value: false +vocab_size: + value: 4096 +waleed_beta: + value: 10 +warmup_steps: + value: 0 +weight_decay: + value: 0.01 diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/output.log b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/output.log new file mode 100644 index 0000000000000000000000000000000000000000..e81dadd30541a5d77ef5169cc8a5615883bf11eb --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/output.log @@ -0,0 +1,112 @@ + 0%| | 0/1000 [00:00>> FINISHED glu-waleed10-21L successfully diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/requirements.txt b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..f36b19895172d9637438a445ceacb71d09dc5d02 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/requirements.txt @@ -0,0 +1,149 @@ +asttokens==3.0.1 +comm==0.2.3 +debugpy==1.8.21 +decorator==5.3.1 +executing==2.2.1 +nest-asyncio==1.6.0 +parso==0.8.7 +platformdirs==4.11.0 +psutil==7.2.2 +ptyprocess==0.7.0 +pure_eval==0.2.3 +Pygments==2.20.0 +pyzmq==27.1.0 +setuptools==83.0.0 +six==1.17.0 +tornado==6.5.7 +traitlets==5.15.0 +fsspec==2026.4.0 +wcwidth==0.8.2 +ipython_pygments_lexers==1.1.1 +jedi==0.20.0 +jupyter_core==5.9.1 +matplotlib-inline==0.2.2 +pexpect==4.9.0 +prompt_toolkit==3.0.53 +python-dateutil==2.9.0.post0 +stack_data==0.6.3 +wheel==0.47.0 +jupyter_client==8.9.1 +pip==26.1.2 +ipython==9.15.0 +ipykernel==7.2.0 +threadpoolctl==3.6.0 +pyparsing==3.3.2 +typing_extensions==4.15.0 +Jinja2==3.1.6 +narwhals==2.24.0 +kiwisolver==1.5.0 +joblib==1.5.3 +fonttools==4.63.0 +cycler==0.12.1 +scipy==1.17.1 +pandas==3.0.5 +contourpy==1.3.3 +scikit-learn==1.9.0 +matplotlib==3.11.1 +urllib3==2.7.0 +tqdm==4.70.0 +idna==3.18 +charset-normalizer==3.4.9 +certifi==2026.7.22 +requests==2.34.2 +seaborn==0.13.2 +uv==0.12.0 +shellingham==1.5.4 +mpmath==1.3.0 +attrs==26.1.0 +hf-xet==1.5.2 +nvidia-nccl-cu12==2.21.5 +MarkupSafe==3.0.3 +regex==2026.7.19 +importlib_metadata==9.0.0 +httpcore==1.0.9 +annotated-doc==0.0.5 +multidict==6.7.1 +aiohttp==3.14.3 +aiosignal==1.4.0 +xxhash==3.8.1 +aiohappyeyeballs==2.7.1 +mdurl==0.1.2 +cuda-toolkit==13.0.3.0 +networkx==3.6.1 +PyYAML==6.0.3 +nvidia-cufile==1.15.1.6 +typer==0.27.0 +torchaudio==2.6.0+cu124 +rich==15.0.0 +nvidia-cufft-cu12==11.2.1.3 +h11==0.16.0 +dill==0.4.1 +cuda-pathfinder==1.6.0 +filelock==3.29.0 +nvidia-nvtx-cu12==12.4.127 +httpx==0.28.1 +anyio==4.14.2 +numpy==2.4.4 +yarl==1.24.5 +click==8.4.2 +triton==3.2.0 +frozenlist==1.8.0 +zipp==4.1.0 +propcache==0.5.2 +tokenizers==0.22.2 +markdown-it-py==4.2.0 +nvidia-cuda-runtime==13.0.96 +cuda-bindings==13.3.1 +nvidia-cuda-cupti==13.0.85 +torch==2.6.0+cu124 +multiprocess==0.70.19 +pillow==12.2.0 +transformers==5.16.0.dev0 +wandb==0.28.1 +nvidia-curand==10.4.0.35 +sympy==1.13.1 +nvidia-cusparse==12.6.3.3 +nvidia-cuda-nvrtc==13.0.88 +typing-inspection==0.4.2 +nvidia-cusolver==12.0.4.66 +nvidia-cufft==12.0.0.61 +nvidia-cudnn-cu13==9.20.0.48 +nvidia-cublas==13.1.1.3 +pyarrow==25.0.0 +evaluate==0.4.6 +diffusers==0.39.0 +pydantic==2.13.4 +annotated-types==0.8.0 +protobuf==7.35.1 +sentry-sdk==2.66.1 +einops==0.8.2 +packaging==26.2 +nvidia-nvjitlink-cu12==12.4.127 +nvidia-curand-cu12==10.3.5.147 +nvidia-cusparselt-cu12==0.6.2 +nvidia-cusparse-cu12==12.3.1.170 +nvidia-cuda-runtime-cu12==12.4.127 +torchvision==0.21.0+cu124 +nvidia-cuda-nvrtc-cu12==12.4.127 +nvidia-cuda-cupti-cu12==12.4.127 +nvidia-cusolver-cu12==11.6.1.9 +nvidia-cublas-cu12==12.4.5.8 +nvidia-cudnn-cu12==9.1.0.70 +huggingface_hub==1.26.0 +datasets==5.0.1 +safetensors==0.8.0 +accelerate==1.14.0 +pydantic_core==2.46.4 +ninja==1.13.0 +autocommand==2.2.2 +backports.tarfile==1.2.0 +importlib_metadata==8.7.1 +jaraco.text==4.0.0 +jaraco.context==6.1.0 +jaraco.functools==4.4.0 +more-itertools==10.8.0 +packaging==26.0 +platformdirs==4.4.0 +tomli==2.4.0 +wheel==0.46.3 +zipp==3.23.0 diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/wandb-metadata.json b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/wandb-metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6a4b06b6a141722cfa9f18682569db3f8021ed82 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/wandb-metadata.json @@ -0,0 +1,101 @@ +{ + "os": "Linux-5.15.0-126-generic-x86_64-with-glibc2.35", + "python": "CPython 3.11.15", + "startedAt": "2026-08-14T21:31:53.799629Z", + "args": [ + "--config", + "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml", + "--variants", + "glu-waleed10", + "glu-silu-waleed10", + "mlp-waleed10", + "mlp-silu-waleed10", + "glu-waleed", + "glu-situglu_low", + "glu-waleedglu_low" + ], + "program": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py", + "codePath": "sweep.py", + "codePathLocal": "sweep.py", + "git": { + "remote": "https://github.com/w-ahmad1a10/Activation.git", + "commit": "3fb47c6943d6927eafc39bd399a3d3a520bae74a" + }, + "email": "deepnevro@gmail.com", + "root": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation", + "host": "deeplens-k3s-node1", + "executable": "/mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python", + "cpu_count": 112, + "cpu_count_logical": 224, + "gpu": "NVIDIA H100 80GB HBM3", + "gpu_count": 8, + "disk": { + "/": { + "total": "1560765693952", + "used": "709060804608" + } + }, + "memory": { + "total": "2164089937920" + }, + "gpu_nvidia": [ + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-39c684a5-fde6-83d7-1663-0859795881ae" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-132944c4-b689-2b5f-89a4-d730401677ab" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bc6c3e3c-9b90-09ca-c034-774961847c54" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea" + } + ], + "cudaVersion": "12.4", + "writerId": "avd7wjdsk6ji6qbb7ojy8wwedk5izmcm" +} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..4b794f990ef6f0d4c0aa2551328674ecf21d8878 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/files/wandb-summary.json @@ -0,0 +1 @@ +{"train/global_step":1000,"train_runtime":825.1777,"_runtime":834,"train/epoch":0.5392289026691831,"train_samples_per_second":620.472,"train_loss":11.851402679443359,"train/grad_norm":2.09375,"train_steps_per_second":1.212,"eval/loss":2.1011769771575928,"total_flos":5.420315836416e+15,"eval/samples_per_second":982.674,"eval/steps_per_second":7.736,"eval/runtime":9.695,"_step":52,"train/learning_rate":0.001,"_wandb":{"runtime":834},"train/loss":8.39569091796875,"_timestamp":1.7867439483697834e+09} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug-core.log b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug-core.log new file mode 100644 index 0000000000000000000000000000000000000000..588056df10bc8cabd5e2682287953b3e2b130ffe --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug-core.log @@ -0,0 +1,62 @@ +{"time":"2026-08-14T21:31:53.539152008Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp650y3sc2/port-1229423.txt","pid":1229423,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false} +{"time":"2026-08-14T21:31:53.540676876Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1229423} +{"time":"2026-08-14T21:31:53.540659035Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T21:31:53.717724383Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"} +{"time":"2026-08-14T21:31:53.802092564Z","level":"INFO","msg":"handleInformInit: received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:54.062397411Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:59.411924652Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.396635188Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.929834053Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.931620539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:48.932491197Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:49.464388506Z","level":"INFO","msg":"handleInformInit: received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:49.733853364Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:55.329792921Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.158537572Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.914569559Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.916156477Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:01.917534874Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:02.524868682Z","level":"INFO","msg":"handleInformInit: received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:02.786199875Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:08.192858634Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:53.643379161Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.173579992Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.17490129Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.175479186Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.694802218Z","level":"INFO","msg":"handleInformInit: received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:11:54.956642682Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:12:00.441973218Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:55.662446346Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.296327526Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.297810539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.29856977Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.82560196Z","level":"INFO","msg":"handleInformInit: received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:24:57.085772231Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:25:02.571913681Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:28.612623637Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.208727987Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.210106262Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.210705504Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.839594174Z","level":"INFO","msg":"handleInformInit: received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:30.099444184Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:35.620666929Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:17.704425638Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.35448297Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.355720686Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.356578583Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.899799881Z","level":"INFO","msg":"handleInformInit: received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:19.159494071Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:24.671083701Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.214192173Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.783107435Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.784410941Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:51.78516667Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088192277Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088229648Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088239651Z","level":"INFO","msg":"connection: closing","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088243873Z","level":"INFO","msg":"server: is shutting down"} +{"time":"2026-08-14T23:05:52.0882852Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088299095Z","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088345917Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088452051Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T23:05:52.088520571Z","level":"INFO","msg":"server: all connections closed"} diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug-internal.log new file mode 100644 index 0000000000000000000000000000000000000000..19ddb84d841b3e1d04c8bf174b9bd4bf062b9f33 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug-internal.log @@ -0,0 +1,127 @@ +{"time":"2026-08-14T21:31:53.802248642Z","level":"INFO","msg":"wandb-core"} +{"time":"2026-08-14T21:31:53.802536305Z","level":"INFO","msg":"stream: starting","core version":"0.28.1"} +{"time":"2026-08-14T21:31:54.062231678Z","level":"INFO","msg":"stream: created new stream","id":"q9vcmje5"} +{"time":"2026-08-14T21:31:54.062319994Z","level":"INFO","msg":"handler: started"} +{"time":"2026-08-14T21:31:54.062390309Z","level":"INFO","msg":"stream: started"} +{"time":"2026-08-14T21:31:54.062412718Z","level":"INFO","msg":"writer: started","stream_id":"q9vcmje5"} +{"time":"2026-08-14T21:31:54.062420181Z","level":"INFO","msg":"sender: started"} +{"time":"2026-08-14T21:31:54.416838412Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:31:54.510348438Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:09.41771802Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":0,"events_lines":1,"console_offset":0,"console_lines":3,"uploaded_len":2} +{"time":"2026-08-14T21:32:09.578280298Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:24.417605975Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":0,"history_lines":1,"events_offset":1,"events_lines":2,"console_offset":2,"console_lines":2} +{"time":"2026-08-14T21:32:24.549935438Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:39.41719538Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":1,"history_lines":1,"events_offset":3,"events_lines":2,"console_offset":2,"console_lines":1} +{"time":"2026-08-14T21:32:39.566802488Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:32:54.417058222Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":5,"events_lines":2,"console_offset":2,"console_lines":1} +{"time":"2026-08-14T21:32:54.526885785Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:09.417361811Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":2,"history_lines":1,"events_offset":7,"events_lines":2,"console_offset":2,"console_lines":1} +{"time":"2026-08-14T21:33:09.541292122Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:24.41766197Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":3,"history_lines":1,"events_offset":9,"events_lines":2,"console_offset":2,"console_lines":1} +{"time":"2026-08-14T21:33:24.549198743Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:39.417110044Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":11,"events_lines":2,"console_offset":2,"console_lines":1} +{"time":"2026-08-14T21:33:39.551776959Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:33:54.417538658Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":4,"history_lines":1,"events_offset":13,"events_lines":2,"console_offset":2,"console_lines":1} +{"time":"2026-08-14T21:33:54.555356936Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:09.417890895Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":5,"history_lines":1,"events_offset":15,"events_lines":2,"console_offset":4,"console_lines":10} +{"time":"2026-08-14T21:34:09.592684206Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:24.417179414Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":6,"history_lines":1,"events_offset":17,"events_lines":2,"console_offset":12,"console_lines":1} +{"time":"2026-08-14T21:34:24.534710617Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:39.417073445Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":19,"events_lines":2,"console_offset":12,"console_lines":1} +{"time":"2026-08-14T21:34:39.570434803Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:34:54.417818476Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":7,"history_lines":1,"events_offset":21,"events_lines":2,"console_offset":12,"console_lines":1} +{"time":"2026-08-14T21:34:54.541208219Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:35:09.41711858Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":8,"history_lines":1,"events_offset":23,"events_lines":2,"console_offset":12,"console_lines":1} +{"time":"2026-08-14T21:35:09.54503893Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:35:24.417709905Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":9,"history_lines":1,"events_offset":25,"events_lines":2,"console_offset":12,"console_lines":1} +{"time":"2026-08-14T21:35:24.538800467Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:35:39.417929262Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":10,"history_lines":1,"events_offset":27,"events_lines":2,"console_offset":14,"console_lines":10} +{"time":"2026-08-14T21:35:39.553593527Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:35:54.417850428Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":1,"events_offset":29,"events_lines":2,"console_offset":22,"console_lines":1} +{"time":"2026-08-14T21:35:54.591775915Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:36:09.417299708Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":12,"history_lines":1,"events_offset":31,"events_lines":2,"console_offset":22,"console_lines":1} +{"time":"2026-08-14T21:36:09.609766076Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:36:24.417329624Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":22,"console_lines":1} +{"time":"2026-08-14T21:36:24.528837524Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:36:39.417160496Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":14,"history_lines":1,"events_offset":35,"events_lines":2,"console_offset":22,"console_lines":1} +{"time":"2026-08-14T21:36:39.539248212Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:36:54.417072248Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":15,"history_lines":1,"events_offset":37,"events_lines":2,"console_offset":24,"console_lines":10} +{"time":"2026-08-14T21:36:54.609691461Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:37:09.417049312Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":1,"events_offset":39,"events_lines":2,"console_offset":32,"console_lines":1} +{"time":"2026-08-14T21:37:09.684843081Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:37:24.417401277Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":17,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":32,"console_lines":1} +{"time":"2026-08-14T21:37:24.537403552Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:37:39.417411849Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":18,"history_lines":1,"events_offset":43,"events_lines":2,"console_offset":32,"console_lines":1} +{"time":"2026-08-14T21:37:39.548230786Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:37:54.417198892Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":32,"console_lines":1} +{"time":"2026-08-14T21:37:54.542467252Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:38:09.417264097Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":20,"history_lines":1,"events_offset":47,"events_lines":2,"console_offset":34,"console_lines":10} +{"time":"2026-08-14T21:38:09.674000615Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:38:24.417681287Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":1,"events_offset":49,"events_lines":2,"console_offset":42,"console_lines":1} +{"time":"2026-08-14T21:38:24.609965411Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:38:39.416972046Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":22,"history_lines":1,"events_offset":51,"events_lines":2,"console_offset":42,"console_lines":1} +{"time":"2026-08-14T21:38:39.549505398Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:38:54.41796769Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":23,"history_lines":1,"events_offset":53,"events_lines":2,"console_offset":42,"console_lines":1} +{"time":"2026-08-14T21:38:54.528895634Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:39:09.417164029Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":1,"events_offset":55,"events_lines":2,"console_offset":42,"console_lines":1} +{"time":"2026-08-14T21:39:09.602030487Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:39:24.417256613Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":25,"history_lines":1,"events_offset":57,"events_lines":2,"console_offset":44,"console_lines":10} +{"time":"2026-08-14T21:39:24.553135287Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:39:39.417138993Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":1,"events_offset":59,"events_lines":2,"console_offset":52,"console_lines":1} +{"time":"2026-08-14T21:39:39.529219292Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:39:54.417226385Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":27,"history_lines":1,"events_offset":61,"events_lines":2,"console_offset":52,"console_lines":1} +{"time":"2026-08-14T21:39:54.536465485Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:40:09.417043824Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":28,"history_lines":1,"events_offset":63,"events_lines":2,"console_offset":52,"console_lines":1} +{"time":"2026-08-14T21:40:09.68305832Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:40:24.41710473Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":65,"events_lines":2,"console_offset":52,"console_lines":1} +{"time":"2026-08-14T21:40:24.529074175Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:40:39.417102784Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":1,"events_offset":67,"events_lines":2,"console_offset":52,"console_lines":1} +{"time":"2026-08-14T21:40:39.548617714Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:40:54.417536563Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":30,"history_lines":1,"events_offset":69,"events_lines":2,"console_offset":54,"console_lines":10} +{"time":"2026-08-14T21:40:54.540414529Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:41:09.417723797Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":1,"events_offset":71,"events_lines":2,"console_offset":62,"console_lines":1} +{"time":"2026-08-14T21:41:09.626110005Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:41:24.417000398Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":32,"history_lines":1,"events_offset":73,"events_lines":2,"console_offset":62,"console_lines":1} +{"time":"2026-08-14T21:41:24.58002245Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:41:39.41726191Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":1,"events_offset":75,"events_lines":2,"console_offset":62,"console_lines":1} +{"time":"2026-08-14T21:41:39.526092005Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:41:54.417072409Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":34,"history_lines":1,"events_offset":77,"events_lines":2,"console_offset":62,"console_lines":1} +{"time":"2026-08-14T21:41:54.546246802Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:42:09.417123682Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":1,"events_offset":79,"events_lines":2,"console_offset":64,"console_lines":10} +{"time":"2026-08-14T21:42:09.666318822Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:42:24.417686885Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":36,"history_lines":1,"events_offset":81,"events_lines":2,"console_offset":72,"console_lines":1} +{"time":"2026-08-14T21:42:24.564587405Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:42:39.41722149Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":1,"events_offset":83,"events_lines":2,"console_offset":72,"console_lines":1} +{"time":"2026-08-14T21:42:39.582243173Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:42:54.417586987Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":38,"history_lines":1,"events_offset":85,"events_lines":2,"console_offset":72,"console_lines":1} +{"time":"2026-08-14T21:42:54.557667192Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:43:09.417207023Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":87,"events_lines":2,"console_offset":72,"console_lines":1} +{"time":"2026-08-14T21:43:09.610669065Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:43:24.417672133Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":1,"events_offset":89,"events_lines":2,"console_offset":74,"console_lines":10} +{"time":"2026-08-14T21:43:24.574587398Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:43:39.417277064Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":41,"history_lines":1,"events_offset":91,"events_lines":2,"console_offset":82,"console_lines":1} +{"time":"2026-08-14T21:43:39.572880967Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:43:54.417379547Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":1,"events_offset":93,"events_lines":2,"console_offset":82,"console_lines":1} +{"time":"2026-08-14T21:43:54.581169619Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:44:09.417088108Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":43,"history_lines":1,"events_offset":95,"events_lines":2,"console_offset":82,"console_lines":1} +{"time":"2026-08-14T21:44:09.637501534Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:44:24.417552656Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":1,"events_offset":97,"events_lines":2,"console_offset":82,"console_lines":1} +{"time":"2026-08-14T21:44:24.556148219Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:44:39.417501955Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":45,"history_lines":1,"events_offset":99,"events_lines":2,"console_offset":84,"console_lines":10} +{"time":"2026-08-14T21:44:39.686483667Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:44:54.417133976Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":101,"events_lines":2,"console_offset":92,"console_lines":1} +{"time":"2026-08-14T21:44:54.554631841Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:45:09.41749322Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":1,"events_offset":103,"events_lines":2,"console_offset":92,"console_lines":1} +{"time":"2026-08-14T21:45:09.642217292Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:45:24.417641844Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":48,"history_lines":1,"events_offset":105,"events_lines":2,"console_offset":92,"console_lines":1} +{"time":"2026-08-14T21:45:24.579697512Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:45:39.41776998Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":3,"events_offset":107,"events_lines":2,"console_offset":92,"console_lines":1} +{"time":"2026-08-14T21:45:39.679514648Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:45:48.819105986Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T21:45:48.819431393Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":52,"history_lines":1,"events_offset":109,"events_lines":1,"console_offset":94,"console_lines":18,"uploaded_len":3,"complete":true,"exit_code":0} +{"time":"2026-08-14T21:45:48.927022721Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:45:48.928319845Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T21:45:48.931654382Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T21:45:48.931674323Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T21:45:48.931746361Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T21:45:48.931750382Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug.log b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..de1a29b0d2a39e4b475bf5b8a3afd97deddbe837 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug.log @@ -0,0 +1,27 @@ +2026-08-14 21:31:53,800 INFO MainThread:1229423 [wandb_setup.py:_flush():81] Current SDK version is 0.28.1 +2026-08-14 21:31:53,800 INFO MainThread:1229423 [wandb_setup.py:_flush():81] Configure stats pid to 1229423 +2026-08-14 21:31:53,800 INFO MainThread:1229423 [wandb_setup.py:_flush():81] Loading settings from environment variables +2026-08-14 21:31:53,800 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():729] Logging user logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug.log +2026-08-14 21:31:53,800 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():730] Logging internal logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_213153-q9vcmje5/logs/debug-internal.log +2026-08-14 21:31:53,800 INFO MainThread:1229423 [wandb_init.py:init():772] calling init triggers +2026-08-14 21:31:53,801 INFO MainThread:1229423 [wandb_init.py:init():777] wandb.init called with sweep_config: {} +config: {'_wandb': {}} +2026-08-14 21:31:53,801 INFO MainThread:1229423 [wandb_init.py:init():820] starting backend +2026-08-14 21:31:53,801 INFO MainThread:1229423 [wandb_init.py:init():835] sending inform_init request +2026-08-14 21:31:54,062 INFO MainThread:1229423 [wandb_init.py:init():840] backend started and connected +2026-08-14 21:31:54,063 INFO MainThread:1229423 [wandb_init.py:init():910] updated telemetry +2026-08-14 21:31:54,070 INFO MainThread:1229423 [wandb_init.py:init():933] communicating run to backend with 90.0 second timeout +2026-08-14 21:31:54,336 INFO MainThread:1229423 [wandb_init.py:init():978] starting run threads in backend +2026-08-14 21:31:54,409 INFO MainThread:1229423 [wandb_run.py:_console_start():2621] atexit reg +2026-08-14 21:31:54,409 INFO MainThread:1229423 [wandb_run.py:_redirect():2471] redirect: wrap_raw +2026-08-14 21:31:54,409 INFO MainThread:1229423 [wandb_run.py:_redirect():2540] Wrapping output streams. +2026-08-14 21:31:54,409 INFO MainThread:1229423 [wandb_run.py:_redirect():2563] Redirects installed. +2026-08-14 21:31:54,412 INFO MainThread:1229423 [wandb_init.py:init():1016] run started, returning control to user process +2026-08-14 21:31:54,413 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 21, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'glu', 'activation': 'waleed10', 'waleed_beta': 10.0, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/glu-waleed10-21L_run', 'per_device_train_batch_size': 128, 'num_train_epochs': 1, 'max_steps': 1000, 'learning_rate': 0.001, 'lr_scheduler_type': 'constant', 'lr_scheduler_kwargs': None, 'warmup_steps': 0, 'optim': 'adamw_torch', 'optim_args': None, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 4, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-glu-waleed10-21L-4.0M-20260814-213152', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 60000, 'eval_delay': 0, 'per_device_eval_batch_size': 128, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 100, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/A-glu-waleed10-21L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} +2026-08-14 21:31:54,415 INFO MainThread:1229423 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 3970432 - > +2026-08-14 21:31:54,415 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 3970432 None +2026-08-14 21:45:48,395 INFO MainThread:1229423 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research1/q9vcmje5 +2026-08-14 21:45:48,396 INFO MainThread:1229423 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 21:45:48,396 INFO MainThread:1229423 [wandb_run.py:_restore():2570] restore +2026-08-14 21:45:48,396 INFO MainThread:1229423 [wandb_run.py:_restore():2576] restore done +2026-08-14 21:45:48,931 INFO MainThread:1229423 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_213153-q9vcmje5/run-q9vcmje5.wandb b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/run-q9vcmje5.wandb new file mode 100644 index 0000000000000000000000000000000000000000..6cde4afbcd913a61e192ad13d4b6f9f5d7fa0c79 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_213153-q9vcmje5/run-q9vcmje5.wandb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b17e37a7a3faacb5cb0161e4035cde89ebd95dc08857df4bc973553e68655012 +size 584687 diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/config.yaml b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..aea6b33d51a57b149527734c8f10b2c701ec7928 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/config.yaml @@ -0,0 +1,439 @@ +_name_or_path: + value: "" +_wandb: + value: + cli_version: 0.28.1 + e: + wtedta21p9zfavbt9r6hw8vts8caym1b: + args: + - --config + - /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml + - --variants + - glu-waleed10 + - glu-silu-waleed10 + - mlp-waleed10 + - mlp-silu-waleed10 + - glu-waleed + - glu-situglu_low + - glu-waleedglu_low + codePath: sweep.py + codePathLocal: sweep.py + cpu_count: 112 + cpu_count_logical: 224 + cudaVersion: "12.4" + disk: + /: + total: "1560765693952" + used: "709081337856" + email: deepnevro@gmail.com + executable: /mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python + git: + commit: 3fb47c6943d6927eafc39bd399a3d3a520bae74a + remote: https://github.com/w-ahmad1a10/Activation.git + gpu: NVIDIA H100 80GB HBM3 + gpu_count: 8 + gpu_nvidia: + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-39c684a5-fde6-83d7-1663-0859795881ae + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-132944c4-b689-2b5f-89a4-d730401677ab + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bc6c3e3c-9b90-09ca-c034-774961847c54 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea + host: deeplens-k3s-node1 + memory: + total: "2164089937920" + os: Linux-5.15.0-126-generic-x86_64-with-glibc2.35 + program: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py + python: CPython 3.11.15 + root: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation + startedAt: "2026-08-14T21:45:49.461932Z" + writerId: wtedta21p9zfavbt9r6hw8vts8caym1b + m: + - "1": train/global_step + "6": + - 3 + "7": [] + - "2": '*' + "5": 1 + "6": + - 1 + "7": [] + python_version: 3.11.15 + t: + "1": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "2": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "3": + - 2 + - 7 + - 13 + - 19 + - 62 + - 66 + "4": 3.11.15 + "5": 0.28.1 + "6": 5.16.0.dev0 + "9": + "1": transformers_trainer + "12": 0.28.1 + "13": linux-x86_64 +accelerator_config: + value: + dispatch_batches: null + even_batches: true + gradient_accumulation_kwargs: null + non_blocking: false + split_batches: false + use_seedable_sampler: true +activation: + value: silu-waleed10 +adam_beta1: + value: 0.9 +adam_beta2: + value: 0.999 +adam_epsilon: + value: 1e-08 +architectures: + value: null +attention_bias: + value: false +attention_dropout: + value: 0 +auto_find_batch_size: + value: false +average_tokens_across_devices: + value: true +batch_eval_metrics: + value: false +bf16: + value: true +bf16_full_eval: + value: false +bos_token_id: + value: 1 +chunk_size_feed_forward: + value: 0 +data_seed: + value: 42 +dataloader_drop_last: + value: false +dataloader_in_order: + value: true +dataloader_multiprocessing_context: + value: null +dataloader_num_workers: + value: 0 +dataloader_persistent_workers: + value: false +dataloader_pin_memory: + value: true +dataloader_prefetch_factor: + value: null +ddp_backend: + value: null +ddp_broadcast_buffers: + value: null +ddp_bucket_cap_mb: + value: null +ddp_find_unused_parameters: + value: null +ddp_static_graph: + value: null +ddp_timeout: + value: 1800 +debug: + value: [] +deepspeed: + value: null +disable_tqdm: + value: false +do_eval: + value: true +do_predict: + value: false +do_train: + value: false +dtype: + value: null +enable_jit_checkpoint: + value: false +eos_token_id: + value: 2 +eval_accumulation_steps: + value: null +eval_delay: + value: 0 +eval_do_concat_batches: + value: true +eval_on_start: + value: false +eval_steps: + value: 60000 +eval_strategy: + value: steps +eval_use_gather_object: + value: false +fp16: + value: false +fp16_full_eval: + value: false +fsdp: + value: null +fsdp_config: + value: null +full_determinism: + value: false +gradient_accumulation_steps: + value: 4 +gradient_checkpointing: + value: false +gradient_checkpointing_kwargs: + value: null +greater_is_better: + value: null +head_dim: + value: 32 +hidden_act: + value: silu +hidden_size: + value: 128 +hub_always_push: + value: false +hub_model_id: + value: w-ahmad/A-glu-silu-waleed10-21L +hub_private_repo: + value: null +hub_revision: + value: null +hub_strategy: + value: every_save +hub_token: + value: +id2label: + value: + "0": LABEL_0 + "1": LABEL_1 +ignore_data_skip: + value: false +include_for_metrics: + value: [] +include_num_input_tokens_seen: + value: "no" +initializer_range: + value: 0.02 +intermediate_size: + value: 256 +is_encoder_decoder: + value: false +label_names: + value: null +label_smoothing_factor: + value: 0 +label2id: + value: + LABEL_0: 0 + LABEL_1: 1 +learning_rate: + value: 0.001 +length_column_name: + value: length +liger_kernel_config: + value: null +load_best_model_at_end: + value: false +local_rank: + value: -1 +log_level: + value: passive +log_level_replica: + value: warning +log_on_each_node: + value: true +logging_first_step: + value: false +logging_nan_inf_filter: + value: true +logging_steps: + value: 20 +logging_strategy: + value: steps +lr_scheduler_kwargs: + value: null +lr_scheduler_type: + value: constant +max_grad_norm: + value: 1 +max_position_embeddings: + value: 512 +max_steps: + value: 1000 +metric_for_best_model: + value: null +mlp_bias: + value: false +mlp_type: + value: glu +model/num_parameters: + value: 3970432 +model_type: + value: tiny_llama +neftune_noise_alpha: + value: null +num_attention_heads: + value: 4 +num_hidden_layers: + value: 21 +num_key_value_heads: + value: 4 +num_train_epochs: + value: 1 +optim: + value: adamw_torch +optim_args: + value: null +optim_target_modules: + value: null +output_attentions: + value: false +output_dir: + value: out/glu-silu-waleed10-21L_run +output_hidden_states: + value: false +pad_token_id: + value: 0 +parallelism_config: + value: null +per_device_eval_batch_size: + value: 128 +per_device_train_batch_size: + value: 128 +prediction_loss_only: + value: false +pretraining_tp: + value: 1 +problem_type: + value: null +project: + value: huggingface +push_to_hub: + value: false +remove_unused_columns: + value: false +report_to: + value: + - wandb +restore_callback_states_from_checkpoint: + value: false +resume_from_checkpoint: + value: null +return_dict: + value: true +rms_norm_eps: + value: 1e-06 +rope_parameters: + value: + rope_theta: 10000 + rope_type: default +run_name: + value: LM-glu-silu-waleed10-21L-4.0M-20260814-214549 +save_on_each_node: + value: false +save_only_model: + value: false +save_steps: + value: 100 +save_strategy: + value: steps +save_total_limit: + value: null +seed: + value: 42 +skip_memory_metrics: + value: true +tf32: + value: null +tie_word_embeddings: + value: true +tokenizer_name: + value: w-ahmad/tiny-stories-tokenizer +torch_compile: + value: false +torch_compile_backend: + value: null +torch_compile_mode: + value: null +torch_empty_cache_steps: + value: null +trackio_bucket_id: + value: null +trackio_space_id: + value: null +trackio_static_space_id: + value: null +train_sampling_strategy: + value: random +transformers_version: + value: 5.16.0.dev0 +use_cache: + value: false +use_cpu: + value: false +use_liger_kernel: + value: false +vocab_size: + value: 4096 +waleed_beta: + value: 10 +warmup_steps: + value: 0 +weight_decay: + value: 0.01 diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/output.log b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/output.log new file mode 100644 index 0000000000000000000000000000000000000000..b14107892e691059fe4151738a5399f74029a07d --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/output.log @@ -0,0 +1,110 @@ + 10%|█ | 100/1000 [01:17<11:33, 1.30it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '28.93', 'grad_norm': '4.031', 'learning_rate': '0.001', 'epoch': '0.01078', 'train/total_time_seconds': '7.849', 'train/time_per_step_avg': '0.3925', 'train/epoch_time_elapsed': '15.51', 'train/estimated_remaining_minutes': '6.41'} +{'loss': '24.24', 'grad_norm': '1.328', 'learning_rate': '0.001', 'epoch': '0.02157', 'train/total_time_seconds': '15.63', 'train/time_per_step_avg': '0.3907', 'train/epoch_time_elapsed': '31.12', 'train/estimated_remaining_minutes': '6.251'} +{'loss': '22.72', 'grad_norm': '2.844', 'learning_rate': '0.001', 'epoch': '0.03235', 'train/total_time_seconds': '23.41', 'train/time_per_step_avg': '0.3902', 'train/epoch_time_elapsed': '46.51', 'train/estimated_remaining_minutes': '6.113'} +{'loss': '21.38', 'grad_norm': '2.031', 'learning_rate': '0.001', 'epoch': '0.04314', 'train/total_time_seconds': '31.19', 'train/time_per_step_avg': '0.3899', 'train/epoch_time_elapsed': '62.01', 'train/estimated_remaining_minutes': '5.979'} +{'loss': '19.96', 'grad_norm': '3.125', 'learning_rate': '0.001', 'epoch': '0.05392', 'train/total_time_seconds': '38.99', 'train/time_per_step_avg': '0.3899', 'train/epoch_time_elapsed': '77.38', 'train/estimated_remaining_minutes': '5.848'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 75.82it/s] + 20%|██ | 200/1000 [02:34<10:05, 1.32it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '18.43', 'grad_norm': '3.078', 'learning_rate': '0.001', 'epoch': '0.06471', 'train/total_time_seconds': '46.81', 'train/time_per_step_avg': '0.3896', 'train/epoch_time_elapsed': '92.97', 'train/estimated_remaining_minutes': '5.721'} +{'loss': '17.22', 'grad_norm': '3.656', 'learning_rate': '0.001', 'epoch': '0.07549', 'train/total_time_seconds': '54.61', 'train/time_per_step_avg': '0.3898', 'train/epoch_time_elapsed': '108.3', 'train/estimated_remaining_minutes': '5.591'} +{'loss': '16.27', 'grad_norm': '3.156', 'learning_rate': '0.001', 'epoch': '0.08628', 'train/total_time_seconds': '62.38', 'train/time_per_step_avg': '0.3896', 'train/epoch_time_elapsed': '124.1', 'train/estimated_remaining_minutes': '5.458'} +{'loss': '15.49', 'grad_norm': '2.359', 'learning_rate': '0.001', 'epoch': '0.09706', 'train/total_time_seconds': '70.13', 'train/time_per_step_avg': '0.3894', 'train/epoch_time_elapsed': '139.3', 'train/estimated_remaining_minutes': '5.325'} +{'loss': '14.8', 'grad_norm': '2.797', 'learning_rate': '0.001', 'epoch': '0.1078', 'train/total_time_seconds': '77.89', 'train/time_per_step_avg': '0.3891', 'train/epoch_time_elapsed': '154.6', 'train/estimated_remaining_minutes': '5.193'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.86it/s] + 30%|███ | 300/1000 [03:51<08:52, 1.31it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '14.22', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.1186', 'train/total_time_seconds': '85.67', 'train/time_per_step_avg': '0.3887', 'train/epoch_time_elapsed': '170.1', 'train/estimated_remaining_minutes': '5.062'} +{'loss': '13.73', 'grad_norm': '2.641', 'learning_rate': '0.001', 'epoch': '0.1294', 'train/total_time_seconds': '93.44', 'train/time_per_step_avg': '0.3883', 'train/epoch_time_elapsed': '185.5', 'train/estimated_remaining_minutes': '4.931'} +{'loss': '13.25', 'grad_norm': '2.172', 'learning_rate': '0.001', 'epoch': '0.1402', 'train/total_time_seconds': '101.2', 'train/time_per_step_avg': '0.3881', 'train/epoch_time_elapsed': '200.7', 'train/estimated_remaining_minutes': '4.8'} +{'loss': '12.85', 'grad_norm': '1.781', 'learning_rate': '0.001', 'epoch': '0.151', 'train/total_time_seconds': '109', 'train/time_per_step_avg': '0.3886', 'train/epoch_time_elapsed': '216.3', 'train/estimated_remaining_minutes': '4.671'} +{'loss': '12.48', 'grad_norm': '2.5', 'learning_rate': '0.001', 'epoch': '0.1618', 'train/total_time_seconds': '116.8', 'train/time_per_step_avg': '0.3888', 'train/epoch_time_elapsed': '231.6', 'train/estimated_remaining_minutes': '4.541'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 52.93it/s] + 40%|████ | 400/1000 [05:08<07:39, 1.31it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '12.14', 'grad_norm': '3.062', 'learning_rate': '0.001', 'epoch': '0.1726', 'train/total_time_seconds': '124.5', 'train/time_per_step_avg': '0.3886', 'train/epoch_time_elapsed': '246.9', 'train/estimated_remaining_minutes': '4.41'} +{'loss': '11.86', 'grad_norm': '2.359', 'learning_rate': '0.001', 'epoch': '0.1833', 'train/total_time_seconds': '132.3', 'train/time_per_step_avg': '0.3884', 'train/epoch_time_elapsed': '262.1', 'train/estimated_remaining_minutes': '4.279'} +{'loss': '11.56', 'grad_norm': '2.516', 'learning_rate': '0.001', 'epoch': '0.1941', 'train/total_time_seconds': '140.1', 'train/time_per_step_avg': '0.3888', 'train/epoch_time_elapsed': '277.5', 'train/estimated_remaining_minutes': '4.15'} +{'loss': '11.3', 'grad_norm': '2.234', 'learning_rate': '0.001', 'epoch': '0.2049', 'train/total_time_seconds': '147.9', 'train/time_per_step_avg': '0.3887', 'train/epoch_time_elapsed': '292.9', 'train/estimated_remaining_minutes': '4.021'} +{'loss': '11.09', 'grad_norm': '1.984', 'learning_rate': '0.001', 'epoch': '0.2157', 'train/total_time_seconds': '155.7', 'train/time_per_step_avg': '0.3888', 'train/epoch_time_elapsed': '308.3', 'train/estimated_remaining_minutes': '3.891'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 77.32it/s] + 50%|█████ | 500/1000 [06:25<06:19, 1.32it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.86', 'grad_norm': '3.016', 'learning_rate': '0.001', 'epoch': '0.2265', 'train/total_time_seconds': '163.5', 'train/time_per_step_avg': '0.3893', 'train/epoch_time_elapsed': '323.7', 'train/estimated_remaining_minutes': '3.762'} +{'loss': '10.72', 'grad_norm': '2.234', 'learning_rate': '0.001', 'epoch': '0.2373', 'train/total_time_seconds': '171.2', 'train/time_per_step_avg': '0.3896', 'train/epoch_time_elapsed': '339.4', 'train/estimated_remaining_minutes': '3.632'} +{'loss': '10.52', 'grad_norm': '2.328', 'learning_rate': '0.001', 'epoch': '0.248', 'train/total_time_seconds': '179', 'train/time_per_step_avg': '0.3892', 'train/epoch_time_elapsed': '354.7', 'train/estimated_remaining_minutes': '3.502'} +{'loss': '10.35', 'grad_norm': '2.938', 'learning_rate': '0.001', 'epoch': '0.2588', 'train/total_time_seconds': '186.7', 'train/time_per_step_avg': '0.3889', 'train/epoch_time_elapsed': '370', 'train/estimated_remaining_minutes': '3.372'} +{'loss': '10.2', 'grad_norm': '2.438', 'learning_rate': '0.001', 'epoch': '0.2696', 'train/total_time_seconds': '194.5', 'train/time_per_step_avg': '0.3885', 'train/epoch_time_elapsed': '385.3', 'train/estimated_remaining_minutes': '3.242'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 81.80it/s] + 60%|██████ | 600/1000 [07:42<05:09, 1.29it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.09', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.2804', 'train/total_time_seconds': '202.3', 'train/time_per_step_avg': '0.3884', 'train/epoch_time_elapsed': '400.9', 'train/estimated_remaining_minutes': '3.112'} +{'loss': '9.965', 'grad_norm': '2', 'learning_rate': '0.001', 'epoch': '0.2912', 'train/total_time_seconds': '210.1', 'train/time_per_step_avg': '0.3886', 'train/epoch_time_elapsed': '416.3', 'train/estimated_remaining_minutes': '2.983'} +{'loss': '9.853', 'grad_norm': '1.992', 'learning_rate': '0.001', 'epoch': '0.302', 'train/total_time_seconds': '217.9', 'train/time_per_step_avg': '0.389', 'train/epoch_time_elapsed': '431.8', 'train/estimated_remaining_minutes': '2.853'} +{'loss': '9.736', 'grad_norm': '1.852', 'learning_rate': '0.001', 'epoch': '0.3128', 'train/total_time_seconds': '225.7', 'train/time_per_step_avg': '0.3895', 'train/epoch_time_elapsed': '447.2', 'train/estimated_remaining_minutes': '2.724'} +{'loss': '9.628', 'grad_norm': '2.656', 'learning_rate': '0.001', 'epoch': '0.3235', 'train/total_time_seconds': '233.5', 'train/time_per_step_avg': '0.3899', 'train/epoch_time_elapsed': '462.9', 'train/estimated_remaining_minutes': '2.594'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.34it/s] + 70%|███████ | 700/1000 [08:59<03:50, 1.30it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.536', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.3343', 'train/total_time_seconds': '241.3', 'train/time_per_step_avg': '0.3897', 'train/epoch_time_elapsed': '478.3', 'train/estimated_remaining_minutes': '2.465'} +{'loss': '9.441', 'grad_norm': '2.031', 'learning_rate': '0.001', 'epoch': '0.3451', 'train/total_time_seconds': '249', 'train/time_per_step_avg': '0.3893', 'train/epoch_time_elapsed': '493.7', 'train/estimated_remaining_minutes': '2.335'} +{'loss': '9.348', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.3559', 'train/total_time_seconds': '256.8', 'train/time_per_step_avg': '0.389', 'train/epoch_time_elapsed': '509.1', 'train/estimated_remaining_minutes': '2.205'} +{'loss': '9.265', 'grad_norm': '2.234', 'learning_rate': '0.001', 'epoch': '0.3667', 'train/total_time_seconds': '264.6', 'train/time_per_step_avg': '0.3889', 'train/epoch_time_elapsed': '524.6', 'train/estimated_remaining_minutes': '2.075'} +{'loss': '9.19', 'grad_norm': '1.641', 'learning_rate': '0.001', 'epoch': '0.3775', 'train/total_time_seconds': '272.4', 'train/time_per_step_avg': '0.3888', 'train/epoch_time_elapsed': '540', 'train/estimated_remaining_minutes': '1.945'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 46.72it/s] + 80%|████████ | 800/1000 [10:17<02:35, 1.29it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.1', 'grad_norm': '2.156', 'learning_rate': '0.001', 'epoch': '0.3882', 'train/total_time_seconds': '280.1', 'train/time_per_step_avg': '0.3887', 'train/epoch_time_elapsed': '555.5', 'train/estimated_remaining_minutes': '1.816'} +{'loss': '9.055', 'grad_norm': '2.531', 'learning_rate': '0.001', 'epoch': '0.399', 'train/total_time_seconds': '287.9', 'train/time_per_step_avg': '0.3888', 'train/epoch_time_elapsed': '570.9', 'train/estimated_remaining_minutes': '1.686'} +{'loss': '8.981', 'grad_norm': '2.016', 'learning_rate': '0.001', 'epoch': '0.4098', 'train/total_time_seconds': '295.7', 'train/time_per_step_avg': '0.3889', 'train/epoch_time_elapsed': '586.4', 'train/estimated_remaining_minutes': '1.556'} +{'loss': '8.936', 'grad_norm': '2.359', 'learning_rate': '0.001', 'epoch': '0.4206', 'train/total_time_seconds': '303.5', 'train/time_per_step_avg': '0.3889', 'train/epoch_time_elapsed': '601.8', 'train/estimated_remaining_minutes': '1.427'} +{'loss': '8.839', 'grad_norm': '2.156', 'learning_rate': '0.001', 'epoch': '0.4314', 'train/total_time_seconds': '311.3', 'train/time_per_step_avg': '0.3893', 'train/epoch_time_elapsed': '617.3', 'train/estimated_remaining_minutes': '1.297'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 77.94it/s] + 90%|█████████ | 900/1000 [11:34<01:16, 1.30it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.804', 'grad_norm': '2.344', 'learning_rate': '0.001', 'epoch': '0.4422', 'train/total_time_seconds': '319.1', 'train/time_per_step_avg': '0.3893', 'train/epoch_time_elapsed': '632.8', 'train/estimated_remaining_minutes': '1.167'} +{'loss': '8.767', 'grad_norm': '2.281', 'learning_rate': '0.001', 'epoch': '0.453', 'train/total_time_seconds': '326.9', 'train/time_per_step_avg': '0.3898', 'train/epoch_time_elapsed': '648.4', 'train/estimated_remaining_minutes': '1.038'} +{'loss': '8.691', 'grad_norm': '2.438', 'learning_rate': '0.001', 'epoch': '0.4637', 'train/total_time_seconds': '334.7', 'train/time_per_step_avg': '0.3898', 'train/epoch_time_elapsed': '663.8', 'train/estimated_remaining_minutes': '0.908'} +{'loss': '8.652', 'grad_norm': '1.828', 'learning_rate': '0.001', 'epoch': '0.4745', 'train/total_time_seconds': '342.4', 'train/time_per_step_avg': '0.3895', 'train/epoch_time_elapsed': '679.2', 'train/estimated_remaining_minutes': '0.7782'} +{'loss': '8.596', 'grad_norm': '2.5', 'learning_rate': '0.001', 'epoch': '0.4853', 'train/total_time_seconds': '350.2', 'train/time_per_step_avg': '0.3895', 'train/epoch_time_elapsed': '694.6', 'train/estimated_remaining_minutes': '0.6486'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 76.93it/s] +100%|██████████| 1000/1000 [13:01<00:00, 1.30i[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.548', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.4961', 'train/total_time_seconds': '358', 'train/time_per_step_avg': '0.3897', 'train/epoch_time_elapsed': '710.2', 'train/estimated_remaining_minutes': '0.5189'} +{'loss': '8.527', 'grad_norm': '2.188', 'learning_rate': '0.001', 'epoch': '0.5069', 'train/total_time_seconds': '365.9', 'train/time_per_step_avg': '0.3897', 'train/epoch_time_elapsed': '725.5', 'train/estimated_remaining_minutes': '0.3892'} +{'loss': '8.503', 'grad_norm': '1.906', 'learning_rate': '0.001', 'epoch': '0.5177', 'train/total_time_seconds': '373.7', 'train/time_per_step_avg': '0.39', 'train/epoch_time_elapsed': '741', 'train/estimated_remaining_minutes': '0.2595'} +{'loss': '8.467', 'grad_norm': '2.141', 'learning_rate': '0.001', 'epoch': '0.5284', 'train/total_time_seconds': '381.5', 'train/time_per_step_avg': '0.3903', 'train/epoch_time_elapsed': '756.3', 'train/estimated_remaining_minutes': '0.1297'} +{'loss': '8.423', 'grad_norm': '2', 'learning_rate': '0.001', 'epoch': '0.5392', 'train/total_time_seconds': '389.3', 'train/time_per_step_avg': '0.3902', 'train/epoch_time_elapsed': '771.8', 'train/estimated_remaining_minutes': '0'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes +{'eval_loss': '2.102', 'eval_runtime': '9.474', 'eval_samples_per_second': '1006', 'eval_steps_per_second': '7.916', 'epoch': '0.5392', 'train/total_time_seconds': '389.3', 'train/time_per_step_avg': '0.3902', 'train/epoch_time_elapsed': '781.3', 'train/estimated_remaining_minutes': '0'} + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.01it/s] +100%|██████████| 1000/1000 [13:01<00:00, 1.28it/s], ?it/s] +{'train_runtime': '782.2', 'train_samples_per_second': '654.5', 'train_steps_per_second': '1.278', 'train_loss': '12.07', 'epoch': '0.5392', 'train/total_time_seconds': '389.3', 'train/time_per_step_avg': '0.3902', 'train/epoch_time_elapsed': '781.3', 'train/estimated_remaining_minutes': '0'} +100%|██████████| 75/75 [00:09<00:00, 8.08it/s] +[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 81.19it/s] +>>> FINISHED glu-silu-waleed10-21L successfully diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/requirements.txt b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..f36b19895172d9637438a445ceacb71d09dc5d02 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/requirements.txt @@ -0,0 +1,149 @@ +asttokens==3.0.1 +comm==0.2.3 +debugpy==1.8.21 +decorator==5.3.1 +executing==2.2.1 +nest-asyncio==1.6.0 +parso==0.8.7 +platformdirs==4.11.0 +psutil==7.2.2 +ptyprocess==0.7.0 +pure_eval==0.2.3 +Pygments==2.20.0 +pyzmq==27.1.0 +setuptools==83.0.0 +six==1.17.0 +tornado==6.5.7 +traitlets==5.15.0 +fsspec==2026.4.0 +wcwidth==0.8.2 +ipython_pygments_lexers==1.1.1 +jedi==0.20.0 +jupyter_core==5.9.1 +matplotlib-inline==0.2.2 +pexpect==4.9.0 +prompt_toolkit==3.0.53 +python-dateutil==2.9.0.post0 +stack_data==0.6.3 +wheel==0.47.0 +jupyter_client==8.9.1 +pip==26.1.2 +ipython==9.15.0 +ipykernel==7.2.0 +threadpoolctl==3.6.0 +pyparsing==3.3.2 +typing_extensions==4.15.0 +Jinja2==3.1.6 +narwhals==2.24.0 +kiwisolver==1.5.0 +joblib==1.5.3 +fonttools==4.63.0 +cycler==0.12.1 +scipy==1.17.1 +pandas==3.0.5 +contourpy==1.3.3 +scikit-learn==1.9.0 +matplotlib==3.11.1 +urllib3==2.7.0 +tqdm==4.70.0 +idna==3.18 +charset-normalizer==3.4.9 +certifi==2026.7.22 +requests==2.34.2 +seaborn==0.13.2 +uv==0.12.0 +shellingham==1.5.4 +mpmath==1.3.0 +attrs==26.1.0 +hf-xet==1.5.2 +nvidia-nccl-cu12==2.21.5 +MarkupSafe==3.0.3 +regex==2026.7.19 +importlib_metadata==9.0.0 +httpcore==1.0.9 +annotated-doc==0.0.5 +multidict==6.7.1 +aiohttp==3.14.3 +aiosignal==1.4.0 +xxhash==3.8.1 +aiohappyeyeballs==2.7.1 +mdurl==0.1.2 +cuda-toolkit==13.0.3.0 +networkx==3.6.1 +PyYAML==6.0.3 +nvidia-cufile==1.15.1.6 +typer==0.27.0 +torchaudio==2.6.0+cu124 +rich==15.0.0 +nvidia-cufft-cu12==11.2.1.3 +h11==0.16.0 +dill==0.4.1 +cuda-pathfinder==1.6.0 +filelock==3.29.0 +nvidia-nvtx-cu12==12.4.127 +httpx==0.28.1 +anyio==4.14.2 +numpy==2.4.4 +yarl==1.24.5 +click==8.4.2 +triton==3.2.0 +frozenlist==1.8.0 +zipp==4.1.0 +propcache==0.5.2 +tokenizers==0.22.2 +markdown-it-py==4.2.0 +nvidia-cuda-runtime==13.0.96 +cuda-bindings==13.3.1 +nvidia-cuda-cupti==13.0.85 +torch==2.6.0+cu124 +multiprocess==0.70.19 +pillow==12.2.0 +transformers==5.16.0.dev0 +wandb==0.28.1 +nvidia-curand==10.4.0.35 +sympy==1.13.1 +nvidia-cusparse==12.6.3.3 +nvidia-cuda-nvrtc==13.0.88 +typing-inspection==0.4.2 +nvidia-cusolver==12.0.4.66 +nvidia-cufft==12.0.0.61 +nvidia-cudnn-cu13==9.20.0.48 +nvidia-cublas==13.1.1.3 +pyarrow==25.0.0 +evaluate==0.4.6 +diffusers==0.39.0 +pydantic==2.13.4 +annotated-types==0.8.0 +protobuf==7.35.1 +sentry-sdk==2.66.1 +einops==0.8.2 +packaging==26.2 +nvidia-nvjitlink-cu12==12.4.127 +nvidia-curand-cu12==10.3.5.147 +nvidia-cusparselt-cu12==0.6.2 +nvidia-cusparse-cu12==12.3.1.170 +nvidia-cuda-runtime-cu12==12.4.127 +torchvision==0.21.0+cu124 +nvidia-cuda-nvrtc-cu12==12.4.127 +nvidia-cuda-cupti-cu12==12.4.127 +nvidia-cusolver-cu12==11.6.1.9 +nvidia-cublas-cu12==12.4.5.8 +nvidia-cudnn-cu12==9.1.0.70 +huggingface_hub==1.26.0 +datasets==5.0.1 +safetensors==0.8.0 +accelerate==1.14.0 +pydantic_core==2.46.4 +ninja==1.13.0 +autocommand==2.2.2 +backports.tarfile==1.2.0 +importlib_metadata==8.7.1 +jaraco.text==4.0.0 +jaraco.context==6.1.0 +jaraco.functools==4.4.0 +more-itertools==10.8.0 +packaging==26.0 +platformdirs==4.4.0 +tomli==2.4.0 +wheel==0.46.3 +zipp==3.23.0 diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/wandb-metadata.json b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/wandb-metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..17a13ae3a2e994ee8619d2ccc73faf833e3950c5 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/wandb-metadata.json @@ -0,0 +1,101 @@ +{ + "os": "Linux-5.15.0-126-generic-x86_64-with-glibc2.35", + "python": "CPython 3.11.15", + "startedAt": "2026-08-14T21:45:49.461932Z", + "args": [ + "--config", + "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml", + "--variants", + "glu-waleed10", + "glu-silu-waleed10", + "mlp-waleed10", + "mlp-silu-waleed10", + "glu-waleed", + "glu-situglu_low", + "glu-waleedglu_low" + ], + "program": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py", + "codePath": "sweep.py", + "codePathLocal": "sweep.py", + "git": { + "remote": "https://github.com/w-ahmad1a10/Activation.git", + "commit": "3fb47c6943d6927eafc39bd399a3d3a520bae74a" + }, + "email": "deepnevro@gmail.com", + "root": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation", + "host": "deeplens-k3s-node1", + "executable": "/mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python", + "cpu_count": 112, + "cpu_count_logical": 224, + "gpu": "NVIDIA H100 80GB HBM3", + "gpu_count": 8, + "disk": { + "/": { + "total": "1560765693952", + "used": "709081337856" + } + }, + "memory": { + "total": "2164089937920" + }, + "gpu_nvidia": [ + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-39c684a5-fde6-83d7-1663-0859795881ae" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-132944c4-b689-2b5f-89a4-d730401677ab" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bc6c3e3c-9b90-09ca-c034-774961847c54" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea" + } + ], + "cudaVersion": "12.4", + "writerId": "wtedta21p9zfavbt9r6hw8vts8caym1b" +} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..2b56e6eb75d873d308697450d7a62838f0f25566 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/files/wandb-summary.json @@ -0,0 +1 @@ +{"train_samples_per_second":654.524,"eval/steps_per_second":7.91,"train_steps_per_second":1.278,"_runtime":790,"total_flos":5.420315836416e+15,"_step":52,"train_loss":12.069926330566407,"train/grad_norm":2,"eval/runtime":9.4813,"_timestamp":1.7867447411310596e+09,"train/global_step":1000,"train_runtime":782.2476,"eval/samples_per_second":1004.817,"train/epoch":0.5392289026691831,"train/learning_rate":0.001,"_wandb":{"runtime":790},"train/loss":8.422817993164063,"eval/loss":2.101808547973633} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug-core.log b/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug-core.log new file mode 100644 index 0000000000000000000000000000000000000000..588056df10bc8cabd5e2682287953b3e2b130ffe --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug-core.log @@ -0,0 +1,62 @@ +{"time":"2026-08-14T21:31:53.539152008Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp650y3sc2/port-1229423.txt","pid":1229423,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false} +{"time":"2026-08-14T21:31:53.540676876Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1229423} +{"time":"2026-08-14T21:31:53.540659035Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T21:31:53.717724383Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"} +{"time":"2026-08-14T21:31:53.802092564Z","level":"INFO","msg":"handleInformInit: received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:54.062397411Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:59.411924652Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.396635188Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.929834053Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.931620539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:48.932491197Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:49.464388506Z","level":"INFO","msg":"handleInformInit: received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:49.733853364Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:55.329792921Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.158537572Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.914569559Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.916156477Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:01.917534874Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:02.524868682Z","level":"INFO","msg":"handleInformInit: received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:02.786199875Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:08.192858634Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:53.643379161Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.173579992Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.17490129Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.175479186Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.694802218Z","level":"INFO","msg":"handleInformInit: received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:11:54.956642682Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:12:00.441973218Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:55.662446346Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.296327526Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.297810539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.29856977Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.82560196Z","level":"INFO","msg":"handleInformInit: received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:24:57.085772231Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:25:02.571913681Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:28.612623637Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.208727987Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.210106262Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.210705504Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.839594174Z","level":"INFO","msg":"handleInformInit: received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:30.099444184Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:35.620666929Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:17.704425638Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.35448297Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.355720686Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.356578583Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.899799881Z","level":"INFO","msg":"handleInformInit: received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:19.159494071Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:24.671083701Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.214192173Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.783107435Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.784410941Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:51.78516667Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088192277Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088229648Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088239651Z","level":"INFO","msg":"connection: closing","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088243873Z","level":"INFO","msg":"server: is shutting down"} +{"time":"2026-08-14T23:05:52.0882852Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088299095Z","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088345917Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088452051Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T23:05:52.088520571Z","level":"INFO","msg":"server: all connections closed"} diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug-internal.log new file mode 100644 index 0000000000000000000000000000000000000000..b4f62cc834d6c956cd959a34c876ded11c9f6396 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug-internal.log @@ -0,0 +1,123 @@ +{"time":"2026-08-14T21:45:49.464511458Z","level":"INFO","msg":"wandb-core"} +{"time":"2026-08-14T21:45:49.464655404Z","level":"INFO","msg":"stream: starting","core version":"0.28.1"} +{"time":"2026-08-14T21:45:49.733697059Z","level":"INFO","msg":"stream: created new stream","id":"k24y90ws"} +{"time":"2026-08-14T21:45:49.733775349Z","level":"INFO","msg":"handler: started"} +{"time":"2026-08-14T21:45:49.733847443Z","level":"INFO","msg":"stream: started"} +{"time":"2026-08-14T21:45:49.733855982Z","level":"INFO","msg":"writer: started","stream_id":"k24y90ws"} +{"time":"2026-08-14T21:45:49.733908797Z","level":"INFO","msg":"sender: started"} +{"time":"2026-08-14T21:45:50.320642056Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:45:50.446758607Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:46:05.320807587Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":0,"events_lines":1,"console_offset":0,"console_lines":1,"uploaded_len":2} +{"time":"2026-08-14T21:46:05.495295275Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:46:20.321362864Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":0,"history_lines":1,"events_offset":1,"events_lines":2,"console_offset":0,"console_lines":2} +{"time":"2026-08-14T21:46:20.588211042Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:46:35.321170502Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":1,"history_lines":1,"events_offset":3,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:46:35.523653777Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:46:50.321411954Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":2,"history_lines":1,"events_offset":5,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:46:50.578357633Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:47:05.320973253Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":3,"history_lines":1,"events_offset":7,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:47:05.494256218Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:47:20.321280351Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":4,"history_lines":1,"events_offset":9,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:47:20.471515967Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:47:35.321022569Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":5,"history_lines":1,"events_offset":11,"events_lines":2,"console_offset":2,"console_lines":10} +{"time":"2026-08-14T21:47:35.496353897Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:47:50.320853074Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":6,"history_lines":1,"events_offset":13,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T21:47:50.483129663Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:48:05.32118415Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":7,"history_lines":1,"events_offset":15,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T21:48:05.503672796Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:48:20.321043028Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":8,"history_lines":1,"events_offset":17,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T21:48:20.530090466Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:48:35.321781731Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":9,"history_lines":1,"events_offset":19,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T21:48:35.508321767Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:48:50.321120237Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":10,"history_lines":1,"events_offset":21,"events_lines":2,"console_offset":12,"console_lines":10} +{"time":"2026-08-14T21:48:50.442521449Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:49:05.320862907Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":1,"events_offset":23,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T21:49:05.535886019Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:49:20.321534303Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":12,"history_lines":1,"events_offset":25,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T21:49:20.473500807Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:49:35.320974561Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":1,"events_offset":27,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T21:49:35.452315273Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:49:50.321583693Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":14,"history_lines":1,"events_offset":29,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T21:49:50.475924855Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:50:05.321095412Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":15,"history_lines":1,"events_offset":31,"events_lines":2,"console_offset":22,"console_lines":10} +{"time":"2026-08-14T21:50:05.510373585Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:50:20.321072924Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T21:50:20.441403714Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:50:35.320763733Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":17,"history_lines":1,"events_offset":35,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T21:50:35.466119892Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:50:50.321438704Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":18,"history_lines":1,"events_offset":37,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T21:50:50.456280209Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:51:05.320761985Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":1,"events_offset":39,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T21:51:05.477583946Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:51:20.321080069Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":20,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":32,"console_lines":10} +{"time":"2026-08-14T21:51:20.453021894Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:51:35.320796936Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":1,"events_offset":43,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T21:51:35.466525803Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:51:50.320955591Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":22,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T21:51:50.469319734Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:52:05.321309506Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":23,"history_lines":1,"events_offset":47,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T21:52:05.501402407Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:52:20.321650872Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":1,"events_offset":49,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T21:52:20.4534208Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:52:35.32088535Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":25,"history_lines":1,"events_offset":51,"events_lines":2,"console_offset":42,"console_lines":10} +{"time":"2026-08-14T21:52:35.461275831Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:52:50.321176023Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":1,"events_offset":53,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T21:52:50.466826609Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:53:05.321699021Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":27,"history_lines":1,"events_offset":55,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T21:53:05.486511079Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:53:20.321112129Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":28,"history_lines":1,"events_offset":57,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T21:53:20.457369802Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:53:35.320797688Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":1,"events_offset":59,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T21:53:35.452345981Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:53:50.321742882Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":30,"history_lines":1,"events_offset":61,"events_lines":2,"console_offset":52,"console_lines":10} +{"time":"2026-08-14T21:53:50.434938647Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:54:05.321429864Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":1,"events_offset":63,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T21:54:05.548284337Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:54:20.32114048Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":32,"history_lines":1,"events_offset":65,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T21:54:20.45358234Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:54:35.321129351Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":1,"events_offset":67,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T21:54:35.474897162Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:54:50.321706148Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":34,"history_lines":1,"events_offset":69,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T21:54:50.608442776Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:55:05.321443012Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":71,"events_lines":2,"console_offset":62,"console_lines":9} +{"time":"2026-08-14T21:55:05.477345408Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:55:20.320811721Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":1,"events_offset":73,"events_lines":2,"console_offset":70,"console_lines":2} +{"time":"2026-08-14T21:55:20.499424268Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:55:35.321629144Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":36,"history_lines":1,"events_offset":75,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T21:55:35.445775297Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:55:50.321311027Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":1,"events_offset":77,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T21:55:50.431282685Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:56:05.320926776Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":38,"history_lines":1,"events_offset":79,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T21:56:05.517090556Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:56:20.321730866Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":81,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T21:56:20.570971062Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:56:35.321155047Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":1,"events_offset":83,"events_lines":2,"console_offset":72,"console_lines":10} +{"time":"2026-08-14T21:56:35.437882867Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:56:50.32102971Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":41,"history_lines":1,"events_offset":85,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T21:56:50.524971529Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:57:05.321204634Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":1,"events_offset":87,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T21:57:05.531275159Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:57:20.32172013Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":43,"history_lines":1,"events_offset":89,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T21:57:20.528511373Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:57:35.321170348Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":1,"events_offset":91,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T21:57:35.458549034Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:57:50.320862066Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":45,"history_lines":1,"events_offset":93,"events_lines":2,"console_offset":82,"console_lines":10} +{"time":"2026-08-14T21:57:50.444730503Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:58:05.321650213Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":95,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T21:58:05.534409376Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:58:20.321042008Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":1,"events_offset":97,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T21:58:20.442037868Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:58:35.32097911Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":48,"history_lines":1,"events_offset":99,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T21:58:35.451754594Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:58:50.321128605Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":1,"events_offset":101,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T21:58:50.456537571Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:59:01.564627783Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T21:59:01.564836113Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":50,"history_lines":3,"events_offset":103,"events_lines":1,"console_offset":90,"console_lines":1,"uploaded_len":3} +{"time":"2026-08-14T21:59:01.777581888Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:59:01.777738913Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":92,"console_lines":18,"complete":true,"exit_code":0} +{"time":"2026-08-14T21:59:01.912393768Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:59:01.913517802Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T21:59:01.916175371Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T21:59:01.916187489Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T21:59:01.916300031Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T21:59:01.916303725Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug.log b/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..3c64bbfba0035f04ddbd2e898ae6e789b7f61a19 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug.log @@ -0,0 +1,24 @@ +2026-08-14 21:45:49,463 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():729] Logging user logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug.log +2026-08-14 21:45:49,463 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():730] Logging internal logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_214549-k24y90ws/logs/debug-internal.log +2026-08-14 21:45:49,463 INFO MainThread:1229423 [wandb_init.py:init():772] calling init triggers +2026-08-14 21:45:49,463 INFO MainThread:1229423 [wandb_init.py:init():777] wandb.init called with sweep_config: {} +config: {'_wandb': {}} +2026-08-14 21:45:49,463 INFO MainThread:1229423 [wandb_init.py:init():820] starting backend +2026-08-14 21:45:49,463 INFO MainThread:1229423 [wandb_init.py:init():835] sending inform_init request +2026-08-14 21:45:49,734 INFO MainThread:1229423 [wandb_init.py:init():840] backend started and connected +2026-08-14 21:45:49,735 INFO MainThread:1229423 [wandb_init.py:init():910] updated telemetry +2026-08-14 21:45:49,742 INFO MainThread:1229423 [wandb_init.py:init():933] communicating run to backend with 90.0 second timeout +2026-08-14 21:45:50,240 INFO MainThread:1229423 [wandb_init.py:init():978] starting run threads in backend +2026-08-14 21:45:50,315 INFO MainThread:1229423 [wandb_run.py:_console_start():2621] atexit reg +2026-08-14 21:45:50,315 INFO MainThread:1229423 [wandb_run.py:_redirect():2471] redirect: wrap_raw +2026-08-14 21:45:50,315 INFO MainThread:1229423 [wandb_run.py:_redirect():2540] Wrapping output streams. +2026-08-14 21:45:50,316 INFO MainThread:1229423 [wandb_run.py:_redirect():2563] Redirects installed. +2026-08-14 21:45:50,316 INFO MainThread:1229423 [wandb_init.py:init():1016] run started, returning control to user process +2026-08-14 21:45:50,317 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 21, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'glu', 'activation': 'silu-waleed10', 'waleed_beta': 10.0, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/glu-silu-waleed10-21L_run', 'per_device_train_batch_size': 128, 'num_train_epochs': 1, 'max_steps': 1000, 'learning_rate': 0.001, 'lr_scheduler_type': 'constant', 'lr_scheduler_kwargs': None, 'warmup_steps': 0, 'optim': 'adamw_torch', 'optim_args': None, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 4, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-glu-silu-waleed10-21L-4.0M-20260814-214549', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 60000, 'eval_delay': 0, 'per_device_eval_batch_size': 128, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 100, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/A-glu-silu-waleed10-21L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} +2026-08-14 21:45:50,319 INFO MainThread:1229423 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 3970432 - > +2026-08-14 21:45:50,319 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 3970432 None +2026-08-14 21:59:01,157 INFO MainThread:1229423 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research1/k24y90ws +2026-08-14 21:59:01,158 INFO MainThread:1229423 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 21:59:01,158 INFO MainThread:1229423 [wandb_run.py:_restore():2570] restore +2026-08-14 21:59:01,158 INFO MainThread:1229423 [wandb_run.py:_restore():2576] restore done +2026-08-14 21:59:01,915 INFO MainThread:1229423 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_214549-k24y90ws/run-k24y90ws.wandb b/zain/Activation/wandb/run-20260814_214549-k24y90ws/run-k24y90ws.wandb new file mode 100644 index 0000000000000000000000000000000000000000..662e271e5c2aa06228e269c43400285f29fbf90e --- /dev/null +++ b/zain/Activation/wandb/run-20260814_214549-k24y90ws/run-k24y90ws.wandb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:848becffaac977b2c893e1398902ee9ad2c0cf5857f01232497e0c98c9682d39 +size 563608 diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/config.yaml b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..a3293a2f3a68d00bf04995ee68560a7b1410d0f9 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/config.yaml @@ -0,0 +1,439 @@ +_name_or_path: + value: "" +_wandb: + value: + cli_version: 0.28.1 + e: + uyw2t8oqcydlnz3g1p7dvon8gcx3um7p: + args: + - --config + - /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml + - --variants + - glu-waleed10 + - glu-silu-waleed10 + - mlp-waleed10 + - mlp-silu-waleed10 + - glu-waleed + - glu-situglu_low + - glu-waleedglu_low + codePath: sweep.py + codePathLocal: sweep.py + cpu_count: 112 + cpu_count_logical: 224 + cudaVersion: "12.4" + disk: + /: + total: "1560765693952" + used: "709083283456" + email: deepnevro@gmail.com + executable: /mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python + git: + commit: 3fb47c6943d6927eafc39bd399a3d3a520bae74a + remote: https://github.com/w-ahmad1a10/Activation.git + gpu: NVIDIA H100 80GB HBM3 + gpu_count: 8 + gpu_nvidia: + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-39c684a5-fde6-83d7-1663-0859795881ae + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-132944c4-b689-2b5f-89a4-d730401677ab + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bc6c3e3c-9b90-09ca-c034-774961847c54 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea + host: deeplens-k3s-node1 + memory: + total: "2164089937920" + os: Linux-5.15.0-126-generic-x86_64-with-glibc2.35 + program: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py + python: CPython 3.11.15 + root: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation + startedAt: "2026-08-14T21:59:02.522762Z" + writerId: uyw2t8oqcydlnz3g1p7dvon8gcx3um7p + m: + - "1": train/global_step + "6": + - 3 + "7": [] + - "2": '*' + "5": 1 + "6": + - 1 + "7": [] + python_version: 3.11.15 + t: + "1": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "2": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "3": + - 2 + - 7 + - 13 + - 19 + - 62 + - 66 + "4": 3.11.15 + "5": 0.28.1 + "6": 5.16.0.dev0 + "9": + "1": transformers_trainer + "12": 0.28.1 + "13": linux-x86_64 +accelerator_config: + value: + dispatch_batches: null + even_batches: true + gradient_accumulation_kwargs: null + non_blocking: false + split_batches: false + use_seedable_sampler: true +activation: + value: waleed10 +adam_beta1: + value: 0.9 +adam_beta2: + value: 0.999 +adam_epsilon: + value: 1e-08 +architectures: + value: null +attention_bias: + value: false +attention_dropout: + value: 0 +auto_find_batch_size: + value: false +average_tokens_across_devices: + value: true +batch_eval_metrics: + value: false +bf16: + value: true +bf16_full_eval: + value: false +bos_token_id: + value: 1 +chunk_size_feed_forward: + value: 0 +data_seed: + value: 42 +dataloader_drop_last: + value: false +dataloader_in_order: + value: true +dataloader_multiprocessing_context: + value: null +dataloader_num_workers: + value: 0 +dataloader_persistent_workers: + value: false +dataloader_pin_memory: + value: true +dataloader_prefetch_factor: + value: null +ddp_backend: + value: null +ddp_broadcast_buffers: + value: null +ddp_bucket_cap_mb: + value: null +ddp_find_unused_parameters: + value: null +ddp_static_graph: + value: null +ddp_timeout: + value: 1800 +debug: + value: [] +deepspeed: + value: null +disable_tqdm: + value: false +do_eval: + value: true +do_predict: + value: false +do_train: + value: false +dtype: + value: null +enable_jit_checkpoint: + value: false +eos_token_id: + value: 2 +eval_accumulation_steps: + value: null +eval_delay: + value: 0 +eval_do_concat_batches: + value: true +eval_on_start: + value: false +eval_steps: + value: 60000 +eval_strategy: + value: steps +eval_use_gather_object: + value: false +fp16: + value: false +fp16_full_eval: + value: false +fsdp: + value: null +fsdp_config: + value: null +full_determinism: + value: false +gradient_accumulation_steps: + value: 4 +gradient_checkpointing: + value: false +gradient_checkpointing_kwargs: + value: null +greater_is_better: + value: null +head_dim: + value: 32 +hidden_act: + value: silu +hidden_size: + value: 128 +hub_always_push: + value: false +hub_model_id: + value: w-ahmad/A-mlp-waleed10-21L +hub_private_repo: + value: null +hub_revision: + value: null +hub_strategy: + value: every_save +hub_token: + value: +id2label: + value: + "0": LABEL_0 + "1": LABEL_1 +ignore_data_skip: + value: false +include_for_metrics: + value: [] +include_num_input_tokens_seen: + value: "no" +initializer_range: + value: 0.02 +intermediate_size: + value: 256 +is_encoder_decoder: + value: false +label_names: + value: null +label_smoothing_factor: + value: 0 +label2id: + value: + LABEL_0: 0 + LABEL_1: 1 +learning_rate: + value: 0.001 +length_column_name: + value: length +liger_kernel_config: + value: null +load_best_model_at_end: + value: false +local_rank: + value: -1 +log_level: + value: passive +log_level_replica: + value: warning +log_on_each_node: + value: true +logging_first_step: + value: false +logging_nan_inf_filter: + value: true +logging_steps: + value: 20 +logging_strategy: + value: steps +lr_scheduler_kwargs: + value: null +lr_scheduler_type: + value: constant +max_grad_norm: + value: 1 +max_position_embeddings: + value: 512 +max_steps: + value: 1000 +metric_for_best_model: + value: null +mlp_bias: + value: false +mlp_type: + value: mlp +model/num_parameters: + value: 3970432 +model_type: + value: tiny_llama +neftune_noise_alpha: + value: null +num_attention_heads: + value: 4 +num_hidden_layers: + value: 21 +num_key_value_heads: + value: 4 +num_train_epochs: + value: 1 +optim: + value: adamw_torch +optim_args: + value: null +optim_target_modules: + value: null +output_attentions: + value: false +output_dir: + value: out/mlp-waleed10-21L_run +output_hidden_states: + value: false +pad_token_id: + value: 0 +parallelism_config: + value: null +per_device_eval_batch_size: + value: 128 +per_device_train_batch_size: + value: 128 +prediction_loss_only: + value: false +pretraining_tp: + value: 1 +problem_type: + value: null +project: + value: huggingface +push_to_hub: + value: false +remove_unused_columns: + value: false +report_to: + value: + - wandb +restore_callback_states_from_checkpoint: + value: false +resume_from_checkpoint: + value: null +return_dict: + value: true +rms_norm_eps: + value: 1e-06 +rope_parameters: + value: + rope_theta: 10000 + rope_type: default +run_name: + value: LM-mlp-waleed10-21L-4.0M-20260814-215902 +save_on_each_node: + value: false +save_only_model: + value: false +save_steps: + value: 100 +save_strategy: + value: steps +save_total_limit: + value: null +seed: + value: 42 +skip_memory_metrics: + value: true +tf32: + value: null +tie_word_embeddings: + value: true +tokenizer_name: + value: w-ahmad/tiny-stories-tokenizer +torch_compile: + value: false +torch_compile_backend: + value: null +torch_compile_mode: + value: null +torch_empty_cache_steps: + value: null +trackio_bucket_id: + value: null +trackio_space_id: + value: null +trackio_static_space_id: + value: null +train_sampling_strategy: + value: random +transformers_version: + value: 5.16.0.dev0 +use_cache: + value: false +use_cpu: + value: false +use_liger_kernel: + value: false +vocab_size: + value: 4096 +waleed_beta: + value: 10 +warmup_steps: + value: 0 +weight_decay: + value: 0.01 diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/output.log b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/output.log new file mode 100644 index 0000000000000000000000000000000000000000..42be723411aee4b04c564e351a0dd4400681a135 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/output.log @@ -0,0 +1,110 @@ + 10%|█ | 100/1000 [01:15<11:13, 1.34it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '29.1', 'grad_norm': '4.219', 'learning_rate': '0.001', 'epoch': '0.01078', 'train/total_time_seconds': '7.459', 'train/time_per_step_avg': '0.3729', 'train/epoch_time_elapsed': '15.17', 'train/estimated_remaining_minutes': '6.091'} +{'loss': '24.02', 'grad_norm': '13.25', 'learning_rate': '0.001', 'epoch': '0.02157', 'train/total_time_seconds': '14.82', 'train/time_per_step_avg': '0.3705', 'train/epoch_time_elapsed': '30.08', 'train/estimated_remaining_minutes': '5.928'} +{'loss': '22.07', 'grad_norm': '2.156', 'learning_rate': '0.001', 'epoch': '0.03235', 'train/total_time_seconds': '22.18', 'train/time_per_step_avg': '0.3696', 'train/epoch_time_elapsed': '45.2', 'train/estimated_remaining_minutes': '5.79'} +{'loss': '20.73', 'grad_norm': '2.984', 'learning_rate': '0.001', 'epoch': '0.04314', 'train/total_time_seconds': '29.57', 'train/time_per_step_avg': '0.3696', 'train/epoch_time_elapsed': '60.31', 'train/estimated_remaining_minutes': '5.668'} +{'loss': '19.62', 'grad_norm': '2.562', 'learning_rate': '0.001', 'epoch': '0.05392', 'train/total_time_seconds': '36.97', 'train/time_per_step_avg': '0.3697', 'train/epoch_time_elapsed': '75.41', 'train/estimated_remaining_minutes': '5.545'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 43.34it/s] + 20%|██ | 200/1000 [02:30<09:53, 1.35it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '18.49', 'grad_norm': '3.5', 'learning_rate': '0.001', 'epoch': '0.06471', 'train/total_time_seconds': '44.36', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '90.44', 'train/estimated_remaining_minutes': '5.422'} +{'loss': '17.53', 'grad_norm': '2.328', 'learning_rate': '0.001', 'epoch': '0.07549', 'train/total_time_seconds': '51.76', 'train/time_per_step_avg': '0.3694', 'train/epoch_time_elapsed': '105.6', 'train/estimated_remaining_minutes': '5.299'} +{'loss': '16.74', 'grad_norm': '4', 'learning_rate': '0.001', 'epoch': '0.08628', 'train/total_time_seconds': '59.16', 'train/time_per_step_avg': '0.3698', 'train/epoch_time_elapsed': '120.5', 'train/estimated_remaining_minutes': '5.176'} +{'loss': '16.19', 'grad_norm': '2.656', 'learning_rate': '0.001', 'epoch': '0.09706', 'train/total_time_seconds': '66.55', 'train/time_per_step_avg': '0.3698', 'train/epoch_time_elapsed': '135.5', 'train/estimated_remaining_minutes': '5.053'} +{'loss': '15.68', 'grad_norm': '2.047', 'learning_rate': '0.001', 'epoch': '0.1078', 'train/total_time_seconds': '73.92', 'train/time_per_step_avg': '0.3696', 'train/epoch_time_elapsed': '150.5', 'train/estimated_remaining_minutes': '4.928'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 78.32it/s] + 30%|███ | 300/1000 [03:45<08:42, 1.34it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '15.22', 'grad_norm': '3.906', 'learning_rate': '0.001', 'epoch': '0.1186', 'train/total_time_seconds': '81.28', 'train/time_per_step_avg': '0.3691', 'train/epoch_time_elapsed': '165.5', 'train/estimated_remaining_minutes': '4.803'} +{'loss': '14.92', 'grad_norm': '2.75', 'learning_rate': '0.001', 'epoch': '0.1294', 'train/total_time_seconds': '88.63', 'train/time_per_step_avg': '0.3687', 'train/epoch_time_elapsed': '180.4', 'train/estimated_remaining_minutes': '4.678'} +{'loss': '14.54', 'grad_norm': '1.633', 'learning_rate': '0.001', 'epoch': '0.1402', 'train/total_time_seconds': '96', 'train/time_per_step_avg': '0.3684', 'train/epoch_time_elapsed': '195.6', 'train/estimated_remaining_minutes': '4.554'} +{'loss': '14.27', 'grad_norm': '2.688', 'learning_rate': '0.001', 'epoch': '0.151', 'train/total_time_seconds': '103.4', 'train/time_per_step_avg': '0.3684', 'train/epoch_time_elapsed': '210.5', 'train/estimated_remaining_minutes': '4.431'} +{'loss': '14.02', 'grad_norm': '3.156', 'learning_rate': '0.001', 'epoch': '0.1618', 'train/total_time_seconds': '110.8', 'train/time_per_step_avg': '0.3687', 'train/epoch_time_elapsed': '225.5', 'train/estimated_remaining_minutes': '4.308'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 80.87it/s] + 40%|████ | 400/1000 [05:00<07:42, 1.30it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '13.78', 'grad_norm': '3.938', 'learning_rate': '0.001', 'epoch': '0.1726', 'train/total_time_seconds': '118.2', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '240.5', 'train/estimated_remaining_minutes': '4.186'} +{'loss': '13.6', 'grad_norm': '2.906', 'learning_rate': '0.001', 'epoch': '0.1833', 'train/total_time_seconds': '125.6', 'train/time_per_step_avg': '0.3695', 'train/epoch_time_elapsed': '255.5', 'train/estimated_remaining_minutes': '4.063'} +{'loss': '13.37', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.1941', 'train/total_time_seconds': '133', 'train/time_per_step_avg': '0.3697', 'train/epoch_time_elapsed': '270.5', 'train/estimated_remaining_minutes': '3.94'} +{'loss': '13.15', 'grad_norm': '3.797', 'learning_rate': '0.001', 'epoch': '0.2049', 'train/total_time_seconds': '140.3', 'train/time_per_step_avg': '0.3693', 'train/epoch_time_elapsed': '285.5', 'train/estimated_remaining_minutes': '3.816'} +{'loss': '13.03', 'grad_norm': '1.859', 'learning_rate': '0.001', 'epoch': '0.2157', 'train/total_time_seconds': '147.7', 'train/time_per_step_avg': '0.3689', 'train/epoch_time_elapsed': '300.6', 'train/estimated_remaining_minutes': '3.692'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 83.27it/s] + 50%|█████ | 500/1000 [06:15<06:15, 1.33it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '12.84', 'grad_norm': '2.766', 'learning_rate': '0.001', 'epoch': '0.2265', 'train/total_time_seconds': '155', 'train/time_per_step_avg': '0.3687', 'train/epoch_time_elapsed': '315.7', 'train/estimated_remaining_minutes': '3.569'} +{'loss': '12.73', 'grad_norm': '4.531', 'learning_rate': '0.001', 'epoch': '0.2373', 'train/total_time_seconds': '162.4', 'train/time_per_step_avg': '0.3687', 'train/epoch_time_elapsed': '330.6', 'train/estimated_remaining_minutes': '3.446'} +{'loss': '12.55', 'grad_norm': '3.031', 'learning_rate': '0.001', 'epoch': '0.248', 'train/total_time_seconds': '169.8', 'train/time_per_step_avg': '0.3688', 'train/epoch_time_elapsed': '345.7', 'train/estimated_remaining_minutes': '3.323'} +{'loss': '12.41', 'grad_norm': '3.359', 'learning_rate': '0.001', 'epoch': '0.2588', 'train/total_time_seconds': '177.3', 'train/time_per_step_avg': '0.3693', 'train/epoch_time_elapsed': '360.6', 'train/estimated_remaining_minutes': '3.2'} +{'loss': '12.28', 'grad_norm': '3.609', 'learning_rate': '0.001', 'epoch': '0.2696', 'train/total_time_seconds': '184.7', 'train/time_per_step_avg': '0.3698', 'train/epoch_time_elapsed': '375.6', 'train/estimated_remaining_minutes': '3.078'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 66.41it/s] + 60%|██████ | 600/1000 [07:30<04:57, 1.34it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '12.15', 'grad_norm': '2.406', 'learning_rate': '0.001', 'epoch': '0.2804', 'train/total_time_seconds': '192', 'train/time_per_step_avg': '0.37', 'train/epoch_time_elapsed': '390.7', 'train/estimated_remaining_minutes': '2.955'} +{'loss': '12.05', 'grad_norm': '3', 'learning_rate': '0.001', 'epoch': '0.2912', 'train/total_time_seconds': '199.4', 'train/time_per_step_avg': '0.37', 'train/epoch_time_elapsed': '405.9', 'train/estimated_remaining_minutes': '2.832'} +{'loss': '11.96', 'grad_norm': '2.547', 'learning_rate': '0.001', 'epoch': '0.302', 'train/total_time_seconds': '206.8', 'train/time_per_step_avg': '0.3698', 'train/epoch_time_elapsed': '421', 'train/estimated_remaining_minutes': '2.709'} +{'loss': '11.82', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.3128', 'train/total_time_seconds': '214.2', 'train/time_per_step_avg': '0.3697', 'train/epoch_time_elapsed': '436', 'train/estimated_remaining_minutes': '2.585'} +{'loss': '11.72', 'grad_norm': '3.219', 'learning_rate': '0.001', 'epoch': '0.3235', 'train/total_time_seconds': '221.6', 'train/time_per_step_avg': '0.3694', 'train/epoch_time_elapsed': '450.9', 'train/estimated_remaining_minutes': '2.462'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 86.34it/s] + 70%|███████ | 700/1000 [08:45<03:42, 1.35it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '11.65', 'grad_norm': '2.438', 'learning_rate': '0.001', 'epoch': '0.3343', 'train/total_time_seconds': '229', 'train/time_per_step_avg': '0.3692', 'train/epoch_time_elapsed': '465.9', 'train/estimated_remaining_minutes': '2.339'} +{'loss': '11.56', 'grad_norm': '2.469', 'learning_rate': '0.001', 'epoch': '0.3451', 'train/total_time_seconds': '236.4', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '480.9', 'train/estimated_remaining_minutes': '2.216'} +{'loss': '11.47', 'grad_norm': '2.5', 'learning_rate': '0.001', 'epoch': '0.3559', 'train/total_time_seconds': '243.7', 'train/time_per_step_avg': '0.3688', 'train/epoch_time_elapsed': '495.9', 'train/estimated_remaining_minutes': '2.093'} +{'loss': '11.37', 'grad_norm': '2.906', 'learning_rate': '0.001', 'epoch': '0.3667', 'train/total_time_seconds': '251.1', 'train/time_per_step_avg': '0.3689', 'train/epoch_time_elapsed': '511.1', 'train/estimated_remaining_minutes': '1.969'} +{'loss': '11.29', 'grad_norm': '3.141', 'learning_rate': '0.001', 'epoch': '0.3775', 'train/total_time_seconds': '258.5', 'train/time_per_step_avg': '0.3689', 'train/epoch_time_elapsed': '526', 'train/estimated_remaining_minutes': '1.846'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 86.72it/s] + 80%|████████ | 800/1000 [10:01<02:29, 1.34it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '11.2', 'grad_norm': '2.969', 'learning_rate': '0.001', 'epoch': '0.3882', 'train/total_time_seconds': '265.9', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '541.2', 'train/estimated_remaining_minutes': '1.723'} +{'loss': '11.14', 'grad_norm': '2.734', 'learning_rate': '0.001', 'epoch': '0.399', 'train/total_time_seconds': '273.3', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '556.1', 'train/estimated_remaining_minutes': '1.6'} +{'loss': '11.09', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.4098', 'train/total_time_seconds': '280.6', 'train/time_per_step_avg': '0.3692', 'train/epoch_time_elapsed': '571.3', 'train/estimated_remaining_minutes': '1.477'} +{'loss': '11.03', 'grad_norm': '2.797', 'learning_rate': '0.001', 'epoch': '0.4206', 'train/total_time_seconds': '288', 'train/time_per_step_avg': '0.3692', 'train/epoch_time_elapsed': '586.2', 'train/estimated_remaining_minutes': '1.354'} +{'loss': '10.93', 'grad_norm': '3.359', 'learning_rate': '0.001', 'epoch': '0.4314', 'train/total_time_seconds': '295.4', 'train/time_per_step_avg': '0.3693', 'train/epoch_time_elapsed': '601.1', 'train/estimated_remaining_minutes': '1.231'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 83.92it/s] + 90%|█████████ | 900/1000 [11:16<01:14, 1.33it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.88', 'grad_norm': '2.469', 'learning_rate': '0.001', 'epoch': '0.4422', 'train/total_time_seconds': '302.8', 'train/time_per_step_avg': '0.3692', 'train/epoch_time_elapsed': '616.1', 'train/estimated_remaining_minutes': '1.108'} +{'loss': '10.83', 'grad_norm': '2.797', 'learning_rate': '0.001', 'epoch': '0.453', 'train/total_time_seconds': '310.2', 'train/time_per_step_avg': '0.3691', 'train/epoch_time_elapsed': '631.3', 'train/estimated_remaining_minutes': '0.9847'} +{'loss': '10.75', 'grad_norm': '2.469', 'learning_rate': '0.001', 'epoch': '0.4637', 'train/total_time_seconds': '317.5', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '646.3', 'train/estimated_remaining_minutes': '0.8615'} +{'loss': '10.71', 'grad_norm': '2.844', 'learning_rate': '0.001', 'epoch': '0.4745', 'train/total_time_seconds': '324.9', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '661.4', 'train/estimated_remaining_minutes': '0.7385'} +{'loss': '10.66', 'grad_norm': '2.75', 'learning_rate': '0.001', 'epoch': '0.4853', 'train/total_time_seconds': '332.3', 'train/time_per_step_avg': '0.3687', 'train/epoch_time_elapsed': '676.4', 'train/estimated_remaining_minutes': '0.6153'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 85.12it/s] +100%|██████████| 1000/1000 [12:41<00:00, 1.31i[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.58', 'grad_norm': '2.422', 'learning_rate': '0.001', 'epoch': '0.4961', 'train/total_time_seconds': '339.7', 'train/time_per_step_avg': '0.3687', 'train/epoch_time_elapsed': '691.7', 'train/estimated_remaining_minutes': '0.4923'} +{'loss': '10.56', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.5069', 'train/total_time_seconds': '347.1', 'train/time_per_step_avg': '0.369', 'train/epoch_time_elapsed': '706.7', 'train/estimated_remaining_minutes': '0.3692'} +{'loss': '10.53', 'grad_norm': '2.328', 'learning_rate': '0.001', 'epoch': '0.5177', 'train/total_time_seconds': '354.5', 'train/time_per_step_avg': '0.3697', 'train/epoch_time_elapsed': '721.8', 'train/estimated_remaining_minutes': '0.2462'} +{'loss': '10.48', 'grad_norm': '3.453', 'learning_rate': '0.001', 'epoch': '0.5284', 'train/total_time_seconds': '361.9', 'train/time_per_step_avg': '0.3697', 'train/epoch_time_elapsed': '736.7', 'train/estimated_remaining_minutes': '0.1231'} +{'loss': '10.44', 'grad_norm': '2.375', 'learning_rate': '0.001', 'epoch': '0.5392', 'train/total_time_seconds': '369.3', 'train/time_per_step_avg': '0.37', 'train/epoch_time_elapsed': '751.8', 'train/estimated_remaining_minutes': '0'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes +{'eval_loss': '2.601', 'eval_runtime': '9.308', 'eval_samples_per_second': '1024', 'eval_steps_per_second': '8.058', 'epoch': '0.5392', 'train/total_time_seconds': '369.3', 'train/time_per_step_avg': '0.37', 'train/epoch_time_elapsed': '761.1', 'train/estimated_remaining_minutes': '0'} + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 72.91it/s] +100%|██████████| 1000/1000 [12:41<00:00, 1.31it/s], ?it/s] +{'train_runtime': '761.9', 'train_samples_per_second': '672', 'train_steps_per_second': '1.312', 'train_loss': '13.63', 'epoch': '0.5392', 'train/total_time_seconds': '369.3', 'train/time_per_step_avg': '0.37', 'train/epoch_time_elapsed': '761.2', 'train/estimated_remaining_minutes': '0'} +100%|██████████| 75/75 [00:09<00:00, 8.30it/s] +[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 88.27it/s] +>>> FINISHED mlp-waleed10-21L successfully diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/requirements.txt b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..f36b19895172d9637438a445ceacb71d09dc5d02 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/requirements.txt @@ -0,0 +1,149 @@ +asttokens==3.0.1 +comm==0.2.3 +debugpy==1.8.21 +decorator==5.3.1 +executing==2.2.1 +nest-asyncio==1.6.0 +parso==0.8.7 +platformdirs==4.11.0 +psutil==7.2.2 +ptyprocess==0.7.0 +pure_eval==0.2.3 +Pygments==2.20.0 +pyzmq==27.1.0 +setuptools==83.0.0 +six==1.17.0 +tornado==6.5.7 +traitlets==5.15.0 +fsspec==2026.4.0 +wcwidth==0.8.2 +ipython_pygments_lexers==1.1.1 +jedi==0.20.0 +jupyter_core==5.9.1 +matplotlib-inline==0.2.2 +pexpect==4.9.0 +prompt_toolkit==3.0.53 +python-dateutil==2.9.0.post0 +stack_data==0.6.3 +wheel==0.47.0 +jupyter_client==8.9.1 +pip==26.1.2 +ipython==9.15.0 +ipykernel==7.2.0 +threadpoolctl==3.6.0 +pyparsing==3.3.2 +typing_extensions==4.15.0 +Jinja2==3.1.6 +narwhals==2.24.0 +kiwisolver==1.5.0 +joblib==1.5.3 +fonttools==4.63.0 +cycler==0.12.1 +scipy==1.17.1 +pandas==3.0.5 +contourpy==1.3.3 +scikit-learn==1.9.0 +matplotlib==3.11.1 +urllib3==2.7.0 +tqdm==4.70.0 +idna==3.18 +charset-normalizer==3.4.9 +certifi==2026.7.22 +requests==2.34.2 +seaborn==0.13.2 +uv==0.12.0 +shellingham==1.5.4 +mpmath==1.3.0 +attrs==26.1.0 +hf-xet==1.5.2 +nvidia-nccl-cu12==2.21.5 +MarkupSafe==3.0.3 +regex==2026.7.19 +importlib_metadata==9.0.0 +httpcore==1.0.9 +annotated-doc==0.0.5 +multidict==6.7.1 +aiohttp==3.14.3 +aiosignal==1.4.0 +xxhash==3.8.1 +aiohappyeyeballs==2.7.1 +mdurl==0.1.2 +cuda-toolkit==13.0.3.0 +networkx==3.6.1 +PyYAML==6.0.3 +nvidia-cufile==1.15.1.6 +typer==0.27.0 +torchaudio==2.6.0+cu124 +rich==15.0.0 +nvidia-cufft-cu12==11.2.1.3 +h11==0.16.0 +dill==0.4.1 +cuda-pathfinder==1.6.0 +filelock==3.29.0 +nvidia-nvtx-cu12==12.4.127 +httpx==0.28.1 +anyio==4.14.2 +numpy==2.4.4 +yarl==1.24.5 +click==8.4.2 +triton==3.2.0 +frozenlist==1.8.0 +zipp==4.1.0 +propcache==0.5.2 +tokenizers==0.22.2 +markdown-it-py==4.2.0 +nvidia-cuda-runtime==13.0.96 +cuda-bindings==13.3.1 +nvidia-cuda-cupti==13.0.85 +torch==2.6.0+cu124 +multiprocess==0.70.19 +pillow==12.2.0 +transformers==5.16.0.dev0 +wandb==0.28.1 +nvidia-curand==10.4.0.35 +sympy==1.13.1 +nvidia-cusparse==12.6.3.3 +nvidia-cuda-nvrtc==13.0.88 +typing-inspection==0.4.2 +nvidia-cusolver==12.0.4.66 +nvidia-cufft==12.0.0.61 +nvidia-cudnn-cu13==9.20.0.48 +nvidia-cublas==13.1.1.3 +pyarrow==25.0.0 +evaluate==0.4.6 +diffusers==0.39.0 +pydantic==2.13.4 +annotated-types==0.8.0 +protobuf==7.35.1 +sentry-sdk==2.66.1 +einops==0.8.2 +packaging==26.2 +nvidia-nvjitlink-cu12==12.4.127 +nvidia-curand-cu12==10.3.5.147 +nvidia-cusparselt-cu12==0.6.2 +nvidia-cusparse-cu12==12.3.1.170 +nvidia-cuda-runtime-cu12==12.4.127 +torchvision==0.21.0+cu124 +nvidia-cuda-nvrtc-cu12==12.4.127 +nvidia-cuda-cupti-cu12==12.4.127 +nvidia-cusolver-cu12==11.6.1.9 +nvidia-cublas-cu12==12.4.5.8 +nvidia-cudnn-cu12==9.1.0.70 +huggingface_hub==1.26.0 +datasets==5.0.1 +safetensors==0.8.0 +accelerate==1.14.0 +pydantic_core==2.46.4 +ninja==1.13.0 +autocommand==2.2.2 +backports.tarfile==1.2.0 +importlib_metadata==8.7.1 +jaraco.text==4.0.0 +jaraco.context==6.1.0 +jaraco.functools==4.4.0 +more-itertools==10.8.0 +packaging==26.0 +platformdirs==4.4.0 +tomli==2.4.0 +wheel==0.46.3 +zipp==3.23.0 diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/wandb-metadata.json b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/wandb-metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6795b7a589062d94402f05a1be7712dcf332ba5f --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/wandb-metadata.json @@ -0,0 +1,101 @@ +{ + "os": "Linux-5.15.0-126-generic-x86_64-with-glibc2.35", + "python": "CPython 3.11.15", + "startedAt": "2026-08-14T21:59:02.522762Z", + "args": [ + "--config", + "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml", + "--variants", + "glu-waleed10", + "glu-silu-waleed10", + "mlp-waleed10", + "mlp-silu-waleed10", + "glu-waleed", + "glu-situglu_low", + "glu-waleedglu_low" + ], + "program": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py", + "codePath": "sweep.py", + "codePathLocal": "sweep.py", + "git": { + "remote": "https://github.com/w-ahmad1a10/Activation.git", + "commit": "3fb47c6943d6927eafc39bd399a3d3a520bae74a" + }, + "email": "deepnevro@gmail.com", + "root": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation", + "host": "deeplens-k3s-node1", + "executable": "/mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python", + "cpu_count": 112, + "cpu_count_logical": 224, + "gpu": "NVIDIA H100 80GB HBM3", + "gpu_count": 8, + "disk": { + "/": { + "total": "1560765693952", + "used": "709083283456" + } + }, + "memory": { + "total": "2164089937920" + }, + "gpu_nvidia": [ + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-39c684a5-fde6-83d7-1663-0859795881ae" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-132944c4-b689-2b5f-89a4-d730401677ab" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bc6c3e3c-9b90-09ca-c034-774961847c54" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea" + } + ], + "cudaVersion": "12.4", + "writerId": "uyw2t8oqcydlnz3g1p7dvon8gcx3um7p" +} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..ace7c304ae943103f31271e30e68a46a2a5f832d --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/files/wandb-summary.json @@ -0,0 +1 @@ +{"train_samples_per_second":671.984,"_timestamp":1.786745513620422e+09,"eval/steps_per_second":8.125,"train/loss":10.437238311767578,"_step":52,"train/learning_rate":0.001,"total_flos":5.420315836416e+15,"train/global_step":1000,"train/grad_norm":2.375,"_wandb":{"runtime":770},"train_loss":13.634520202636718,"_runtime":770,"train_runtime":761.9233,"eval/samples_per_second":1032.07,"train_steps_per_second":1.312,"eval/loss":2.6014304161071777,"train/epoch":0.5392289026691831,"eval/runtime":9.231} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug-core.log b/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug-core.log new file mode 100644 index 0000000000000000000000000000000000000000..588056df10bc8cabd5e2682287953b3e2b130ffe --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug-core.log @@ -0,0 +1,62 @@ +{"time":"2026-08-14T21:31:53.539152008Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp650y3sc2/port-1229423.txt","pid":1229423,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false} +{"time":"2026-08-14T21:31:53.540676876Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1229423} +{"time":"2026-08-14T21:31:53.540659035Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T21:31:53.717724383Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"} +{"time":"2026-08-14T21:31:53.802092564Z","level":"INFO","msg":"handleInformInit: received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:54.062397411Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:59.411924652Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.396635188Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.929834053Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.931620539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:48.932491197Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:49.464388506Z","level":"INFO","msg":"handleInformInit: received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:49.733853364Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:55.329792921Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.158537572Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.914569559Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.916156477Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:01.917534874Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:02.524868682Z","level":"INFO","msg":"handleInformInit: received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:02.786199875Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:08.192858634Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:53.643379161Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.173579992Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.17490129Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.175479186Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.694802218Z","level":"INFO","msg":"handleInformInit: received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:11:54.956642682Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:12:00.441973218Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:55.662446346Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.296327526Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.297810539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.29856977Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.82560196Z","level":"INFO","msg":"handleInformInit: received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:24:57.085772231Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:25:02.571913681Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:28.612623637Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.208727987Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.210106262Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.210705504Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.839594174Z","level":"INFO","msg":"handleInformInit: received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:30.099444184Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:35.620666929Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:17.704425638Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.35448297Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.355720686Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.356578583Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.899799881Z","level":"INFO","msg":"handleInformInit: received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:19.159494071Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:24.671083701Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.214192173Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.783107435Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.784410941Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:51.78516667Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088192277Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088229648Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088239651Z","level":"INFO","msg":"connection: closing","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088243873Z","level":"INFO","msg":"server: is shutting down"} +{"time":"2026-08-14T23:05:52.0882852Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088299095Z","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088345917Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088452051Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T23:05:52.088520571Z","level":"INFO","msg":"server: all connections closed"} diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug-internal.log new file mode 100644 index 0000000000000000000000000000000000000000..d4e1c857c47f5a2260b4038ca4c780dab075644e --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug-internal.log @@ -0,0 +1,119 @@ +{"time":"2026-08-14T21:59:02.525025917Z","level":"INFO","msg":"wandb-core"} +{"time":"2026-08-14T21:59:02.52517493Z","level":"INFO","msg":"stream: starting","core version":"0.28.1"} +{"time":"2026-08-14T21:59:02.786002054Z","level":"INFO","msg":"stream: created new stream","id":"npap5k3g"} +{"time":"2026-08-14T21:59:02.786104418Z","level":"INFO","msg":"handler: started"} +{"time":"2026-08-14T21:59:02.786187407Z","level":"INFO","msg":"stream: started"} +{"time":"2026-08-14T21:59:02.786198868Z","level":"INFO","msg":"writer: started","stream_id":"npap5k3g"} +{"time":"2026-08-14T21:59:02.786231041Z","level":"INFO","msg":"sender: started"} +{"time":"2026-08-14T21:59:03.16418738Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:59:03.266771519Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:59:18.164192155Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":0,"events_lines":1,"console_offset":0,"console_lines":1,"uploaded_len":2} +{"time":"2026-08-14T21:59:18.299361274Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:59:33.16463703Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":0,"history_lines":1,"events_offset":1,"events_lines":2,"console_offset":0,"console_lines":2} +{"time":"2026-08-14T21:59:33.274514188Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T21:59:48.164822613Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":1,"history_lines":1,"events_offset":3,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T21:59:48.292417282Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:00:03.164276479Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":2,"history_lines":1,"events_offset":5,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:00:03.319883137Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:00:18.164816938Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":3,"history_lines":1,"events_offset":7,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:00:18.291187426Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:00:33.164651236Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":4,"history_lines":1,"events_offset":9,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:00:33.29338924Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:00:48.164357281Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":5,"history_lines":1,"events_offset":11,"events_lines":2,"console_offset":2,"console_lines":10} +{"time":"2026-08-14T22:00:48.312694551Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:01:03.165180094Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":6,"history_lines":1,"events_offset":13,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:01:03.31910489Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:01:18.165111217Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":7,"history_lines":1,"events_offset":15,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:01:18.283466415Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:01:33.164469089Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":8,"history_lines":1,"events_offset":17,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:01:33.272073411Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:01:48.164574676Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":9,"history_lines":1,"events_offset":19,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:01:48.29257985Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:02:03.164529891Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":10,"history_lines":1,"events_offset":21,"events_lines":2,"console_offset":12,"console_lines":10} +{"time":"2026-08-14T22:02:03.386751795Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:02:18.165330631Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":1,"events_offset":23,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:02:18.338577972Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:02:33.164355386Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":12,"history_lines":1,"events_offset":25,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:02:33.267447058Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:02:48.165019444Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":1,"events_offset":27,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:02:48.275154687Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:03:03.165234131Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":14,"history_lines":1,"events_offset":29,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:03:03.302679957Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:03:18.165058948Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":15,"history_lines":1,"events_offset":31,"events_lines":2,"console_offset":22,"console_lines":10} +{"time":"2026-08-14T22:03:18.307459427Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:03:33.164356216Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:03:33.285878685Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:03:48.164932217Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":17,"history_lines":1,"events_offset":35,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:03:48.289134523Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:04:03.164590286Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":18,"history_lines":1,"events_offset":37,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:04:03.318137913Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:04:18.166174086Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":1,"events_offset":39,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:04:18.329722068Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:04:33.165069939Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":20,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":32,"console_lines":10} +{"time":"2026-08-14T22:04:33.293741837Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:04:48.165112624Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":1,"events_offset":43,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:04:48.287164722Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:05:03.164528764Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":22,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:05:03.317551641Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:05:18.164433387Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":23,"history_lines":1,"events_offset":47,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:05:18.295084679Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:05:33.164912254Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":1,"events_offset":49,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:05:33.281370888Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:05:48.164748571Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":25,"history_lines":1,"events_offset":51,"events_lines":2,"console_offset":42,"console_lines":10} +{"time":"2026-08-14T22:05:48.30493606Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:06:03.165189247Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":1,"events_offset":53,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:06:03.328534559Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:06:18.164737424Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":27,"history_lines":1,"events_offset":55,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:06:18.311417814Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:06:33.164319253Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":28,"history_lines":1,"events_offset":57,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:06:33.301054843Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:06:48.1644944Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":1,"events_offset":59,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:06:48.289979667Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:07:03.165264099Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":30,"history_lines":1,"events_offset":61,"events_lines":2,"console_offset":52,"console_lines":10} +{"time":"2026-08-14T22:07:03.343292381Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:07:18.164973207Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":1,"events_offset":63,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:07:18.353573199Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:07:33.164802894Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":32,"history_lines":1,"events_offset":65,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:07:33.305285692Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:07:48.165189676Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":1,"events_offset":67,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:07:48.289133513Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:08:03.164839714Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":34,"history_lines":1,"events_offset":69,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:08:03.354952087Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:08:18.164344367Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":1,"events_offset":71,"events_lines":2,"console_offset":62,"console_lines":10} +{"time":"2026-08-14T22:08:18.29928323Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:08:33.164862978Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":36,"history_lines":1,"events_offset":73,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:08:33.310705554Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:08:48.165076462Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":1,"events_offset":75,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:08:48.287531546Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:09:03.165128392Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":38,"history_lines":1,"events_offset":77,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:09:03.350593875Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:09:18.164398077Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":79,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:09:18.29071046Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:09:33.165180812Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":1,"events_offset":81,"events_lines":2,"console_offset":72,"console_lines":10} +{"time":"2026-08-14T22:09:33.301433559Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:09:48.16448724Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":41,"history_lines":1,"events_offset":83,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:09:48.293952896Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:10:03.16458135Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":1,"events_offset":85,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:10:03.381027285Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:10:18.164382054Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":43,"history_lines":1,"events_offset":87,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:10:18.292431171Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:10:33.16446272Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":1,"events_offset":89,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:10:33.29040667Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:10:48.16469844Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":45,"history_lines":1,"events_offset":91,"events_lines":2,"console_offset":82,"console_lines":10} +{"time":"2026-08-14T22:10:48.299861962Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:11:03.165186395Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":93,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:11:03.387650073Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:11:18.164720722Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":1,"events_offset":95,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:11:18.319368282Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:11:33.165036415Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":48,"history_lines":1,"events_offset":97,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:11:33.406468053Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:11:48.165096137Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":3,"events_offset":99,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:11:48.293634658Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:11:54.032679301Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T22:11:54.033879099Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":52,"history_lines":1,"events_offset":101,"events_lines":1,"console_offset":92,"console_lines":18,"uploaded_len":3,"complete":true,"exit_code":0} +{"time":"2026-08-14T22:11:54.171244354Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:11:54.172499989Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T22:11:54.174925302Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T22:11:54.174947987Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T22:11:54.17500805Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T22:11:54.175011794Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug.log b/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..115def628bb86cf13535aa5be317fa74a53b2e2d --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug.log @@ -0,0 +1,24 @@ +2026-08-14 21:59:02,523 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():729] Logging user logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug.log +2026-08-14 21:59:02,523 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():730] Logging internal logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_215902-npap5k3g/logs/debug-internal.log +2026-08-14 21:59:02,523 INFO MainThread:1229423 [wandb_init.py:init():772] calling init triggers +2026-08-14 21:59:02,523 INFO MainThread:1229423 [wandb_init.py:init():777] wandb.init called with sweep_config: {} +config: {'_wandb': {}} +2026-08-14 21:59:02,523 INFO MainThread:1229423 [wandb_init.py:init():820] starting backend +2026-08-14 21:59:02,523 INFO MainThread:1229423 [wandb_init.py:init():835] sending inform_init request +2026-08-14 21:59:02,786 INFO MainThread:1229423 [wandb_init.py:init():840] backend started and connected +2026-08-14 21:59:02,787 INFO MainThread:1229423 [wandb_init.py:init():910] updated telemetry +2026-08-14 21:59:02,794 INFO MainThread:1229423 [wandb_init.py:init():933] communicating run to backend with 90.0 second timeout +2026-08-14 21:59:03,083 INFO MainThread:1229423 [wandb_init.py:init():978] starting run threads in backend +2026-08-14 21:59:03,159 INFO MainThread:1229423 [wandb_run.py:_console_start():2621] atexit reg +2026-08-14 21:59:03,159 INFO MainThread:1229423 [wandb_run.py:_redirect():2471] redirect: wrap_raw +2026-08-14 21:59:03,159 INFO MainThread:1229423 [wandb_run.py:_redirect():2540] Wrapping output streams. +2026-08-14 21:59:03,159 INFO MainThread:1229423 [wandb_run.py:_redirect():2563] Redirects installed. +2026-08-14 21:59:03,160 INFO MainThread:1229423 [wandb_init.py:init():1016] run started, returning control to user process +2026-08-14 21:59:03,161 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 21, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'mlp', 'activation': 'waleed10', 'waleed_beta': 10.0, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/mlp-waleed10-21L_run', 'per_device_train_batch_size': 128, 'num_train_epochs': 1, 'max_steps': 1000, 'learning_rate': 0.001, 'lr_scheduler_type': 'constant', 'lr_scheduler_kwargs': None, 'warmup_steps': 0, 'optim': 'adamw_torch', 'optim_args': None, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 4, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-mlp-waleed10-21L-4.0M-20260814-215902', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 60000, 'eval_delay': 0, 'per_device_eval_batch_size': 128, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 100, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/A-mlp-waleed10-21L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} +2026-08-14 21:59:03,163 INFO MainThread:1229423 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 3970432 - > +2026-08-14 21:59:03,163 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 3970432 None +2026-08-14 22:11:53,642 INFO MainThread:1229423 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research1/npap5k3g +2026-08-14 22:11:53,643 INFO MainThread:1229423 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 22:11:53,643 INFO MainThread:1229423 [wandb_run.py:_restore():2570] restore +2026-08-14 22:11:53,643 INFO MainThread:1229423 [wandb_run.py:_restore():2576] restore done +2026-08-14 22:11:54,174 INFO MainThread:1229423 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_215902-npap5k3g/run-npap5k3g.wandb b/zain/Activation/wandb/run-20260814_215902-npap5k3g/run-npap5k3g.wandb new file mode 100644 index 0000000000000000000000000000000000000000..601ecd402e4daa102d503f6d69b4e4276b8b0749 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_215902-npap5k3g/run-npap5k3g.wandb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6ad8a8f597b0c7341cdd99b8a47d1d3d1be062d82972d78c7311251214c50c5 +size 563606 diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/files/config.yaml b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..f27a9516a93bfa1c405bd8e8edc7e4ca16f82739 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/config.yaml @@ -0,0 +1,439 @@ +_name_or_path: + value: "" +_wandb: + value: + cli_version: 0.28.1 + e: + uq8vhfi992gapn2qqlagfoc9r0woeq20: + args: + - --config + - /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml + - --variants + - glu-waleed10 + - glu-silu-waleed10 + - mlp-waleed10 + - mlp-silu-waleed10 + - glu-waleed + - glu-situglu_low + - glu-waleedglu_low + codePath: sweep.py + codePathLocal: sweep.py + cpu_count: 112 + cpu_count_logical: 224 + cudaVersion: "12.4" + disk: + /: + total: "1560765693952" + used: "709095636992" + email: deepnevro@gmail.com + executable: /mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python + git: + commit: 3fb47c6943d6927eafc39bd399a3d3a520bae74a + remote: https://github.com/w-ahmad1a10/Activation.git + gpu: NVIDIA H100 80GB HBM3 + gpu_count: 8 + gpu_nvidia: + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-39c684a5-fde6-83d7-1663-0859795881ae + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-132944c4-b689-2b5f-89a4-d730401677ab + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bc6c3e3c-9b90-09ca-c034-774961847c54 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea + host: deeplens-k3s-node1 + memory: + total: "2164089937920" + os: Linux-5.15.0-126-generic-x86_64-with-glibc2.35 + program: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py + python: CPython 3.11.15 + root: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation + startedAt: "2026-08-14T22:11:54.692447Z" + writerId: uq8vhfi992gapn2qqlagfoc9r0woeq20 + m: + - "1": train/global_step + "6": + - 3 + "7": [] + - "2": '*' + "5": 1 + "6": + - 1 + "7": [] + python_version: 3.11.15 + t: + "1": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "2": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "3": + - 2 + - 7 + - 13 + - 19 + - 62 + - 66 + "4": 3.11.15 + "5": 0.28.1 + "6": 5.16.0.dev0 + "9": + "1": transformers_trainer + "12": 0.28.1 + "13": linux-x86_64 +accelerator_config: + value: + dispatch_batches: null + even_batches: true + gradient_accumulation_kwargs: null + non_blocking: false + split_batches: false + use_seedable_sampler: true +activation: + value: silu-waleed10 +adam_beta1: + value: 0.9 +adam_beta2: + value: 0.999 +adam_epsilon: + value: 1e-08 +architectures: + value: null +attention_bias: + value: false +attention_dropout: + value: 0 +auto_find_batch_size: + value: false +average_tokens_across_devices: + value: true +batch_eval_metrics: + value: false +bf16: + value: true +bf16_full_eval: + value: false +bos_token_id: + value: 1 +chunk_size_feed_forward: + value: 0 +data_seed: + value: 42 +dataloader_drop_last: + value: false +dataloader_in_order: + value: true +dataloader_multiprocessing_context: + value: null +dataloader_num_workers: + value: 0 +dataloader_persistent_workers: + value: false +dataloader_pin_memory: + value: true +dataloader_prefetch_factor: + value: null +ddp_backend: + value: null +ddp_broadcast_buffers: + value: null +ddp_bucket_cap_mb: + value: null +ddp_find_unused_parameters: + value: null +ddp_static_graph: + value: null +ddp_timeout: + value: 1800 +debug: + value: [] +deepspeed: + value: null +disable_tqdm: + value: false +do_eval: + value: true +do_predict: + value: false +do_train: + value: false +dtype: + value: null +enable_jit_checkpoint: + value: false +eos_token_id: + value: 2 +eval_accumulation_steps: + value: null +eval_delay: + value: 0 +eval_do_concat_batches: + value: true +eval_on_start: + value: false +eval_steps: + value: 60000 +eval_strategy: + value: steps +eval_use_gather_object: + value: false +fp16: + value: false +fp16_full_eval: + value: false +fsdp: + value: null +fsdp_config: + value: null +full_determinism: + value: false +gradient_accumulation_steps: + value: 4 +gradient_checkpointing: + value: false +gradient_checkpointing_kwargs: + value: null +greater_is_better: + value: null +head_dim: + value: 32 +hidden_act: + value: silu +hidden_size: + value: 128 +hub_always_push: + value: false +hub_model_id: + value: w-ahmad/A-mlp-silu-waleed10-21L +hub_private_repo: + value: null +hub_revision: + value: null +hub_strategy: + value: every_save +hub_token: + value: +id2label: + value: + "0": LABEL_0 + "1": LABEL_1 +ignore_data_skip: + value: false +include_for_metrics: + value: [] +include_num_input_tokens_seen: + value: "no" +initializer_range: + value: 0.02 +intermediate_size: + value: 256 +is_encoder_decoder: + value: false +label_names: + value: null +label_smoothing_factor: + value: 0 +label2id: + value: + LABEL_0: 0 + LABEL_1: 1 +learning_rate: + value: 0.001 +length_column_name: + value: length +liger_kernel_config: + value: null +load_best_model_at_end: + value: false +local_rank: + value: -1 +log_level: + value: passive +log_level_replica: + value: warning +log_on_each_node: + value: true +logging_first_step: + value: false +logging_nan_inf_filter: + value: true +logging_steps: + value: 20 +logging_strategy: + value: steps +lr_scheduler_kwargs: + value: null +lr_scheduler_type: + value: constant +max_grad_norm: + value: 1 +max_position_embeddings: + value: 512 +max_steps: + value: 1000 +metric_for_best_model: + value: null +mlp_bias: + value: false +mlp_type: + value: mlp +model/num_parameters: + value: 3970432 +model_type: + value: tiny_llama +neftune_noise_alpha: + value: null +num_attention_heads: + value: 4 +num_hidden_layers: + value: 21 +num_key_value_heads: + value: 4 +num_train_epochs: + value: 1 +optim: + value: adamw_torch +optim_args: + value: null +optim_target_modules: + value: null +output_attentions: + value: false +output_dir: + value: out/mlp-silu-waleed10-21L_run +output_hidden_states: + value: false +pad_token_id: + value: 0 +parallelism_config: + value: null +per_device_eval_batch_size: + value: 128 +per_device_train_batch_size: + value: 128 +prediction_loss_only: + value: false +pretraining_tp: + value: 1 +problem_type: + value: null +project: + value: huggingface +push_to_hub: + value: false +remove_unused_columns: + value: false +report_to: + value: + - wandb +restore_callback_states_from_checkpoint: + value: false +resume_from_checkpoint: + value: null +return_dict: + value: true +rms_norm_eps: + value: 1e-06 +rope_parameters: + value: + rope_theta: 10000 + rope_type: default +run_name: + value: LM-mlp-silu-waleed10-21L-4.0M-20260814-221154 +save_on_each_node: + value: false +save_only_model: + value: false +save_steps: + value: 100 +save_strategy: + value: steps +save_total_limit: + value: null +seed: + value: 42 +skip_memory_metrics: + value: true +tf32: + value: null +tie_word_embeddings: + value: true +tokenizer_name: + value: w-ahmad/tiny-stories-tokenizer +torch_compile: + value: false +torch_compile_backend: + value: null +torch_compile_mode: + value: null +torch_empty_cache_steps: + value: null +trackio_bucket_id: + value: null +trackio_space_id: + value: null +trackio_static_space_id: + value: null +train_sampling_strategy: + value: random +transformers_version: + value: 5.16.0.dev0 +use_cache: + value: false +use_cpu: + value: false +use_liger_kernel: + value: false +vocab_size: + value: 4096 +waleed_beta: + value: 10 +warmup_steps: + value: 0 +weight_decay: + value: 0.01 diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/files/output.log b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/output.log new file mode 100644 index 0000000000000000000000000000000000000000..a775d120c328fabcb28a18da1dde27c5510dca86 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/output.log @@ -0,0 +1,110 @@ + 10%|█ | 100/1000 [01:16<11:17, 1.33it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '29.1', 'grad_norm': '4.094', 'learning_rate': '0.001', 'epoch': '0.01078', 'train/total_time_seconds': '7.585', 'train/time_per_step_avg': '0.3793', 'train/epoch_time_elapsed': '15.48', 'train/estimated_remaining_minutes': '6.194'} +{'loss': '24.32', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.02157', 'train/total_time_seconds': '15.13', 'train/time_per_step_avg': '0.3783', 'train/epoch_time_elapsed': '30.63', 'train/estimated_remaining_minutes': '6.053'} +{'loss': '22.97', 'grad_norm': '1.391', 'learning_rate': '0.001', 'epoch': '0.03235', 'train/total_time_seconds': '22.67', 'train/time_per_step_avg': '0.3779', 'train/epoch_time_elapsed': '45.83', 'train/estimated_remaining_minutes': '5.92'} +{'loss': '21.97', 'grad_norm': '1.625', 'learning_rate': '0.001', 'epoch': '0.04314', 'train/total_time_seconds': '30.22', 'train/time_per_step_avg': '0.3777', 'train/epoch_time_elapsed': '60.9', 'train/estimated_remaining_minutes': '5.792'} +{'loss': '20.95', 'grad_norm': '3.625', 'learning_rate': '0.001', 'epoch': '0.05392', 'train/total_time_seconds': '37.75', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '76.14', 'train/estimated_remaining_minutes': '5.662'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 47.25it/s] + 20%|██ | 200/1000 [02:31<09:58, 1.34it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '20.18', 'grad_norm': '3.375', 'learning_rate': '0.001', 'epoch': '0.06471', 'train/total_time_seconds': '45.25', 'train/time_per_step_avg': '0.3767', 'train/epoch_time_elapsed': '91.24', 'train/estimated_remaining_minutes': '5.531'} +{'loss': '19.39', 'grad_norm': '3.797', 'learning_rate': '0.001', 'epoch': '0.07549', 'train/total_time_seconds': '52.77', 'train/time_per_step_avg': '0.3763', 'train/epoch_time_elapsed': '106.3', 'train/estimated_remaining_minutes': '5.402'} +{'loss': '18.66', 'grad_norm': '3.219', 'learning_rate': '0.001', 'epoch': '0.08628', 'train/total_time_seconds': '60.34', 'train/time_per_step_avg': '0.3767', 'train/epoch_time_elapsed': '121.4', 'train/estimated_remaining_minutes': '5.28'} +{'loss': '17.96', 'grad_norm': '2.609', 'learning_rate': '0.001', 'epoch': '0.09706', 'train/total_time_seconds': '67.9', 'train/time_per_step_avg': '0.3768', 'train/epoch_time_elapsed': '136.7', 'train/estimated_remaining_minutes': '5.155'} +{'loss': '17.32', 'grad_norm': '2.359', 'learning_rate': '0.001', 'epoch': '0.1078', 'train/total_time_seconds': '75.42', 'train/time_per_step_avg': '0.3767', 'train/epoch_time_elapsed': '151.7', 'train/estimated_remaining_minutes': '5.028'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 85.88it/s] + 30%|███ | 300/1000 [03:47<09:14, 1.26it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '16.67', 'grad_norm': '3.484', 'learning_rate': '0.001', 'epoch': '0.1186', 'train/total_time_seconds': '82.96', 'train/time_per_step_avg': '0.3771', 'train/epoch_time_elapsed': '166.9', 'train/estimated_remaining_minutes': '4.902'} +{'loss': '16.15', 'grad_norm': '4.062', 'learning_rate': '0.001', 'epoch': '0.1294', 'train/total_time_seconds': '90.52', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '182', 'train/estimated_remaining_minutes': '4.777'} +{'loss': '15.67', 'grad_norm': '2.344', 'learning_rate': '0.001', 'epoch': '0.1402', 'train/total_time_seconds': '98.09', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '197.2', 'train/estimated_remaining_minutes': '4.653'} +{'loss': '15.26', 'grad_norm': '2.984', 'learning_rate': '0.001', 'epoch': '0.151', 'train/total_time_seconds': '105.6', 'train/time_per_step_avg': '0.3774', 'train/epoch_time_elapsed': '212.4', 'train/estimated_remaining_minutes': '4.527'} +{'loss': '14.91', 'grad_norm': '1.812', 'learning_rate': '0.001', 'epoch': '0.1618', 'train/total_time_seconds': '113.2', 'train/time_per_step_avg': '0.3778', 'train/epoch_time_elapsed': '227.8', 'train/estimated_remaining_minutes': '4.402'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 67.77it/s] + 40%|████ | 400/1000 [05:04<07:29, 1.33it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '14.61', 'grad_norm': '2.391', 'learning_rate': '0.001', 'epoch': '0.1726', 'train/total_time_seconds': '120.8', 'train/time_per_step_avg': '0.3781', 'train/epoch_time_elapsed': '243.2', 'train/estimated_remaining_minutes': '4.277'} +{'loss': '14.27', 'grad_norm': '1.648', 'learning_rate': '0.001', 'epoch': '0.1833', 'train/total_time_seconds': '128.3', 'train/time_per_step_avg': '0.3783', 'train/epoch_time_elapsed': '258.5', 'train/estimated_remaining_minutes': '4.152'} +{'loss': '13.94', 'grad_norm': '2.031', 'learning_rate': '0.001', 'epoch': '0.1941', 'train/total_time_seconds': '135.9', 'train/time_per_step_avg': '0.3781', 'train/epoch_time_elapsed': '273.8', 'train/estimated_remaining_minutes': '4.027'} +{'loss': '13.63', 'grad_norm': '2.75', 'learning_rate': '0.001', 'epoch': '0.2049', 'train/total_time_seconds': '143.5', 'train/time_per_step_avg': '0.3784', 'train/epoch_time_elapsed': '289.1', 'train/estimated_remaining_minutes': '3.902'} +{'loss': '13.42', 'grad_norm': '2.688', 'learning_rate': '0.001', 'epoch': '0.2157', 'train/total_time_seconds': '151', 'train/time_per_step_avg': '0.3778', 'train/epoch_time_elapsed': '304', 'train/estimated_remaining_minutes': '3.774'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 89.52it/s] + 50%|█████ | 500/1000 [06:20<06:18, 1.32it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '13.16', 'grad_norm': '2.641', 'learning_rate': '0.001', 'epoch': '0.2265', 'train/total_time_seconds': '158.5', 'train/time_per_step_avg': '0.3777', 'train/epoch_time_elapsed': '319.4', 'train/estimated_remaining_minutes': '3.649'} +{'loss': '12.97', 'grad_norm': '2.031', 'learning_rate': '0.001', 'epoch': '0.2373', 'train/total_time_seconds': '166.1', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '334.5', 'train/estimated_remaining_minutes': '3.523'} +{'loss': '12.77', 'grad_norm': '2.375', 'learning_rate': '0.001', 'epoch': '0.248', 'train/total_time_seconds': '173.7', 'train/time_per_step_avg': '0.3776', 'train/epoch_time_elapsed': '349.6', 'train/estimated_remaining_minutes': '3.398'} +{'loss': '12.56', 'grad_norm': '1.961', 'learning_rate': '0.001', 'epoch': '0.2588', 'train/total_time_seconds': '181.2', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '364.9', 'train/estimated_remaining_minutes': '3.272'} +{'loss': '12.36', 'grad_norm': '1.664', 'learning_rate': '0.001', 'epoch': '0.2696', 'train/total_time_seconds': '188.8', 'train/time_per_step_avg': '0.3782', 'train/epoch_time_elapsed': '380.1', 'train/estimated_remaining_minutes': '3.147'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.83it/s] + 60%|██████ | 600/1000 [07:36<05:01, 1.33it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '12.23', 'grad_norm': '2.406', 'learning_rate': '0.001', 'epoch': '0.2804', 'train/total_time_seconds': '196.4', 'train/time_per_step_avg': '0.3783', 'train/epoch_time_elapsed': '395.5', 'train/estimated_remaining_minutes': '3.021'} +{'loss': '12.07', 'grad_norm': '2.188', 'learning_rate': '0.001', 'epoch': '0.2912', 'train/total_time_seconds': '203.9', 'train/time_per_step_avg': '0.3782', 'train/epoch_time_elapsed': '410.7', 'train/estimated_remaining_minutes': '2.895'} +{'loss': '11.92', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.302', 'train/total_time_seconds': '211.4', 'train/time_per_step_avg': '0.3777', 'train/epoch_time_elapsed': '425.8', 'train/estimated_remaining_minutes': '2.769'} +{'loss': '11.78', 'grad_norm': '2.062', 'learning_rate': '0.001', 'epoch': '0.3128', 'train/total_time_seconds': '218.9', 'train/time_per_step_avg': '0.3772', 'train/epoch_time_elapsed': '440.9', 'train/estimated_remaining_minutes': '2.642'} +{'loss': '11.67', 'grad_norm': '2.562', 'learning_rate': '0.001', 'epoch': '0.3235', 'train/total_time_seconds': '226.5', 'train/time_per_step_avg': '0.3767', 'train/epoch_time_elapsed': '456.3', 'train/estimated_remaining_minutes': '2.516'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 78.46it/s] + 70%|███████ | 700/1000 [08:52<03:45, 1.33it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '11.55', 'grad_norm': '1.992', 'learning_rate': '0.001', 'epoch': '0.3343', 'train/total_time_seconds': '234', 'train/time_per_step_avg': '0.3766', 'train/epoch_time_elapsed': '471.7', 'train/estimated_remaining_minutes': '2.391'} +{'loss': '11.43', 'grad_norm': '2', 'learning_rate': '0.001', 'epoch': '0.3451', 'train/total_time_seconds': '241.6', 'train/time_per_step_avg': '0.3767', 'train/epoch_time_elapsed': '486.9', 'train/estimated_remaining_minutes': '2.265'} +{'loss': '11.32', 'grad_norm': '2.203', 'learning_rate': '0.001', 'epoch': '0.3559', 'train/total_time_seconds': '249.1', 'train/time_per_step_avg': '0.377', 'train/epoch_time_elapsed': '502.2', 'train/estimated_remaining_minutes': '2.139'} +{'loss': '11.2', 'grad_norm': '2.156', 'learning_rate': '0.001', 'epoch': '0.3667', 'train/total_time_seconds': '256.7', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '517.4', 'train/estimated_remaining_minutes': '2.013'} +{'loss': '11.1', 'grad_norm': '2.422', 'learning_rate': '0.001', 'epoch': '0.3775', 'train/total_time_seconds': '264.3', 'train/time_per_step_avg': '0.3779', 'train/epoch_time_elapsed': '532.6', 'train/estimated_remaining_minutes': '1.888'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 47.34it/s] + 80%|████████ | 800/1000 [10:08<02:32, 1.31it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.98', 'grad_norm': '2.031', 'learning_rate': '0.001', 'epoch': '0.3882', 'train/total_time_seconds': '271.8', 'train/time_per_step_avg': '0.378', 'train/epoch_time_elapsed': '547.8', 'train/estimated_remaining_minutes': '1.762'} +{'loss': '10.93', 'grad_norm': '2.781', 'learning_rate': '0.001', 'epoch': '0.399', 'train/total_time_seconds': '279.4', 'train/time_per_step_avg': '0.378', 'train/epoch_time_elapsed': '563.1', 'train/estimated_remaining_minutes': '1.636'} +{'loss': '10.84', 'grad_norm': '2.266', 'learning_rate': '0.001', 'epoch': '0.4098', 'train/total_time_seconds': '286.9', 'train/time_per_step_avg': '0.3777', 'train/epoch_time_elapsed': '578.3', 'train/estimated_remaining_minutes': '1.51'} +{'loss': '10.76', 'grad_norm': '2.531', 'learning_rate': '0.001', 'epoch': '0.4206', 'train/total_time_seconds': '294.5', 'train/time_per_step_avg': '0.3779', 'train/epoch_time_elapsed': '593.6', 'train/estimated_remaining_minutes': '1.384'} +{'loss': '10.63', 'grad_norm': '2.375', 'learning_rate': '0.001', 'epoch': '0.4314', 'train/total_time_seconds': '302', 'train/time_per_step_avg': '0.3777', 'train/epoch_time_elapsed': '609', 'train/estimated_remaining_minutes': '1.258'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 64.95it/s] + 90%|█████████ | 900/1000 [11:25<01:18, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.59', 'grad_norm': '2.469', 'learning_rate': '0.001', 'epoch': '0.4422', 'train/total_time_seconds': '309.6', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '624.2', 'train/estimated_remaining_minutes': '1.133'} +{'loss': '10.52', 'grad_norm': '2.531', 'learning_rate': '0.001', 'epoch': '0.453', 'train/total_time_seconds': '317.1', 'train/time_per_step_avg': '0.3776', 'train/epoch_time_elapsed': '639.5', 'train/estimated_remaining_minutes': '1.007'} +{'loss': '10.43', 'grad_norm': '2.484', 'learning_rate': '0.001', 'epoch': '0.4637', 'train/total_time_seconds': '324.7', 'train/time_per_step_avg': '0.3777', 'train/epoch_time_elapsed': '654.6', 'train/estimated_remaining_minutes': '0.8809'} +{'loss': '10.36', 'grad_norm': '2.234', 'learning_rate': '0.001', 'epoch': '0.4745', 'train/total_time_seconds': '332.2', 'train/time_per_step_avg': '0.3775', 'train/epoch_time_elapsed': '670', 'train/estimated_remaining_minutes': '0.7551'} +{'loss': '10.3', 'grad_norm': '2.641', 'learning_rate': '0.001', 'epoch': '0.4853', 'train/total_time_seconds': '339.8', 'train/time_per_step_avg': '0.3778', 'train/epoch_time_elapsed': '685.3', 'train/estimated_remaining_minutes': '0.6293'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 82.06it/s] +100%|██████████| 1000/1000 [12:50<00:00, 1.28i[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.23', 'grad_norm': '2.703', 'learning_rate': '0.001', 'epoch': '0.4961', 'train/total_time_seconds': '347.4', 'train/time_per_step_avg': '0.3777', 'train/epoch_time_elapsed': '700.6', 'train/estimated_remaining_minutes': '0.5034'} +{'loss': '10.2', 'grad_norm': '2.984', 'learning_rate': '0.001', 'epoch': '0.5069', 'train/total_time_seconds': '354.9', 'train/time_per_step_avg': '0.3773', 'train/epoch_time_elapsed': '715.6', 'train/estimated_remaining_minutes': '0.3775'} +{'loss': '10.15', 'grad_norm': '2.234', 'learning_rate': '0.001', 'epoch': '0.5177', 'train/total_time_seconds': '362.4', 'train/time_per_step_avg': '0.377', 'train/epoch_time_elapsed': '730.9', 'train/estimated_remaining_minutes': '0.2516'} +{'loss': '10.11', 'grad_norm': '2.469', 'learning_rate': '0.001', 'epoch': '0.5284', 'train/total_time_seconds': '369.9', 'train/time_per_step_avg': '0.3765', 'train/epoch_time_elapsed': '746', 'train/estimated_remaining_minutes': '0.1258'} +{'loss': '10.05', 'grad_norm': '2.609', 'learning_rate': '0.001', 'epoch': '0.5392', 'train/total_time_seconds': '377.4', 'train/time_per_step_avg': '0.376', 'train/epoch_time_elapsed': '761.2', 'train/estimated_remaining_minutes': '0'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes +{'eval_loss': '2.505', 'eval_runtime': '9.414', 'eval_samples_per_second': '1012', 'eval_steps_per_second': '7.967', 'epoch': '0.5392', 'train/total_time_seconds': '377.4', 'train/time_per_step_avg': '0.376', 'train/epoch_time_elapsed': '770.6', 'train/estimated_remaining_minutes': '0'} + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 78.21it/s] +100%|██████████| 1000/1000 [12:50<00:00, 1.30it/s], ?it/s] +{'train_runtime': '771.5', 'train_samples_per_second': '663.7', 'train_steps_per_second': '1.296', 'train_loss': '13.97', 'epoch': '0.5392', 'train/total_time_seconds': '377.4', 'train/time_per_step_avg': '0.376', 'train/epoch_time_elapsed': '770.7', 'train/estimated_remaining_minutes': '0'} +100%|██████████| 75/75 [00:09<00:00, 8.05it/s] +[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 81.46it/s] +>>> FINISHED mlp-silu-waleed10-21L successfully diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/files/requirements.txt b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..f36b19895172d9637438a445ceacb71d09dc5d02 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/requirements.txt @@ -0,0 +1,149 @@ +asttokens==3.0.1 +comm==0.2.3 +debugpy==1.8.21 +decorator==5.3.1 +executing==2.2.1 +nest-asyncio==1.6.0 +parso==0.8.7 +platformdirs==4.11.0 +psutil==7.2.2 +ptyprocess==0.7.0 +pure_eval==0.2.3 +Pygments==2.20.0 +pyzmq==27.1.0 +setuptools==83.0.0 +six==1.17.0 +tornado==6.5.7 +traitlets==5.15.0 +fsspec==2026.4.0 +wcwidth==0.8.2 +ipython_pygments_lexers==1.1.1 +jedi==0.20.0 +jupyter_core==5.9.1 +matplotlib-inline==0.2.2 +pexpect==4.9.0 +prompt_toolkit==3.0.53 +python-dateutil==2.9.0.post0 +stack_data==0.6.3 +wheel==0.47.0 +jupyter_client==8.9.1 +pip==26.1.2 +ipython==9.15.0 +ipykernel==7.2.0 +threadpoolctl==3.6.0 +pyparsing==3.3.2 +typing_extensions==4.15.0 +Jinja2==3.1.6 +narwhals==2.24.0 +kiwisolver==1.5.0 +joblib==1.5.3 +fonttools==4.63.0 +cycler==0.12.1 +scipy==1.17.1 +pandas==3.0.5 +contourpy==1.3.3 +scikit-learn==1.9.0 +matplotlib==3.11.1 +urllib3==2.7.0 +tqdm==4.70.0 +idna==3.18 +charset-normalizer==3.4.9 +certifi==2026.7.22 +requests==2.34.2 +seaborn==0.13.2 +uv==0.12.0 +shellingham==1.5.4 +mpmath==1.3.0 +attrs==26.1.0 +hf-xet==1.5.2 +nvidia-nccl-cu12==2.21.5 +MarkupSafe==3.0.3 +regex==2026.7.19 +importlib_metadata==9.0.0 +httpcore==1.0.9 +annotated-doc==0.0.5 +multidict==6.7.1 +aiohttp==3.14.3 +aiosignal==1.4.0 +xxhash==3.8.1 +aiohappyeyeballs==2.7.1 +mdurl==0.1.2 +cuda-toolkit==13.0.3.0 +networkx==3.6.1 +PyYAML==6.0.3 +nvidia-cufile==1.15.1.6 +typer==0.27.0 +torchaudio==2.6.0+cu124 +rich==15.0.0 +nvidia-cufft-cu12==11.2.1.3 +h11==0.16.0 +dill==0.4.1 +cuda-pathfinder==1.6.0 +filelock==3.29.0 +nvidia-nvtx-cu12==12.4.127 +httpx==0.28.1 +anyio==4.14.2 +numpy==2.4.4 +yarl==1.24.5 +click==8.4.2 +triton==3.2.0 +frozenlist==1.8.0 +zipp==4.1.0 +propcache==0.5.2 +tokenizers==0.22.2 +markdown-it-py==4.2.0 +nvidia-cuda-runtime==13.0.96 +cuda-bindings==13.3.1 +nvidia-cuda-cupti==13.0.85 +torch==2.6.0+cu124 +multiprocess==0.70.19 +pillow==12.2.0 +transformers==5.16.0.dev0 +wandb==0.28.1 +nvidia-curand==10.4.0.35 +sympy==1.13.1 +nvidia-cusparse==12.6.3.3 +nvidia-cuda-nvrtc==13.0.88 +typing-inspection==0.4.2 +nvidia-cusolver==12.0.4.66 +nvidia-cufft==12.0.0.61 +nvidia-cudnn-cu13==9.20.0.48 +nvidia-cublas==13.1.1.3 +pyarrow==25.0.0 +evaluate==0.4.6 +diffusers==0.39.0 +pydantic==2.13.4 +annotated-types==0.8.0 +protobuf==7.35.1 +sentry-sdk==2.66.1 +einops==0.8.2 +packaging==26.2 +nvidia-nvjitlink-cu12==12.4.127 +nvidia-curand-cu12==10.3.5.147 +nvidia-cusparselt-cu12==0.6.2 +nvidia-cusparse-cu12==12.3.1.170 +nvidia-cuda-runtime-cu12==12.4.127 +torchvision==0.21.0+cu124 +nvidia-cuda-nvrtc-cu12==12.4.127 +nvidia-cuda-cupti-cu12==12.4.127 +nvidia-cusolver-cu12==11.6.1.9 +nvidia-cublas-cu12==12.4.5.8 +nvidia-cudnn-cu12==9.1.0.70 +huggingface_hub==1.26.0 +datasets==5.0.1 +safetensors==0.8.0 +accelerate==1.14.0 +pydantic_core==2.46.4 +ninja==1.13.0 +autocommand==2.2.2 +backports.tarfile==1.2.0 +importlib_metadata==8.7.1 +jaraco.text==4.0.0 +jaraco.context==6.1.0 +jaraco.functools==4.4.0 +more-itertools==10.8.0 +packaging==26.0 +platformdirs==4.4.0 +tomli==2.4.0 +wheel==0.46.3 +zipp==3.23.0 diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/files/wandb-metadata.json b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/wandb-metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..9deffabdebf1529a3981fa08e2e37285c02facb5 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/wandb-metadata.json @@ -0,0 +1,101 @@ +{ + "os": "Linux-5.15.0-126-generic-x86_64-with-glibc2.35", + "python": "CPython 3.11.15", + "startedAt": "2026-08-14T22:11:54.692447Z", + "args": [ + "--config", + "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml", + "--variants", + "glu-waleed10", + "glu-silu-waleed10", + "mlp-waleed10", + "mlp-silu-waleed10", + "glu-waleed", + "glu-situglu_low", + "glu-waleedglu_low" + ], + "program": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py", + "codePath": "sweep.py", + "codePathLocal": "sweep.py", + "git": { + "remote": "https://github.com/w-ahmad1a10/Activation.git", + "commit": "3fb47c6943d6927eafc39bd399a3d3a520bae74a" + }, + "email": "deepnevro@gmail.com", + "root": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation", + "host": "deeplens-k3s-node1", + "executable": "/mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python", + "cpu_count": 112, + "cpu_count_logical": 224, + "gpu": "NVIDIA H100 80GB HBM3", + "gpu_count": 8, + "disk": { + "/": { + "total": "1560765693952", + "used": "709095636992" + } + }, + "memory": { + "total": "2164089937920" + }, + "gpu_nvidia": [ + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-39c684a5-fde6-83d7-1663-0859795881ae" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-132944c4-b689-2b5f-89a4-d730401677ab" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bc6c3e3c-9b90-09ca-c034-774961847c54" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea" + } + ], + "cudaVersion": "12.4", + "writerId": "uq8vhfi992gapn2qqlagfoc9r0woeq20" +} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..7d05b76ceb36b4dc7fc62c2d41ba0ca481bbe35f --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/files/wandb-summary.json @@ -0,0 +1 @@ +{"total_flos":5.420315836416e+15,"train_samples_per_second":663.654,"eval/steps_per_second":7.88,"train_runtime":771.486,"train_loss":13.971002395629883,"eval/runtime":9.5175,"_wandb":{"runtime":780},"train/global_step":1000,"train/loss":10.053606414794922,"_timestamp":1.7867462956361027e+09,"eval/loss":2.5052456855773926,"train/learning_rate":0.001,"_step":52,"train/grad_norm":2.609375,"_runtime":780,"eval/samples_per_second":1001.003,"train_steps_per_second":1.296,"train/epoch":0.5392289026691831} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug-core.log b/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug-core.log new file mode 100644 index 0000000000000000000000000000000000000000..588056df10bc8cabd5e2682287953b3e2b130ffe --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug-core.log @@ -0,0 +1,62 @@ +{"time":"2026-08-14T21:31:53.539152008Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp650y3sc2/port-1229423.txt","pid":1229423,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false} +{"time":"2026-08-14T21:31:53.540676876Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1229423} +{"time":"2026-08-14T21:31:53.540659035Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T21:31:53.717724383Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"} +{"time":"2026-08-14T21:31:53.802092564Z","level":"INFO","msg":"handleInformInit: received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:54.062397411Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:59.411924652Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.396635188Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.929834053Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.931620539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:48.932491197Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:49.464388506Z","level":"INFO","msg":"handleInformInit: received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:49.733853364Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:55.329792921Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.158537572Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.914569559Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.916156477Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:01.917534874Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:02.524868682Z","level":"INFO","msg":"handleInformInit: received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:02.786199875Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:08.192858634Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:53.643379161Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.173579992Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.17490129Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.175479186Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.694802218Z","level":"INFO","msg":"handleInformInit: received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:11:54.956642682Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:12:00.441973218Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:55.662446346Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.296327526Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.297810539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.29856977Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.82560196Z","level":"INFO","msg":"handleInformInit: received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:24:57.085772231Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:25:02.571913681Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:28.612623637Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.208727987Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.210106262Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.210705504Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.839594174Z","level":"INFO","msg":"handleInformInit: received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:30.099444184Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:35.620666929Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:17.704425638Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.35448297Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.355720686Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.356578583Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.899799881Z","level":"INFO","msg":"handleInformInit: received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:19.159494071Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:24.671083701Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.214192173Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.783107435Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.784410941Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:51.78516667Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088192277Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088229648Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088239651Z","level":"INFO","msg":"connection: closing","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088243873Z","level":"INFO","msg":"server: is shutting down"} +{"time":"2026-08-14T23:05:52.0882852Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088299095Z","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088345917Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088452051Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T23:05:52.088520571Z","level":"INFO","msg":"server: all connections closed"} diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug-internal.log new file mode 100644 index 0000000000000000000000000000000000000000..38ff0e8ba586d18e7156ee891f4ac27478d3e810 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug-internal.log @@ -0,0 +1,121 @@ +{"time":"2026-08-14T22:11:54.694930488Z","level":"INFO","msg":"wandb-core"} +{"time":"2026-08-14T22:11:54.695061314Z","level":"INFO","msg":"stream: starting","core version":"0.28.1"} +{"time":"2026-08-14T22:11:54.956465986Z","level":"INFO","msg":"stream: created new stream","id":"a3607q94"} +{"time":"2026-08-14T22:11:54.956541268Z","level":"INFO","msg":"handler: started"} +{"time":"2026-08-14T22:11:54.95663725Z","level":"INFO","msg":"stream: started"} +{"time":"2026-08-14T22:11:54.956647578Z","level":"INFO","msg":"writer: started","stream_id":"a3607q94"} +{"time":"2026-08-14T22:11:54.9566745Z","level":"INFO","msg":"sender: started"} +{"time":"2026-08-14T22:11:55.405026716Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:11:55.518319573Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:12:10.405391964Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":0,"events_lines":1,"console_offset":0,"console_lines":1,"uploaded_len":2} +{"time":"2026-08-14T22:12:10.559461859Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:12:25.405485607Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":0,"history_lines":1,"events_offset":1,"events_lines":2,"console_offset":0,"console_lines":2} +{"time":"2026-08-14T22:12:25.550239937Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:12:40.405268298Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":1,"history_lines":1,"events_offset":3,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:12:40.5116701Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:12:55.406850003Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":2,"history_lines":1,"events_offset":5,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:12:55.555408593Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:13:10.405496619Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":3,"history_lines":1,"events_offset":7,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:13:10.5558936Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:13:25.405663028Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":4,"history_lines":1,"events_offset":9,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:13:25.538451856Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:13:40.405804156Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":5,"history_lines":1,"events_offset":11,"events_lines":2,"console_offset":2,"console_lines":10} +{"time":"2026-08-14T22:13:40.533262745Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:13:55.405648498Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":6,"history_lines":1,"events_offset":13,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:13:55.529428811Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:14:10.405699023Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":7,"history_lines":1,"events_offset":15,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:14:10.554609811Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:14:25.405296636Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":8,"history_lines":1,"events_offset":17,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:14:25.544426629Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:14:40.405384171Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":9,"history_lines":1,"events_offset":19,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:14:40.53709818Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:14:55.405885358Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":10,"history_lines":1,"events_offset":21,"events_lines":2,"console_offset":12,"console_lines":10} +{"time":"2026-08-14T22:14:55.538990888Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:15:10.405566821Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":1,"events_offset":23,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:15:10.550080507Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:15:25.405918015Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":12,"history_lines":1,"events_offset":25,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:15:25.513709253Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:15:40.405536965Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":1,"events_offset":27,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:15:40.505023493Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:15:55.405149988Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":14,"history_lines":1,"events_offset":29,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:15:55.51412857Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:16:10.405144989Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":15,"history_lines":1,"events_offset":31,"events_lines":2,"console_offset":22,"console_lines":10} +{"time":"2026-08-14T22:16:10.541001666Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:16:25.405968509Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:16:25.508694054Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:16:40.405798516Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":17,"history_lines":1,"events_offset":35,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:16:40.533724449Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:16:55.405448681Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":18,"history_lines":1,"events_offset":37,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:16:55.528250746Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:17:10.406148823Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":1,"events_offset":39,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:17:10.549567524Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:17:25.405294454Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":20,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":32,"console_lines":10} +{"time":"2026-08-14T22:17:25.532071927Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:17:40.405762467Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":1,"events_offset":43,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:17:40.528553533Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:17:55.405245047Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":22,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:17:55.535139755Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:18:10.405494472Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":23,"history_lines":1,"events_offset":47,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:18:10.556962534Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:18:25.405906254Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":1,"events_offset":49,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:18:25.543583377Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:18:40.405684146Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":25,"history_lines":1,"events_offset":51,"events_lines":2,"console_offset":42,"console_lines":10} +{"time":"2026-08-14T22:18:40.517192829Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:18:55.405352723Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":1,"events_offset":53,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:18:55.556085442Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:19:10.405214712Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":27,"history_lines":1,"events_offset":55,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:19:10.549268441Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:19:25.405984196Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":28,"history_lines":1,"events_offset":57,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:19:25.524400916Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:19:40.406034794Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":1,"events_offset":59,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:19:40.529752452Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:19:55.405562656Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":30,"history_lines":1,"events_offset":61,"events_lines":2,"console_offset":52,"console_lines":10} +{"time":"2026-08-14T22:19:55.522207405Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:20:10.405470142Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":1,"events_offset":63,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:20:10.520670867Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:20:25.40533656Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":32,"history_lines":1,"events_offset":65,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:20:25.548953084Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:20:40.405524309Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":1,"events_offset":67,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:20:40.546057054Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:20:55.405924314Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":34,"history_lines":1,"events_offset":69,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:20:55.51483369Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:21:10.405655258Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":1,"events_offset":71,"events_lines":2,"console_offset":62,"console_lines":10} +{"time":"2026-08-14T22:21:10.538990403Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:21:25.405590498Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":36,"history_lines":1,"events_offset":73,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:21:25.560039008Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:21:40.406004609Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":1,"events_offset":75,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:21:40.520368533Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:21:55.40554939Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":38,"history_lines":1,"events_offset":77,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:21:55.510133609Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:22:10.406035723Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":79,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:22:10.564369093Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:22:25.405797515Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":1,"events_offset":81,"events_lines":2,"console_offset":72,"console_lines":10} +{"time":"2026-08-14T22:22:25.518466895Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:22:40.40573414Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":41,"history_lines":1,"events_offset":83,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:22:40.565407197Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:22:55.405279582Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":1,"events_offset":85,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:22:55.572896644Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:23:10.406123059Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":43,"history_lines":1,"events_offset":87,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:23:10.525921175Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:23:25.405432496Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":1,"events_offset":89,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:23:25.587919865Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:23:40.405843481Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":45,"history_lines":1,"events_offset":91,"events_lines":2,"console_offset":82,"console_lines":10} +{"time":"2026-08-14T22:23:40.555604866Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:23:55.405877105Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":93,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:23:55.513658618Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:24:10.405810362Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":1,"events_offset":95,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:24:10.565224461Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:24:25.406151072Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":48,"history_lines":1,"events_offset":97,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:24:25.552886878Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:24:40.405160685Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":1,"events_offset":99,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:24:40.52176699Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:24:55.406073053Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":50,"history_lines":2,"events_offset":101,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:24:55.539961111Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:24:56.042584292Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T22:24:56.042849717Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":52,"history_lines":1,"events_offset":103,"events_lines":1,"console_offset":92,"console_lines":18,"uploaded_len":3,"complete":true,"exit_code":0} +{"time":"2026-08-14T22:24:56.293989988Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:24:56.295208855Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T22:24:56.297829968Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T22:24:56.297843095Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T22:24:56.297914854Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T22:24:56.297920806Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug.log b/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..55bb1e21df451804a7baf677bdb05dffc5bac9a9 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug.log @@ -0,0 +1,24 @@ +2026-08-14 22:11:54,693 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():729] Logging user logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug.log +2026-08-14 22:11:54,693 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():730] Logging internal logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_221154-a3607q94/logs/debug-internal.log +2026-08-14 22:11:54,693 INFO MainThread:1229423 [wandb_init.py:init():772] calling init triggers +2026-08-14 22:11:54,693 INFO MainThread:1229423 [wandb_init.py:init():777] wandb.init called with sweep_config: {} +config: {'_wandb': {}} +2026-08-14 22:11:54,693 INFO MainThread:1229423 [wandb_init.py:init():820] starting backend +2026-08-14 22:11:54,693 INFO MainThread:1229423 [wandb_init.py:init():835] sending inform_init request +2026-08-14 22:11:54,957 INFO MainThread:1229423 [wandb_init.py:init():840] backend started and connected +2026-08-14 22:11:54,958 INFO MainThread:1229423 [wandb_init.py:init():910] updated telemetry +2026-08-14 22:11:54,965 INFO MainThread:1229423 [wandb_init.py:init():933] communicating run to backend with 90.0 second timeout +2026-08-14 22:11:55,322 INFO MainThread:1229423 [wandb_init.py:init():978] starting run threads in backend +2026-08-14 22:11:55,399 INFO MainThread:1229423 [wandb_run.py:_console_start():2621] atexit reg +2026-08-14 22:11:55,399 INFO MainThread:1229423 [wandb_run.py:_redirect():2471] redirect: wrap_raw +2026-08-14 22:11:55,399 INFO MainThread:1229423 [wandb_run.py:_redirect():2540] Wrapping output streams. +2026-08-14 22:11:55,399 INFO MainThread:1229423 [wandb_run.py:_redirect():2563] Redirects installed. +2026-08-14 22:11:55,400 INFO MainThread:1229423 [wandb_init.py:init():1016] run started, returning control to user process +2026-08-14 22:11:55,401 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 21, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'mlp', 'activation': 'silu-waleed10', 'waleed_beta': 10.0, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/mlp-silu-waleed10-21L_run', 'per_device_train_batch_size': 128, 'num_train_epochs': 1, 'max_steps': 1000, 'learning_rate': 0.001, 'lr_scheduler_type': 'constant', 'lr_scheduler_kwargs': None, 'warmup_steps': 0, 'optim': 'adamw_torch', 'optim_args': None, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 4, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-mlp-silu-waleed10-21L-4.0M-20260814-221154', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 60000, 'eval_delay': 0, 'per_device_eval_batch_size': 128, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 100, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/A-mlp-silu-waleed10-21L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} +2026-08-14 22:11:55,404 INFO MainThread:1229423 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 3970432 - > +2026-08-14 22:11:55,404 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 3970432 None +2026-08-14 22:24:55,661 INFO MainThread:1229423 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research1/a3607q94 +2026-08-14 22:24:55,662 INFO MainThread:1229423 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 22:24:55,662 INFO MainThread:1229423 [wandb_run.py:_restore():2570] restore +2026-08-14 22:24:55,662 INFO MainThread:1229423 [wandb_run.py:_restore():2576] restore done +2026-08-14 22:24:56,297 INFO MainThread:1229423 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_221154-a3607q94/run-a3607q94.wandb b/zain/Activation/wandb/run-20260814_221154-a3607q94/run-a3607q94.wandb new file mode 100644 index 0000000000000000000000000000000000000000..e089613426e80da4de3894eff456e514da62edeb --- /dev/null +++ b/zain/Activation/wandb/run-20260814_221154-a3607q94/run-a3607q94.wandb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fd675d3647f508d189770d90b8b24f154361d7cb4eb0a260d80b438c86d0d42b +size 574747 diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/config.yaml b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..620371c1dccd997597bb50a184b48a78500143f8 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/config.yaml @@ -0,0 +1,439 @@ +_name_or_path: + value: "" +_wandb: + value: + cli_version: 0.28.1 + e: + ry559cdhs8pu8maqeo36io5wxu2p8d4g: + args: + - --config + - /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml + - --variants + - glu-waleed10 + - glu-silu-waleed10 + - mlp-waleed10 + - mlp-silu-waleed10 + - glu-waleed + - glu-situglu_low + - glu-waleedglu_low + codePath: sweep.py + codePathLocal: sweep.py + cpu_count: 112 + cpu_count_logical: 224 + cudaVersion: "12.4" + disk: + /: + total: "1560765693952" + used: "709103636480" + email: deepnevro@gmail.com + executable: /mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python + git: + commit: 3fb47c6943d6927eafc39bd399a3d3a520bae74a + remote: https://github.com/w-ahmad1a10/Activation.git + gpu: NVIDIA H100 80GB HBM3 + gpu_count: 8 + gpu_nvidia: + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-39c684a5-fde6-83d7-1663-0859795881ae + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-132944c4-b689-2b5f-89a4-d730401677ab + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bc6c3e3c-9b90-09ca-c034-774961847c54 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea + host: deeplens-k3s-node1 + memory: + total: "2164089937920" + os: Linux-5.15.0-126-generic-x86_64-with-glibc2.35 + program: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py + python: CPython 3.11.15 + root: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation + startedAt: "2026-08-14T22:24:56.823342Z" + writerId: ry559cdhs8pu8maqeo36io5wxu2p8d4g + m: + - "1": train/global_step + "6": + - 3 + "7": [] + - "2": '*' + "5": 1 + "6": + - 1 + "7": [] + python_version: 3.11.15 + t: + "1": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "2": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "3": + - 2 + - 7 + - 13 + - 19 + - 62 + - 66 + "4": 3.11.15 + "5": 0.28.1 + "6": 5.16.0.dev0 + "9": + "1": transformers_trainer + "12": 0.28.1 + "13": linux-x86_64 +accelerator_config: + value: + dispatch_batches: null + even_batches: true + gradient_accumulation_kwargs: null + non_blocking: false + split_batches: false + use_seedable_sampler: true +activation: + value: waleed +adam_beta1: + value: 0.9 +adam_beta2: + value: 0.999 +adam_epsilon: + value: 1e-08 +architectures: + value: null +attention_bias: + value: false +attention_dropout: + value: 0 +auto_find_batch_size: + value: false +average_tokens_across_devices: + value: true +batch_eval_metrics: + value: false +bf16: + value: true +bf16_full_eval: + value: false +bos_token_id: + value: 1 +chunk_size_feed_forward: + value: 0 +data_seed: + value: 42 +dataloader_drop_last: + value: false +dataloader_in_order: + value: true +dataloader_multiprocessing_context: + value: null +dataloader_num_workers: + value: 0 +dataloader_persistent_workers: + value: false +dataloader_pin_memory: + value: true +dataloader_prefetch_factor: + value: null +ddp_backend: + value: null +ddp_broadcast_buffers: + value: null +ddp_bucket_cap_mb: + value: null +ddp_find_unused_parameters: + value: null +ddp_static_graph: + value: null +ddp_timeout: + value: 1800 +debug: + value: [] +deepspeed: + value: null +disable_tqdm: + value: false +do_eval: + value: true +do_predict: + value: false +do_train: + value: false +dtype: + value: null +enable_jit_checkpoint: + value: false +eos_token_id: + value: 2 +eval_accumulation_steps: + value: null +eval_delay: + value: 0 +eval_do_concat_batches: + value: true +eval_on_start: + value: false +eval_steps: + value: 60000 +eval_strategy: + value: steps +eval_use_gather_object: + value: false +fp16: + value: false +fp16_full_eval: + value: false +fsdp: + value: null +fsdp_config: + value: null +full_determinism: + value: false +gradient_accumulation_steps: + value: 4 +gradient_checkpointing: + value: false +gradient_checkpointing_kwargs: + value: null +greater_is_better: + value: null +head_dim: + value: 32 +hidden_act: + value: silu +hidden_size: + value: 128 +hub_always_push: + value: false +hub_model_id: + value: w-ahmad/A-glu-waleed-21L +hub_private_repo: + value: null +hub_revision: + value: null +hub_strategy: + value: every_save +hub_token: + value: +id2label: + value: + "0": LABEL_0 + "1": LABEL_1 +ignore_data_skip: + value: false +include_for_metrics: + value: [] +include_num_input_tokens_seen: + value: "no" +initializer_range: + value: 0.02 +intermediate_size: + value: 256 +is_encoder_decoder: + value: false +label_names: + value: null +label_smoothing_factor: + value: 0 +label2id: + value: + LABEL_0: 0 + LABEL_1: 1 +learning_rate: + value: 0.001 +length_column_name: + value: length +liger_kernel_config: + value: null +load_best_model_at_end: + value: false +local_rank: + value: -1 +log_level: + value: passive +log_level_replica: + value: warning +log_on_each_node: + value: true +logging_first_step: + value: false +logging_nan_inf_filter: + value: true +logging_steps: + value: 20 +logging_strategy: + value: steps +lr_scheduler_kwargs: + value: null +lr_scheduler_type: + value: constant +max_grad_norm: + value: 1 +max_position_embeddings: + value: 512 +max_steps: + value: 1000 +metric_for_best_model: + value: null +mlp_bias: + value: false +mlp_type: + value: glu +model/num_parameters: + value: 3970432 +model_type: + value: tiny_llama +neftune_noise_alpha: + value: null +num_attention_heads: + value: 4 +num_hidden_layers: + value: 21 +num_key_value_heads: + value: 4 +num_train_epochs: + value: 1 +optim: + value: adamw_torch +optim_args: + value: null +optim_target_modules: + value: null +output_attentions: + value: false +output_dir: + value: out/glu-waleed-21L_run +output_hidden_states: + value: false +pad_token_id: + value: 0 +parallelism_config: + value: null +per_device_eval_batch_size: + value: 128 +per_device_train_batch_size: + value: 128 +prediction_loss_only: + value: false +pretraining_tp: + value: 1 +problem_type: + value: null +project: + value: huggingface +push_to_hub: + value: false +remove_unused_columns: + value: false +report_to: + value: + - wandb +restore_callback_states_from_checkpoint: + value: false +resume_from_checkpoint: + value: null +return_dict: + value: true +rms_norm_eps: + value: 1e-06 +rope_parameters: + value: + rope_theta: 10000 + rope_type: default +run_name: + value: LM-glu-waleed-21L-4.0M-20260814-222456 +save_on_each_node: + value: false +save_only_model: + value: false +save_steps: + value: 100 +save_strategy: + value: steps +save_total_limit: + value: null +seed: + value: 42 +skip_memory_metrics: + value: true +tf32: + value: null +tie_word_embeddings: + value: true +tokenizer_name: + value: w-ahmad/tiny-stories-tokenizer +torch_compile: + value: false +torch_compile_backend: + value: null +torch_compile_mode: + value: null +torch_empty_cache_steps: + value: null +trackio_bucket_id: + value: null +trackio_space_id: + value: null +trackio_static_space_id: + value: null +train_sampling_strategy: + value: random +transformers_version: + value: 5.16.0.dev0 +use_cache: + value: false +use_cpu: + value: false +use_liger_kernel: + value: false +vocab_size: + value: 4096 +waleed_beta: + value: 10 +warmup_steps: + value: 0 +weight_decay: + value: 0.01 diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/output.log b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/output.log new file mode 100644 index 0000000000000000000000000000000000000000..ea8ab309d27962827c1307d0ca251923ea22cee5 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/output.log @@ -0,0 +1,110 @@ + 10%|█ | 100/1000 [01:19<11:47, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '28.88', 'grad_norm': '4', 'learning_rate': '0.001', 'epoch': '0.01078', 'train/total_time_seconds': '8.163', 'train/time_per_step_avg': '0.4082', 'train/epoch_time_elapsed': '15.97', 'train/estimated_remaining_minutes': '6.667'} +{'loss': '24.01', 'grad_norm': '1.969', 'learning_rate': '0.001', 'epoch': '0.02157', 'train/total_time_seconds': '16.29', 'train/time_per_step_avg': '0.4071', 'train/epoch_time_elapsed': '31.88', 'train/estimated_remaining_minutes': '6.514'} +{'loss': '22.25', 'grad_norm': '1.992', 'learning_rate': '0.001', 'epoch': '0.03235', 'train/total_time_seconds': '24.42', 'train/time_per_step_avg': '0.407', 'train/epoch_time_elapsed': '47.66', 'train/estimated_remaining_minutes': '6.376'} +{'loss': '20.39', 'grad_norm': '1.625', 'learning_rate': '0.001', 'epoch': '0.04314', 'train/total_time_seconds': '32.55', 'train/time_per_step_avg': '0.4069', 'train/epoch_time_elapsed': '63.47', 'train/estimated_remaining_minutes': '6.239'} +{'loss': '18.54', 'grad_norm': '2.062', 'learning_rate': '0.001', 'epoch': '0.05392', 'train/total_time_seconds': '40.67', 'train/time_per_step_avg': '0.4067', 'train/epoch_time_elapsed': '79.22', 'train/estimated_remaining_minutes': '6.1'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 75.75it/s] + 20%|██ | 200/1000 [02:38<10:30, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '17.23', 'grad_norm': '2.094', 'learning_rate': '0.001', 'epoch': '0.06471', 'train/total_time_seconds': '48.8', 'train/time_per_step_avg': '0.4064', 'train/epoch_time_elapsed': '95.07', 'train/estimated_remaining_minutes': '5.964'} +{'loss': '16.2', 'grad_norm': '2.938', 'learning_rate': '0.001', 'epoch': '0.07549', 'train/total_time_seconds': '56.96', 'train/time_per_step_avg': '0.4067', 'train/epoch_time_elapsed': '110.9', 'train/estimated_remaining_minutes': '5.831'} +{'loss': '15.42', 'grad_norm': '1.352', 'learning_rate': '0.001', 'epoch': '0.08628', 'train/total_time_seconds': '65.09', 'train/time_per_step_avg': '0.4067', 'train/epoch_time_elapsed': '126.6', 'train/estimated_remaining_minutes': '5.695'} +{'loss': '14.78', 'grad_norm': '2.297', 'learning_rate': '0.001', 'epoch': '0.09706', 'train/total_time_seconds': '73.21', 'train/time_per_step_avg': '0.4065', 'train/epoch_time_elapsed': '142.4', 'train/estimated_remaining_minutes': '5.558'} +{'loss': '14.2', 'grad_norm': '2.234', 'learning_rate': '0.001', 'epoch': '0.1078', 'train/total_time_seconds': '81.33', 'train/time_per_step_avg': '0.4066', 'train/epoch_time_elapsed': '158.2', 'train/estimated_remaining_minutes': '5.422'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 76.92it/s] + 30%|███ | 300/1000 [03:57<09:04, 1.29it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '13.72', 'grad_norm': '2.484', 'learning_rate': '0.001', 'epoch': '0.1186', 'train/total_time_seconds': '89.45', 'train/time_per_step_avg': '0.4066', 'train/epoch_time_elapsed': '174.1', 'train/estimated_remaining_minutes': '5.286'} +{'loss': '13.31', 'grad_norm': '1.75', 'learning_rate': '0.001', 'epoch': '0.1294', 'train/total_time_seconds': '97.59', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '189.9', 'train/estimated_remaining_minutes': '5.151'} +{'loss': '12.89', 'grad_norm': '2.547', 'learning_rate': '0.001', 'epoch': '0.1402', 'train/total_time_seconds': '105.7', 'train/time_per_step_avg': '0.4064', 'train/epoch_time_elapsed': '205.8', 'train/estimated_remaining_minutes': '5.015'} +{'loss': '12.52', 'grad_norm': '2.438', 'learning_rate': '0.001', 'epoch': '0.151', 'train/total_time_seconds': '113.8', 'train/time_per_step_avg': '0.4064', 'train/epoch_time_elapsed': '221.6', 'train/estimated_remaining_minutes': '4.879'} +{'loss': '12.18', 'grad_norm': '2.547', 'learning_rate': '0.001', 'epoch': '0.1618', 'train/total_time_seconds': '121.9', 'train/time_per_step_avg': '0.4062', 'train/epoch_time_elapsed': '237.4', 'train/estimated_remaining_minutes': '4.742'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.43it/s] + 40%|████ | 400/1000 [05:16<07:52, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '11.86', 'grad_norm': '2.094', 'learning_rate': '0.001', 'epoch': '0.1726', 'train/total_time_seconds': '130.1', 'train/time_per_step_avg': '0.406', 'train/epoch_time_elapsed': '253.1', 'train/estimated_remaining_minutes': '4.606'} +{'loss': '11.58', 'grad_norm': '1.992', 'learning_rate': '0.001', 'epoch': '0.1833', 'train/total_time_seconds': '138.1', 'train/time_per_step_avg': '0.4055', 'train/epoch_time_elapsed': '268.9', 'train/estimated_remaining_minutes': '4.469'} +{'loss': '11.32', 'grad_norm': '2.188', 'learning_rate': '0.001', 'epoch': '0.1941', 'train/total_time_seconds': '146.2', 'train/time_per_step_avg': '0.4051', 'train/epoch_time_elapsed': '284.6', 'train/estimated_remaining_minutes': '4.333'} +{'loss': '11.07', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.2049', 'train/total_time_seconds': '154.4', 'train/time_per_step_avg': '0.4051', 'train/epoch_time_elapsed': '300.4', 'train/estimated_remaining_minutes': '4.198'} +{'loss': '10.86', 'grad_norm': '2.234', 'learning_rate': '0.001', 'epoch': '0.2157', 'train/total_time_seconds': '162.5', 'train/time_per_step_avg': '0.4053', 'train/epoch_time_elapsed': '316.2', 'train/estimated_remaining_minutes': '4.062'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 74.34it/s] + 50%|█████ | 500/1000 [06:35<06:30, 1.28it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.65', 'grad_norm': '2.516', 'learning_rate': '0.001', 'epoch': '0.2265', 'train/total_time_seconds': '170.6', 'train/time_per_step_avg': '0.4052', 'train/epoch_time_elapsed': '332.1', 'train/estimated_remaining_minutes': '3.926'} +{'loss': '10.53', 'grad_norm': '2.656', 'learning_rate': '0.001', 'epoch': '0.2373', 'train/total_time_seconds': '178.7', 'train/time_per_step_avg': '0.4053', 'train/epoch_time_elapsed': '347.9', 'train/estimated_remaining_minutes': '3.79'} +{'loss': '10.34', 'grad_norm': '2.016', 'learning_rate': '0.001', 'epoch': '0.248', 'train/total_time_seconds': '186.8', 'train/time_per_step_avg': '0.4055', 'train/epoch_time_elapsed': '363.8', 'train/estimated_remaining_minutes': '3.655'} +{'loss': '10.18', 'grad_norm': '2.281', 'learning_rate': '0.001', 'epoch': '0.2588', 'train/total_time_seconds': '194.9', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '379.7', 'train/estimated_remaining_minutes': '3.519'} +{'loss': '10.04', 'grad_norm': '2.422', 'learning_rate': '0.001', 'epoch': '0.2696', 'train/total_time_seconds': '203.1', 'train/time_per_step_avg': '0.4058', 'train/epoch_time_elapsed': '395.4', 'train/estimated_remaining_minutes': '3.384'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 41.75it/s] + 60%|██████ | 600/1000 [07:54<05:19, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.927', 'grad_norm': '1.977', 'learning_rate': '0.001', 'epoch': '0.2804', 'train/total_time_seconds': '211.2', 'train/time_per_step_avg': '0.4062', 'train/epoch_time_elapsed': '411.3', 'train/estimated_remaining_minutes': '3.249'} +{'loss': '9.82', 'grad_norm': '2.172', 'learning_rate': '0.001', 'epoch': '0.2912', 'train/total_time_seconds': '219.3', 'train/time_per_step_avg': '0.4066', 'train/epoch_time_elapsed': '427.1', 'train/estimated_remaining_minutes': '3.114'} +{'loss': '9.708', 'grad_norm': '2.594', 'learning_rate': '0.001', 'epoch': '0.302', 'train/total_time_seconds': '227.5', 'train/time_per_step_avg': '0.4066', 'train/epoch_time_elapsed': '442.8', 'train/estimated_remaining_minutes': '2.979'} +{'loss': '9.597', 'grad_norm': '2.109', 'learning_rate': '0.001', 'epoch': '0.3128', 'train/total_time_seconds': '235.5', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '458.5', 'train/estimated_remaining_minutes': '2.843'} +{'loss': '9.496', 'grad_norm': '2.547', 'learning_rate': '0.001', 'epoch': '0.3235', 'train/total_time_seconds': '243.6', 'train/time_per_step_avg': '0.4059', 'train/epoch_time_elapsed': '474.5', 'train/estimated_remaining_minutes': '2.707'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 80.61it/s] + 70%|███████ | 700/1000 [09:13<03:55, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.41', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.3343', 'train/total_time_seconds': '251.7', 'train/time_per_step_avg': '0.4053', 'train/epoch_time_elapsed': '490.4', 'train/estimated_remaining_minutes': '2.571'} +{'loss': '9.316', 'grad_norm': '2.312', 'learning_rate': '0.001', 'epoch': '0.3451', 'train/total_time_seconds': '259.8', 'train/time_per_step_avg': '0.405', 'train/epoch_time_elapsed': '506.1', 'train/estimated_remaining_minutes': '2.436'} +{'loss': '9.233', 'grad_norm': '1.945', 'learning_rate': '0.001', 'epoch': '0.3559', 'train/total_time_seconds': '268', 'train/time_per_step_avg': '0.4051', 'train/epoch_time_elapsed': '522', 'train/estimated_remaining_minutes': '2.301'} +{'loss': '9.158', 'grad_norm': '2.125', 'learning_rate': '0.001', 'epoch': '0.3667', 'train/total_time_seconds': '276.1', 'train/time_per_step_avg': '0.4054', 'train/epoch_time_elapsed': '537.9', 'train/estimated_remaining_minutes': '2.165'} +{'loss': '9.085', 'grad_norm': '2.094', 'learning_rate': '0.001', 'epoch': '0.3775', 'train/total_time_seconds': '284.2', 'train/time_per_step_avg': '0.4058', 'train/epoch_time_elapsed': '553.8', 'train/estimated_remaining_minutes': '2.03'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 78.59it/s] + 80%|████████ | 800/1000 [10:33<02:41, 1.24it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.995', 'grad_norm': '1.977', 'learning_rate': '0.001', 'epoch': '0.3882', 'train/total_time_seconds': '292.4', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '569.8', 'train/estimated_remaining_minutes': '1.895'} +{'loss': '8.963', 'grad_norm': '2', 'learning_rate': '0.001', 'epoch': '0.399', 'train/total_time_seconds': '300.5', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '585.6', 'train/estimated_remaining_minutes': '1.759'} +{'loss': '8.899', 'grad_norm': '1.867', 'learning_rate': '0.001', 'epoch': '0.4098', 'train/total_time_seconds': '308.5', 'train/time_per_step_avg': '0.4059', 'train/epoch_time_elapsed': '601.6', 'train/estimated_remaining_minutes': '1.624'} +{'loss': '8.858', 'grad_norm': '1.891', 'learning_rate': '0.001', 'epoch': '0.4206', 'train/total_time_seconds': '316.6', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '617.4', 'train/estimated_remaining_minutes': '1.489'} +{'loss': '8.76', 'grad_norm': '2.109', 'learning_rate': '0.001', 'epoch': '0.4314', 'train/total_time_seconds': '324.7', 'train/time_per_step_avg': '0.4052', 'train/epoch_time_elapsed': '633.3', 'train/estimated_remaining_minutes': '1.353'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.93it/s] + 90%|█████████ | 900/1000 [11:52<01:19, 1.26it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.735', 'grad_norm': '2.594', 'learning_rate': '0.001', 'epoch': '0.4422', 'train/total_time_seconds': '332.8', 'train/time_per_step_avg': '0.4048', 'train/epoch_time_elapsed': '649.1', 'train/estimated_remaining_minutes': '1.218'} +{'loss': '8.695', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.453', 'train/total_time_seconds': '340.9', 'train/time_per_step_avg': '0.4049', 'train/epoch_time_elapsed': '664.9', 'train/estimated_remaining_minutes': '1.082'} +{'loss': '8.626', 'grad_norm': '1.992', 'learning_rate': '0.001', 'epoch': '0.4637', 'train/total_time_seconds': '349.1', 'train/time_per_step_avg': '0.4051', 'train/epoch_time_elapsed': '680.6', 'train/estimated_remaining_minutes': '0.947'} +{'loss': '8.586', 'grad_norm': '1.922', 'learning_rate': '0.001', 'epoch': '0.4745', 'train/total_time_seconds': '357.2', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '696.5', 'train/estimated_remaining_minutes': '0.8118'} +{'loss': '8.538', 'grad_norm': '2.062', 'learning_rate': '0.001', 'epoch': '0.4853', 'train/total_time_seconds': '365.3', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '712.3', 'train/estimated_remaining_minutes': '0.6765'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 65.03it/s] +100%|██████████| 1000/1000 [13:21<00:00, 1.27i[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.492', 'grad_norm': '1.922', 'learning_rate': '0.001', 'epoch': '0.4961', 'train/total_time_seconds': '373.4', 'train/time_per_step_avg': '0.4057', 'train/epoch_time_elapsed': '728.2', 'train/estimated_remaining_minutes': '0.5412'} +{'loss': '8.474', 'grad_norm': '1.891', 'learning_rate': '0.001', 'epoch': '0.5069', 'train/total_time_seconds': '381.5', 'train/time_per_step_avg': '0.4059', 'train/epoch_time_elapsed': '743.9', 'train/estimated_remaining_minutes': '0.4059'} +{'loss': '8.452', 'grad_norm': '1.844', 'learning_rate': '0.001', 'epoch': '0.5177', 'train/total_time_seconds': '389.7', 'train/time_per_step_avg': '0.406', 'train/epoch_time_elapsed': '759.8', 'train/estimated_remaining_minutes': '0.2706'} +{'loss': '8.417', 'grad_norm': '1.891', 'learning_rate': '0.001', 'epoch': '0.5284', 'train/total_time_seconds': '397.8', 'train/time_per_step_avg': '0.4058', 'train/epoch_time_elapsed': '775.7', 'train/estimated_remaining_minutes': '0.1353'} +{'loss': '8.381', 'grad_norm': '1.844', 'learning_rate': '0.001', 'epoch': '0.5392', 'train/total_time_seconds': '405.9', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '791.6', 'train/estimated_remaining_minutes': '0'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes +{'eval_loss': '2.094', 'eval_runtime': '9.59', 'eval_samples_per_second': '993.4', 'eval_steps_per_second': '7.821', 'epoch': '0.5392', 'train/total_time_seconds': '405.9', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '801.2', 'train/estimated_remaining_minutes': '0'} + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 76.03it/s] +100%|██████████| 1000/1000 [13:21<00:00, 1.25it/s], ?it/s] +{'train_runtime': '802.1', 'train_samples_per_second': '638.3', 'train_steps_per_second': '1.247', 'train_loss': '11.79', 'epoch': '0.5392', 'train/total_time_seconds': '405.9', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '801.2', 'train/estimated_remaining_minutes': '0'} +100%|██████████| 75/75 [00:09<00:00, 7.85it/s] +[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.14it/s] +>>> FINISHED glu-waleed-21L successfully diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/requirements.txt b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..f36b19895172d9637438a445ceacb71d09dc5d02 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/requirements.txt @@ -0,0 +1,149 @@ +asttokens==3.0.1 +comm==0.2.3 +debugpy==1.8.21 +decorator==5.3.1 +executing==2.2.1 +nest-asyncio==1.6.0 +parso==0.8.7 +platformdirs==4.11.0 +psutil==7.2.2 +ptyprocess==0.7.0 +pure_eval==0.2.3 +Pygments==2.20.0 +pyzmq==27.1.0 +setuptools==83.0.0 +six==1.17.0 +tornado==6.5.7 +traitlets==5.15.0 +fsspec==2026.4.0 +wcwidth==0.8.2 +ipython_pygments_lexers==1.1.1 +jedi==0.20.0 +jupyter_core==5.9.1 +matplotlib-inline==0.2.2 +pexpect==4.9.0 +prompt_toolkit==3.0.53 +python-dateutil==2.9.0.post0 +stack_data==0.6.3 +wheel==0.47.0 +jupyter_client==8.9.1 +pip==26.1.2 +ipython==9.15.0 +ipykernel==7.2.0 +threadpoolctl==3.6.0 +pyparsing==3.3.2 +typing_extensions==4.15.0 +Jinja2==3.1.6 +narwhals==2.24.0 +kiwisolver==1.5.0 +joblib==1.5.3 +fonttools==4.63.0 +cycler==0.12.1 +scipy==1.17.1 +pandas==3.0.5 +contourpy==1.3.3 +scikit-learn==1.9.0 +matplotlib==3.11.1 +urllib3==2.7.0 +tqdm==4.70.0 +idna==3.18 +charset-normalizer==3.4.9 +certifi==2026.7.22 +requests==2.34.2 +seaborn==0.13.2 +uv==0.12.0 +shellingham==1.5.4 +mpmath==1.3.0 +attrs==26.1.0 +hf-xet==1.5.2 +nvidia-nccl-cu12==2.21.5 +MarkupSafe==3.0.3 +regex==2026.7.19 +importlib_metadata==9.0.0 +httpcore==1.0.9 +annotated-doc==0.0.5 +multidict==6.7.1 +aiohttp==3.14.3 +aiosignal==1.4.0 +xxhash==3.8.1 +aiohappyeyeballs==2.7.1 +mdurl==0.1.2 +cuda-toolkit==13.0.3.0 +networkx==3.6.1 +PyYAML==6.0.3 +nvidia-cufile==1.15.1.6 +typer==0.27.0 +torchaudio==2.6.0+cu124 +rich==15.0.0 +nvidia-cufft-cu12==11.2.1.3 +h11==0.16.0 +dill==0.4.1 +cuda-pathfinder==1.6.0 +filelock==3.29.0 +nvidia-nvtx-cu12==12.4.127 +httpx==0.28.1 +anyio==4.14.2 +numpy==2.4.4 +yarl==1.24.5 +click==8.4.2 +triton==3.2.0 +frozenlist==1.8.0 +zipp==4.1.0 +propcache==0.5.2 +tokenizers==0.22.2 +markdown-it-py==4.2.0 +nvidia-cuda-runtime==13.0.96 +cuda-bindings==13.3.1 +nvidia-cuda-cupti==13.0.85 +torch==2.6.0+cu124 +multiprocess==0.70.19 +pillow==12.2.0 +transformers==5.16.0.dev0 +wandb==0.28.1 +nvidia-curand==10.4.0.35 +sympy==1.13.1 +nvidia-cusparse==12.6.3.3 +nvidia-cuda-nvrtc==13.0.88 +typing-inspection==0.4.2 +nvidia-cusolver==12.0.4.66 +nvidia-cufft==12.0.0.61 +nvidia-cudnn-cu13==9.20.0.48 +nvidia-cublas==13.1.1.3 +pyarrow==25.0.0 +evaluate==0.4.6 +diffusers==0.39.0 +pydantic==2.13.4 +annotated-types==0.8.0 +protobuf==7.35.1 +sentry-sdk==2.66.1 +einops==0.8.2 +packaging==26.2 +nvidia-nvjitlink-cu12==12.4.127 +nvidia-curand-cu12==10.3.5.147 +nvidia-cusparselt-cu12==0.6.2 +nvidia-cusparse-cu12==12.3.1.170 +nvidia-cuda-runtime-cu12==12.4.127 +torchvision==0.21.0+cu124 +nvidia-cuda-nvrtc-cu12==12.4.127 +nvidia-cuda-cupti-cu12==12.4.127 +nvidia-cusolver-cu12==11.6.1.9 +nvidia-cublas-cu12==12.4.5.8 +nvidia-cudnn-cu12==9.1.0.70 +huggingface_hub==1.26.0 +datasets==5.0.1 +safetensors==0.8.0 +accelerate==1.14.0 +pydantic_core==2.46.4 +ninja==1.13.0 +autocommand==2.2.2 +backports.tarfile==1.2.0 +importlib_metadata==8.7.1 +jaraco.text==4.0.0 +jaraco.context==6.1.0 +jaraco.functools==4.4.0 +more-itertools==10.8.0 +packaging==26.0 +platformdirs==4.4.0 +tomli==2.4.0 +wheel==0.46.3 +zipp==3.23.0 diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/wandb-metadata.json b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/wandb-metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..2fa5d6f35ae9b0d7fbfc0fb261dbaf38942c6540 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/wandb-metadata.json @@ -0,0 +1,101 @@ +{ + "os": "Linux-5.15.0-126-generic-x86_64-with-glibc2.35", + "python": "CPython 3.11.15", + "startedAt": "2026-08-14T22:24:56.823342Z", + "args": [ + "--config", + "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml", + "--variants", + "glu-waleed10", + "glu-silu-waleed10", + "mlp-waleed10", + "mlp-silu-waleed10", + "glu-waleed", + "glu-situglu_low", + "glu-waleedglu_low" + ], + "program": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py", + "codePath": "sweep.py", + "codePathLocal": "sweep.py", + "git": { + "remote": "https://github.com/w-ahmad1a10/Activation.git", + "commit": "3fb47c6943d6927eafc39bd399a3d3a520bae74a" + }, + "email": "deepnevro@gmail.com", + "root": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation", + "host": "deeplens-k3s-node1", + "executable": "/mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python", + "cpu_count": 112, + "cpu_count_logical": 224, + "gpu": "NVIDIA H100 80GB HBM3", + "gpu_count": 8, + "disk": { + "/": { + "total": "1560765693952", + "used": "709103636480" + } + }, + "memory": { + "total": "2164089937920" + }, + "gpu_nvidia": [ + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-39c684a5-fde6-83d7-1663-0859795881ae" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-132944c4-b689-2b5f-89a4-d730401677ab" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bc6c3e3c-9b90-09ca-c034-774961847c54" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea" + } + ], + "cudaVersion": "12.4", + "writerId": "ry559cdhs8pu8maqeo36io5wxu2p8d4g" +} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..6692a5e6b9cf35ad32f06cae1d14d499dc60039c --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/files/wandb-summary.json @@ -0,0 +1 @@ +{"train/loss":8.381436920166015,"_timestamp":1.7867471085870068e+09,"eval/steps_per_second":7.683,"_step":52,"train/grad_norm":1.84375,"train_samples_per_second":638.348,"_wandb":{"runtime":811},"train_runtime":802.0703,"train_loss":11.791388671875,"_runtime":811,"train/global_step":1000,"eval/loss":2.0943470001220703,"total_flos":5.420315836416e+15,"eval/samples_per_second":975.891,"train/epoch":0.5392289026691831,"eval/runtime":9.7624,"train_steps_per_second":1.247,"train/learning_rate":0.001} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug-core.log b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug-core.log new file mode 100644 index 0000000000000000000000000000000000000000..588056df10bc8cabd5e2682287953b3e2b130ffe --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug-core.log @@ -0,0 +1,62 @@ +{"time":"2026-08-14T21:31:53.539152008Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp650y3sc2/port-1229423.txt","pid":1229423,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false} +{"time":"2026-08-14T21:31:53.540676876Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1229423} +{"time":"2026-08-14T21:31:53.540659035Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T21:31:53.717724383Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"} +{"time":"2026-08-14T21:31:53.802092564Z","level":"INFO","msg":"handleInformInit: received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:54.062397411Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:59.411924652Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.396635188Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.929834053Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.931620539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:48.932491197Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:49.464388506Z","level":"INFO","msg":"handleInformInit: received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:49.733853364Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:55.329792921Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.158537572Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.914569559Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.916156477Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:01.917534874Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:02.524868682Z","level":"INFO","msg":"handleInformInit: received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:02.786199875Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:08.192858634Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:53.643379161Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.173579992Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.17490129Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.175479186Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.694802218Z","level":"INFO","msg":"handleInformInit: received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:11:54.956642682Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:12:00.441973218Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:55.662446346Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.296327526Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.297810539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.29856977Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.82560196Z","level":"INFO","msg":"handleInformInit: received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:24:57.085772231Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:25:02.571913681Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:28.612623637Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.208727987Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.210106262Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.210705504Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.839594174Z","level":"INFO","msg":"handleInformInit: received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:30.099444184Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:35.620666929Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:17.704425638Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.35448297Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.355720686Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.356578583Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.899799881Z","level":"INFO","msg":"handleInformInit: received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:19.159494071Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:24.671083701Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.214192173Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.783107435Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.784410941Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:51.78516667Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088192277Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088229648Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088239651Z","level":"INFO","msg":"connection: closing","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088243873Z","level":"INFO","msg":"server: is shutting down"} +{"time":"2026-08-14T23:05:52.0882852Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088299095Z","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088345917Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088452051Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T23:05:52.088520571Z","level":"INFO","msg":"server: all connections closed"} diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug-internal.log new file mode 100644 index 0000000000000000000000000000000000000000..f38621df7f4fa96fdc92d2368f797cac1d33c352 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug-internal.log @@ -0,0 +1,125 @@ +{"time":"2026-08-14T22:24:56.825723698Z","level":"INFO","msg":"wandb-core"} +{"time":"2026-08-14T22:24:56.825860434Z","level":"INFO","msg":"stream: starting","core version":"0.28.1"} +{"time":"2026-08-14T22:24:57.085596625Z","level":"INFO","msg":"stream: created new stream","id":"wcvy5vyw"} +{"time":"2026-08-14T22:24:57.085675681Z","level":"INFO","msg":"handler: started"} +{"time":"2026-08-14T22:24:57.085765252Z","level":"INFO","msg":"stream: started"} +{"time":"2026-08-14T22:24:57.085777062Z","level":"INFO","msg":"writer: started","stream_id":"wcvy5vyw"} +{"time":"2026-08-14T22:24:57.085800553Z","level":"INFO","msg":"sender: started"} +{"time":"2026-08-14T22:24:57.57390664Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:24:57.689803675Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:25:12.574037646Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":0,"events_lines":1,"console_offset":0,"console_lines":1,"uploaded_len":2} +{"time":"2026-08-14T22:25:12.716448176Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:25:27.574260399Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":0,"history_lines":1,"events_offset":1,"events_lines":2,"console_offset":0,"console_lines":2} +{"time":"2026-08-14T22:25:27.69115135Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:25:42.57432798Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":1,"history_lines":1,"events_offset":3,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:25:42.701645665Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:25:57.574852945Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":2,"history_lines":1,"events_offset":5,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:25:57.694991424Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:26:12.574234616Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":3,"history_lines":1,"events_offset":7,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:26:12.702125322Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:26:27.574556836Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":4,"history_lines":1,"events_offset":9,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:26:27.713695135Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:26:42.574592485Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":5,"history_lines":1,"events_offset":11,"events_lines":2,"console_offset":2,"console_lines":10} +{"time":"2026-08-14T22:26:42.717599899Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:26:57.574987828Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":6,"history_lines":1,"events_offset":13,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:26:57.697000807Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:27:12.574348283Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":7,"history_lines":1,"events_offset":15,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:27:12.700155262Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:27:27.574749814Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":8,"history_lines":1,"events_offset":17,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:27:27.728833877Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:27:42.574886541Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":9,"history_lines":1,"events_offset":19,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:27:42.725241152Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:27:57.574361574Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":10,"history_lines":1,"events_offset":21,"events_lines":2,"console_offset":12,"console_lines":10} +{"time":"2026-08-14T22:27:57.708700673Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:28:12.574682815Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":1,"events_offset":23,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:28:12.674665971Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:28:27.574667399Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":12,"history_lines":1,"events_offset":25,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:28:27.740964744Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:28:42.574290372Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":1,"events_offset":27,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:28:43.234719315Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:28:57.574853814Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":14,"history_lines":1,"events_offset":29,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:28:57.694726394Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:29:12.574668999Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":15,"history_lines":1,"events_offset":31,"events_lines":2,"console_offset":22,"console_lines":10} +{"time":"2026-08-14T22:29:12.693938985Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:29:27.574781248Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:29:27.684922113Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:29:42.574792655Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":17,"history_lines":1,"events_offset":35,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:29:42.702364027Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:29:57.574828034Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":37,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:29:57.682234649Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:30:12.574145436Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":18,"history_lines":1,"events_offset":39,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:30:12.78125588Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:30:27.574409495Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:30:27.700286132Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:30:42.574579334Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":20,"history_lines":1,"events_offset":43,"events_lines":2,"console_offset":32,"console_lines":10} +{"time":"2026-08-14T22:30:42.708397219Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:30:57.574936618Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:30:57.677385895Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:31:12.574052907Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":22,"history_lines":1,"events_offset":47,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:31:12.698626832Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:31:27.57497536Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":23,"history_lines":1,"events_offset":49,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:31:27.711296606Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:31:42.574875419Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":1,"events_offset":51,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:31:42.686140815Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:31:57.574985842Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":25,"history_lines":1,"events_offset":53,"events_lines":2,"console_offset":42,"console_lines":10} +{"time":"2026-08-14T22:31:57.692228972Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:32:12.574537455Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":1,"events_offset":55,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:32:12.72236182Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:32:27.574897977Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":27,"history_lines":1,"events_offset":57,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:32:27.698118542Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:32:42.574300653Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":28,"history_lines":1,"events_offset":59,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:32:42.690560187Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:32:57.575083653Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":1,"events_offset":61,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:32:57.720449984Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:33:12.574474659Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":30,"history_lines":1,"events_offset":63,"events_lines":2,"console_offset":52,"console_lines":10} +{"time":"2026-08-14T22:33:12.713825792Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:33:27.574130185Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":1,"events_offset":65,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:33:27.727843221Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:33:42.57400035Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":32,"history_lines":1,"events_offset":67,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:33:42.678772321Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:33:57.574860667Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":1,"events_offset":69,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:33:57.703764077Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:34:12.574978134Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":34,"history_lines":1,"events_offset":71,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:34:12.695794566Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:34:27.574270423Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":1,"events_offset":73,"events_lines":2,"console_offset":62,"console_lines":10} +{"time":"2026-08-14T22:34:27.687851697Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:34:42.574471733Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":75,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:34:42.678295361Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:34:57.574760136Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":36,"history_lines":1,"events_offset":77,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:34:57.695664984Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:35:12.574728541Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":1,"events_offset":79,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:35:12.697377259Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:35:27.574941286Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":38,"history_lines":1,"events_offset":81,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:35:27.682936495Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:35:42.574971781Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":83,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:35:42.714357465Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:35:57.574263028Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":1,"events_offset":85,"events_lines":2,"console_offset":72,"console_lines":10} +{"time":"2026-08-14T22:35:57.708124425Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:36:12.574958811Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":41,"history_lines":1,"events_offset":87,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:36:12.697898169Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:36:27.574130752Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":1,"events_offset":89,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:36:27.700264798Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:36:42.574300051Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":43,"history_lines":1,"events_offset":91,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:36:42.709508515Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:36:57.574652978Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":1,"events_offset":93,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:36:57.697730869Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:37:12.574917991Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":45,"history_lines":1,"events_offset":95,"events_lines":2,"console_offset":82,"console_lines":10} +{"time":"2026-08-14T22:37:12.735095383Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:37:27.574633025Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":97,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:37:27.698944697Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:37:42.574415424Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":1,"events_offset":99,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:37:42.734335436Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:37:57.574721565Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":48,"history_lines":1,"events_offset":101,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:37:57.698777829Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:38:12.574423276Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":1,"events_offset":103,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:38:12.775133791Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:38:27.574276913Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":50,"history_lines":2,"events_offset":105,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:38:27.693401254Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:38:29.060944776Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T22:38:29.061209871Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":52,"history_lines":1,"events_offset":107,"events_lines":1,"console_offset":92,"console_lines":18,"uploaded_len":3,"complete":true,"exit_code":0} +{"time":"2026-08-14T22:38:29.206456668Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:38:29.207654339Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T22:38:29.210126011Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T22:38:29.210150844Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T22:38:29.210267138Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T22:38:29.210270442Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug.log b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..a36f65ce0159523a06ef1c79027f24b6e521d4ad --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug.log @@ -0,0 +1,24 @@ +2026-08-14 22:24:56,824 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():729] Logging user logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug.log +2026-08-14 22:24:56,824 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():730] Logging internal logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/logs/debug-internal.log +2026-08-14 22:24:56,824 INFO MainThread:1229423 [wandb_init.py:init():772] calling init triggers +2026-08-14 22:24:56,824 INFO MainThread:1229423 [wandb_init.py:init():777] wandb.init called with sweep_config: {} +config: {'_wandb': {}} +2026-08-14 22:24:56,824 INFO MainThread:1229423 [wandb_init.py:init():820] starting backend +2026-08-14 22:24:56,824 INFO MainThread:1229423 [wandb_init.py:init():835] sending inform_init request +2026-08-14 22:24:57,086 INFO MainThread:1229423 [wandb_init.py:init():840] backend started and connected +2026-08-14 22:24:57,087 INFO MainThread:1229423 [wandb_init.py:init():910] updated telemetry +2026-08-14 22:24:57,094 INFO MainThread:1229423 [wandb_init.py:init():933] communicating run to backend with 90.0 second timeout +2026-08-14 22:24:57,491 INFO MainThread:1229423 [wandb_init.py:init():978] starting run threads in backend +2026-08-14 22:24:57,568 INFO MainThread:1229423 [wandb_run.py:_console_start():2621] atexit reg +2026-08-14 22:24:57,568 INFO MainThread:1229423 [wandb_run.py:_redirect():2471] redirect: wrap_raw +2026-08-14 22:24:57,568 INFO MainThread:1229423 [wandb_run.py:_redirect():2540] Wrapping output streams. +2026-08-14 22:24:57,568 INFO MainThread:1229423 [wandb_run.py:_redirect():2563] Redirects installed. +2026-08-14 22:24:57,569 INFO MainThread:1229423 [wandb_init.py:init():1016] run started, returning control to user process +2026-08-14 22:24:57,570 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 21, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'glu', 'activation': 'waleed', 'waleed_beta': 10.0, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/glu-waleed-21L_run', 'per_device_train_batch_size': 128, 'num_train_epochs': 1, 'max_steps': 1000, 'learning_rate': 0.001, 'lr_scheduler_type': 'constant', 'lr_scheduler_kwargs': None, 'warmup_steps': 0, 'optim': 'adamw_torch', 'optim_args': None, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 4, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-glu-waleed-21L-4.0M-20260814-222456', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 60000, 'eval_delay': 0, 'per_device_eval_batch_size': 128, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 100, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/A-glu-waleed-21L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} +2026-08-14 22:24:57,572 INFO MainThread:1229423 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 3970432 - > +2026-08-14 22:24:57,573 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 3970432 None +2026-08-14 22:38:28,611 INFO MainThread:1229423 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research1/wcvy5vyw +2026-08-14 22:38:28,612 INFO MainThread:1229423 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 22:38:28,612 INFO MainThread:1229423 [wandb_run.py:_restore():2570] restore +2026-08-14 22:38:28,612 INFO MainThread:1229423 [wandb_run.py:_restore():2576] restore done +2026-08-14 22:38:29,209 INFO MainThread:1229423 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/run-wcvy5vyw.wandb b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/run-wcvy5vyw.wandb new file mode 100644 index 0000000000000000000000000000000000000000..c32c67c12c3653e1e1d674220674fe1cc7a42f24 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_222456-wcvy5vyw/run-wcvy5vyw.wandb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44111704e4ab704fe722ebd03f8efcda1d13ee243a9c99a238efa4fa43363850 +size 580518 diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/config.yaml b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..9f9348d912bcbbdf1dec62e029cfd41ed33dedc0 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/config.yaml @@ -0,0 +1,439 @@ +_name_or_path: + value: "" +_wandb: + value: + cli_version: 0.28.1 + e: + ld9jgt7guvta0zype38auuu3dri4f4ao: + args: + - --config + - /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml + - --variants + - glu-waleed10 + - glu-silu-waleed10 + - mlp-waleed10 + - mlp-silu-waleed10 + - glu-waleed + - glu-situglu_low + - glu-waleedglu_low + codePath: sweep.py + codePathLocal: sweep.py + cpu_count: 112 + cpu_count_logical: 224 + cudaVersion: "12.4" + disk: + /: + total: "1560765693952" + used: "709102354432" + email: deepnevro@gmail.com + executable: /mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python + git: + commit: 3fb47c6943d6927eafc39bd399a3d3a520bae74a + remote: https://github.com/w-ahmad1a10/Activation.git + gpu: NVIDIA H100 80GB HBM3 + gpu_count: 8 + gpu_nvidia: + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-39c684a5-fde6-83d7-1663-0859795881ae + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-132944c4-b689-2b5f-89a4-d730401677ab + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bc6c3e3c-9b90-09ca-c034-774961847c54 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea + host: deeplens-k3s-node1 + memory: + total: "2164089937920" + os: Linux-5.15.0-126-generic-x86_64-with-glibc2.35 + program: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py + python: CPython 3.11.15 + root: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation + startedAt: "2026-08-14T22:38:29.837436Z" + writerId: ld9jgt7guvta0zype38auuu3dri4f4ao + m: + - "1": train/global_step + "6": + - 3 + "7": [] + - "2": '*' + "5": 1 + "6": + - 1 + "7": [] + python_version: 3.11.15 + t: + "1": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "2": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "3": + - 2 + - 7 + - 13 + - 19 + - 62 + - 66 + "4": 3.11.15 + "5": 0.28.1 + "6": 5.16.0.dev0 + "9": + "1": transformers_trainer + "12": 0.28.1 + "13": linux-x86_64 +accelerator_config: + value: + dispatch_batches: null + even_batches: true + gradient_accumulation_kwargs: null + non_blocking: false + split_batches: false + use_seedable_sampler: true +activation: + value: situglu_low +adam_beta1: + value: 0.9 +adam_beta2: + value: 0.999 +adam_epsilon: + value: 1e-08 +architectures: + value: null +attention_bias: + value: false +attention_dropout: + value: 0 +auto_find_batch_size: + value: false +average_tokens_across_devices: + value: true +batch_eval_metrics: + value: false +bf16: + value: true +bf16_full_eval: + value: false +bos_token_id: + value: 1 +chunk_size_feed_forward: + value: 0 +data_seed: + value: 42 +dataloader_drop_last: + value: false +dataloader_in_order: + value: true +dataloader_multiprocessing_context: + value: null +dataloader_num_workers: + value: 0 +dataloader_persistent_workers: + value: false +dataloader_pin_memory: + value: true +dataloader_prefetch_factor: + value: null +ddp_backend: + value: null +ddp_broadcast_buffers: + value: null +ddp_bucket_cap_mb: + value: null +ddp_find_unused_parameters: + value: null +ddp_static_graph: + value: null +ddp_timeout: + value: 1800 +debug: + value: [] +deepspeed: + value: null +disable_tqdm: + value: false +do_eval: + value: true +do_predict: + value: false +do_train: + value: false +dtype: + value: null +enable_jit_checkpoint: + value: false +eos_token_id: + value: 2 +eval_accumulation_steps: + value: null +eval_delay: + value: 0 +eval_do_concat_batches: + value: true +eval_on_start: + value: false +eval_steps: + value: 60000 +eval_strategy: + value: steps +eval_use_gather_object: + value: false +fp16: + value: false +fp16_full_eval: + value: false +fsdp: + value: null +fsdp_config: + value: null +full_determinism: + value: false +gradient_accumulation_steps: + value: 4 +gradient_checkpointing: + value: false +gradient_checkpointing_kwargs: + value: null +greater_is_better: + value: null +head_dim: + value: 32 +hidden_act: + value: silu +hidden_size: + value: 128 +hub_always_push: + value: false +hub_model_id: + value: w-ahmad/A-glu-situglu_low-21L +hub_private_repo: + value: null +hub_revision: + value: null +hub_strategy: + value: every_save +hub_token: + value: +id2label: + value: + "0": LABEL_0 + "1": LABEL_1 +ignore_data_skip: + value: false +include_for_metrics: + value: [] +include_num_input_tokens_seen: + value: "no" +initializer_range: + value: 0.02 +intermediate_size: + value: 256 +is_encoder_decoder: + value: false +label_names: + value: null +label_smoothing_factor: + value: 0 +label2id: + value: + LABEL_0: 0 + LABEL_1: 1 +learning_rate: + value: 0.001 +length_column_name: + value: length +liger_kernel_config: + value: null +load_best_model_at_end: + value: false +local_rank: + value: -1 +log_level: + value: passive +log_level_replica: + value: warning +log_on_each_node: + value: true +logging_first_step: + value: false +logging_nan_inf_filter: + value: true +logging_steps: + value: 20 +logging_strategy: + value: steps +lr_scheduler_kwargs: + value: null +lr_scheduler_type: + value: constant +max_grad_norm: + value: 1 +max_position_embeddings: + value: 512 +max_steps: + value: 1000 +metric_for_best_model: + value: null +mlp_bias: + value: false +mlp_type: + value: glu +model/num_parameters: + value: 3970432 +model_type: + value: tiny_llama +neftune_noise_alpha: + value: null +num_attention_heads: + value: 4 +num_hidden_layers: + value: 21 +num_key_value_heads: + value: 4 +num_train_epochs: + value: 1 +optim: + value: adamw_torch +optim_args: + value: null +optim_target_modules: + value: null +output_attentions: + value: false +output_dir: + value: out/glu-situglu_low-21L_run +output_hidden_states: + value: false +pad_token_id: + value: 0 +parallelism_config: + value: null +per_device_eval_batch_size: + value: 128 +per_device_train_batch_size: + value: 128 +prediction_loss_only: + value: false +pretraining_tp: + value: 1 +problem_type: + value: null +project: + value: huggingface +push_to_hub: + value: false +remove_unused_columns: + value: false +report_to: + value: + - wandb +restore_callback_states_from_checkpoint: + value: false +resume_from_checkpoint: + value: null +return_dict: + value: true +rms_norm_eps: + value: 1e-06 +rope_parameters: + value: + rope_theta: 10000 + rope_type: default +run_name: + value: LM-glu-situglu_low-21L-4.0M-20260814-223829 +save_on_each_node: + value: false +save_only_model: + value: false +save_steps: + value: 100 +save_strategy: + value: steps +save_total_limit: + value: null +seed: + value: 42 +skip_memory_metrics: + value: true +tf32: + value: null +tie_word_embeddings: + value: true +tokenizer_name: + value: w-ahmad/tiny-stories-tokenizer +torch_compile: + value: false +torch_compile_backend: + value: null +torch_compile_mode: + value: null +torch_empty_cache_steps: + value: null +trackio_bucket_id: + value: null +trackio_space_id: + value: null +trackio_static_space_id: + value: null +train_sampling_strategy: + value: random +transformers_version: + value: 5.16.0.dev0 +use_cache: + value: false +use_cpu: + value: false +use_liger_kernel: + value: false +vocab_size: + value: 4096 +waleed_beta: + value: 10 +warmup_steps: + value: 0 +weight_decay: + value: 0.01 diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/output.log b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/output.log new file mode 100644 index 0000000000000000000000000000000000000000..efdcb9acddff101fe62222aa9d22a138b00ecf42 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/output.log @@ -0,0 +1,110 @@ + 10%|█ | 100/1000 [01:20<11:58, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '28.93', 'grad_norm': '4.031', 'learning_rate': '0.001', 'epoch': '0.01078', 'train/total_time_seconds': '8.487', 'train/time_per_step_avg': '0.4243', 'train/epoch_time_elapsed': '16.23', 'train/estimated_remaining_minutes': '6.931'} +{'loss': '24.27', 'grad_norm': '1.391', 'learning_rate': '0.001', 'epoch': '0.02157', 'train/total_time_seconds': '16.92', 'train/time_per_step_avg': '0.4231', 'train/epoch_time_elapsed': '32.28', 'train/estimated_remaining_minutes': '6.77'} +{'loss': '22.78', 'grad_norm': '3.031', 'learning_rate': '0.001', 'epoch': '0.03235', 'train/total_time_seconds': '25.4', 'train/time_per_step_avg': '0.4233', 'train/epoch_time_elapsed': '48.38', 'train/estimated_remaining_minutes': '6.631'} +{'loss': '21.39', 'grad_norm': '3.656', 'learning_rate': '0.001', 'epoch': '0.04314', 'train/total_time_seconds': '33.87', 'train/time_per_step_avg': '0.4233', 'train/epoch_time_elapsed': '64.52', 'train/estimated_remaining_minutes': '6.491'} +{'loss': '19.97', 'grad_norm': '4.531', 'learning_rate': '0.001', 'epoch': '0.05392', 'train/total_time_seconds': '42.32', 'train/time_per_step_avg': '0.4232', 'train/epoch_time_elapsed': '80.6', 'train/estimated_remaining_minutes': '6.348'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 67.06it/s] + 20%|██ | 200/1000 [02:41<10:41, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '18.74', 'grad_norm': '4.906', 'learning_rate': '0.001', 'epoch': '0.06471', 'train/total_time_seconds': '50.77', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '97.07', 'train/estimated_remaining_minutes': '6.206'} +{'loss': '17.7', 'grad_norm': '4.344', 'learning_rate': '0.001', 'epoch': '0.07549', 'train/total_time_seconds': '59.22', 'train/time_per_step_avg': '0.423', 'train/epoch_time_elapsed': '113.1', 'train/estimated_remaining_minutes': '6.063'} +{'loss': '16.74', 'grad_norm': '3.688', 'learning_rate': '0.001', 'epoch': '0.08628', 'train/total_time_seconds': '67.69', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '129.3', 'train/estimated_remaining_minutes': '5.923'} +{'loss': '15.89', 'grad_norm': '2.484', 'learning_rate': '0.001', 'epoch': '0.09706', 'train/total_time_seconds': '76.16', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '145.3', 'train/estimated_remaining_minutes': '5.782'} +{'loss': '15.18', 'grad_norm': '4.188', 'learning_rate': '0.001', 'epoch': '0.1078', 'train/total_time_seconds': '84.63', 'train/time_per_step_avg': '0.4231', 'train/epoch_time_elapsed': '161.5', 'train/estimated_remaining_minutes': '5.642'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 74.34it/s] + 30%|███ | 300/1000 [04:01<09:16, 1.26it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '14.59', 'grad_norm': '2.047', 'learning_rate': '0.001', 'epoch': '0.1186', 'train/total_time_seconds': '93.1', 'train/time_per_step_avg': '0.4232', 'train/epoch_time_elapsed': '177.6', 'train/estimated_remaining_minutes': '5.501'} +{'loss': '14.07', 'grad_norm': '1.797', 'learning_rate': '0.001', 'epoch': '0.1294', 'train/total_time_seconds': '101.6', 'train/time_per_step_avg': '0.4233', 'train/epoch_time_elapsed': '193.8', 'train/estimated_remaining_minutes': '5.36'} +{'loss': '13.55', 'grad_norm': '2.547', 'learning_rate': '0.001', 'epoch': '0.1402', 'train/total_time_seconds': '110', 'train/time_per_step_avg': '0.4235', 'train/epoch_time_elapsed': '209.8', 'train/estimated_remaining_minutes': '5.22'} +{'loss': '13.13', 'grad_norm': '2.516', 'learning_rate': '0.001', 'epoch': '0.151', 'train/total_time_seconds': '118.5', 'train/time_per_step_avg': '0.4236', 'train/epoch_time_elapsed': '225.9', 'train/estimated_remaining_minutes': '5.079'} +{'loss': '12.77', 'grad_norm': '2.078', 'learning_rate': '0.001', 'epoch': '0.1618', 'train/total_time_seconds': '127', 'train/time_per_step_avg': '0.4234', 'train/epoch_time_elapsed': '241.8', 'train/estimated_remaining_minutes': '4.938'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 58.58it/s] + 40%|████ | 400/1000 [05:22<08:01, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '12.41', 'grad_norm': '2.484', 'learning_rate': '0.001', 'epoch': '0.1726', 'train/total_time_seconds': '135.4', 'train/time_per_step_avg': '0.4233', 'train/epoch_time_elapsed': '258.3', 'train/estimated_remaining_minutes': '4.796'} +{'loss': '12.1', 'grad_norm': '2.328', 'learning_rate': '0.001', 'epoch': '0.1833', 'train/total_time_seconds': '143.9', 'train/time_per_step_avg': '0.4234', 'train/epoch_time_elapsed': '274.5', 'train/estimated_remaining_minutes': '4.655'} +{'loss': '11.79', 'grad_norm': '1.906', 'learning_rate': '0.001', 'epoch': '0.1941', 'train/total_time_seconds': '152.3', 'train/time_per_step_avg': '0.423', 'train/epoch_time_elapsed': '290.6', 'train/estimated_remaining_minutes': '4.514'} +{'loss': '11.49', 'grad_norm': '2.203', 'learning_rate': '0.001', 'epoch': '0.2049', 'train/total_time_seconds': '160.8', 'train/time_per_step_avg': '0.4227', 'train/epoch_time_elapsed': '306.7', 'train/estimated_remaining_minutes': '4.372'} +{'loss': '11.26', 'grad_norm': '2.219', 'learning_rate': '0.001', 'epoch': '0.2157', 'train/total_time_seconds': '169.3', 'train/time_per_step_avg': '0.4231', 'train/epoch_time_elapsed': '322.9', 'train/estimated_remaining_minutes': '4.232'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 75.33it/s] + 50%|█████ | 500/1000 [06:43<06:40, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '11.03', 'grad_norm': '2.266', 'learning_rate': '0.001', 'epoch': '0.2265', 'train/total_time_seconds': '177.8', 'train/time_per_step_avg': '0.4233', 'train/epoch_time_elapsed': '339', 'train/estimated_remaining_minutes': '4.091'} +{'loss': '10.88', 'grad_norm': '2.5', 'learning_rate': '0.001', 'epoch': '0.2373', 'train/total_time_seconds': '186.2', 'train/time_per_step_avg': '0.4232', 'train/epoch_time_elapsed': '355.1', 'train/estimated_remaining_minutes': '3.95'} +{'loss': '10.67', 'grad_norm': '2.578', 'learning_rate': '0.001', 'epoch': '0.248', 'train/total_time_seconds': '194.7', 'train/time_per_step_avg': '0.4233', 'train/epoch_time_elapsed': '371.2', 'train/estimated_remaining_minutes': '3.809'} +{'loss': '10.49', 'grad_norm': '2.438', 'learning_rate': '0.001', 'epoch': '0.2588', 'train/total_time_seconds': '203.1', 'train/time_per_step_avg': '0.4235', 'train/epoch_time_elapsed': '387.3', 'train/estimated_remaining_minutes': '3.668'} +{'loss': '10.32', 'grad_norm': '1.992', 'learning_rate': '0.001', 'epoch': '0.2696', 'train/total_time_seconds': '211.6', 'train/time_per_step_avg': '0.4234', 'train/epoch_time_elapsed': '403.3', 'train/estimated_remaining_minutes': '3.527'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 78.63it/s] + 60%|██████ | 600/1000 [08:03<05:20, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.19', 'grad_norm': '1.891', 'learning_rate': '0.001', 'epoch': '0.2804', 'train/total_time_seconds': '220.1', 'train/time_per_step_avg': '0.4232', 'train/epoch_time_elapsed': '419.5', 'train/estimated_remaining_minutes': '3.386'} +{'loss': '10.06', 'grad_norm': '2.484', 'learning_rate': '0.001', 'epoch': '0.2912', 'train/total_time_seconds': '228.6', 'train/time_per_step_avg': '0.4235', 'train/epoch_time_elapsed': '435.6', 'train/estimated_remaining_minutes': '3.245'} +{'loss': '9.931', 'grad_norm': '2.266', 'learning_rate': '0.001', 'epoch': '0.302', 'train/total_time_seconds': '237.1', 'train/time_per_step_avg': '0.4239', 'train/epoch_time_elapsed': '451.7', 'train/estimated_remaining_minutes': '3.104'} +{'loss': '9.802', 'grad_norm': '2.344', 'learning_rate': '0.001', 'epoch': '0.3128', 'train/total_time_seconds': '245.5', 'train/time_per_step_avg': '0.4239', 'train/epoch_time_elapsed': '467.7', 'train/estimated_remaining_minutes': '2.963'} +{'loss': '9.69', 'grad_norm': '1.898', 'learning_rate': '0.001', 'epoch': '0.3235', 'train/total_time_seconds': '254', 'train/time_per_step_avg': '0.4239', 'train/epoch_time_elapsed': '483.7', 'train/estimated_remaining_minutes': '2.822'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 78.12it/s] + 70%|███████ | 700/1000 [09:24<04:03, 1.23it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.586', 'grad_norm': '2.172', 'learning_rate': '0.001', 'epoch': '0.3343', 'train/total_time_seconds': '262.5', 'train/time_per_step_avg': '0.424', 'train/epoch_time_elapsed': '499.9', 'train/estimated_remaining_minutes': '2.681'} +{'loss': '9.485', 'grad_norm': '2.703', 'learning_rate': '0.001', 'epoch': '0.3451', 'train/total_time_seconds': '271', 'train/time_per_step_avg': '0.4239', 'train/epoch_time_elapsed': '515.9', 'train/estimated_remaining_minutes': '2.54'} +{'loss': '9.387', 'grad_norm': '1.75', 'learning_rate': '0.001', 'epoch': '0.3559', 'train/total_time_seconds': '279.4', 'train/time_per_step_avg': '0.4236', 'train/epoch_time_elapsed': '532.1', 'train/estimated_remaining_minutes': '2.399'} +{'loss': '9.298', 'grad_norm': '2.25', 'learning_rate': '0.001', 'epoch': '0.3667', 'train/total_time_seconds': '287.9', 'train/time_per_step_avg': '0.4234', 'train/epoch_time_elapsed': '548.2', 'train/estimated_remaining_minutes': '2.258'} +{'loss': '9.215', 'grad_norm': '1.922', 'learning_rate': '0.001', 'epoch': '0.3775', 'train/total_time_seconds': '296.3', 'train/time_per_step_avg': '0.4232', 'train/epoch_time_elapsed': '564.5', 'train/estimated_remaining_minutes': '2.117'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 59.58it/s] + 80%|████████ | 800/1000 [10:45<02:39, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.112', 'grad_norm': '2.203', 'learning_rate': '0.001', 'epoch': '0.3882', 'train/total_time_seconds': '304.8', 'train/time_per_step_avg': '0.423', 'train/epoch_time_elapsed': '580.6', 'train/estimated_remaining_minutes': '1.975'} +{'loss': '9.064', 'grad_norm': '2.062', 'learning_rate': '0.001', 'epoch': '0.399', 'train/total_time_seconds': '313.2', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '596.8', 'train/estimated_remaining_minutes': '1.834'} +{'loss': '8.99', 'grad_norm': '2.484', 'learning_rate': '0.001', 'epoch': '0.4098', 'train/total_time_seconds': '321.7', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '612.9', 'train/estimated_remaining_minutes': '1.693'} +{'loss': '8.943', 'grad_norm': '2.062', 'learning_rate': '0.001', 'epoch': '0.4206', 'train/total_time_seconds': '330.2', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '629.3', 'train/estimated_remaining_minutes': '1.552'} +{'loss': '8.839', 'grad_norm': '2.031', 'learning_rate': '0.001', 'epoch': '0.4314', 'train/total_time_seconds': '338.6', 'train/time_per_step_avg': '0.4227', 'train/epoch_time_elapsed': '645.3', 'train/estimated_remaining_minutes': '1.411'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 81.83it/s] + 90%|█████████ | 900/1000 [12:06<01:21, 1.23it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.803', 'grad_norm': '1.961', 'learning_rate': '0.001', 'epoch': '0.4422', 'train/total_time_seconds': '347.1', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '661.6', 'train/estimated_remaining_minutes': '1.27'} +{'loss': '8.757', 'grad_norm': '2.062', 'learning_rate': '0.001', 'epoch': '0.453', 'train/total_time_seconds': '355.5', 'train/time_per_step_avg': '0.4229', 'train/epoch_time_elapsed': '677.8', 'train/estimated_remaining_minutes': '1.129'} +{'loss': '8.683', 'grad_norm': '1.664', 'learning_rate': '0.001', 'epoch': '0.4637', 'train/total_time_seconds': '364', 'train/time_per_step_avg': '0.4228', 'train/epoch_time_elapsed': '694', 'train/estimated_remaining_minutes': '0.9876'} +{'loss': '8.638', 'grad_norm': '2.188', 'learning_rate': '0.001', 'epoch': '0.4745', 'train/total_time_seconds': '372.4', 'train/time_per_step_avg': '0.4226', 'train/epoch_time_elapsed': '710.1', 'train/estimated_remaining_minutes': '0.8464'} +{'loss': '8.583', 'grad_norm': '1.906', 'learning_rate': '0.001', 'epoch': '0.4853', 'train/total_time_seconds': '380.9', 'train/time_per_step_avg': '0.4231', 'train/epoch_time_elapsed': '726.4', 'train/estimated_remaining_minutes': '0.7054'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 75.88it/s] +100%|██████████| 1000/1000 [13:37<00:00, 1.23i[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.531', 'grad_norm': '2.406', 'learning_rate': '0.001', 'epoch': '0.4961', 'train/total_time_seconds': '389.4', 'train/time_per_step_avg': '0.4233', 'train/epoch_time_elapsed': '742.7', 'train/estimated_remaining_minutes': '0.5643'} +{'loss': '8.513', 'grad_norm': '2.062', 'learning_rate': '0.001', 'epoch': '0.5069', 'train/total_time_seconds': '397.9', 'train/time_per_step_avg': '0.4236', 'train/epoch_time_elapsed': '758.9', 'train/estimated_remaining_minutes': '0.4233'} +{'loss': '8.484', 'grad_norm': '2.047', 'learning_rate': '0.001', 'epoch': '0.5177', 'train/total_time_seconds': '406.4', 'train/time_per_step_avg': '0.4239', 'train/epoch_time_elapsed': '775.1', 'train/estimated_remaining_minutes': '0.2822'} +{'loss': '8.444', 'grad_norm': '1.883', 'learning_rate': '0.001', 'epoch': '0.5284', 'train/total_time_seconds': '414.8', 'train/time_per_step_avg': '0.4243', 'train/epoch_time_elapsed': '791.2', 'train/estimated_remaining_minutes': '0.1411'} +{'loss': '8.405', 'grad_norm': '1.695', 'learning_rate': '0.001', 'epoch': '0.5392', 'train/total_time_seconds': '423.4', 'train/time_per_step_avg': '0.4244', 'train/epoch_time_elapsed': '807.4', 'train/estimated_remaining_minutes': '0'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes +{'eval_loss': '2.097', 'eval_runtime': '9.902', 'eval_samples_per_second': '962.1', 'eval_steps_per_second': '7.574', 'epoch': '0.5392', 'train/total_time_seconds': '423.4', 'train/time_per_step_avg': '0.4244', 'train/epoch_time_elapsed': '817.3', 'train/estimated_remaining_minutes': '0'} + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 67.31it/s] +100%|██████████| 1000/1000 [13:37<00:00, 1.22it/s], ?it/s] +{'train_runtime': '818.2', 'train_samples_per_second': '625.8', 'train_steps_per_second': '1.222', 'train_loss': '12.19', 'epoch': '0.5392', 'train/total_time_seconds': '423.4', 'train/time_per_step_avg': '0.4244', 'train/epoch_time_elapsed': '817.3', 'train/estimated_remaining_minutes': '0'} +100%|██████████| 75/75 [00:09<00:00, 7.90it/s] +[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 55.66it/s] +>>> FINISHED glu-situglu_low-21L successfully diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/requirements.txt b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..f36b19895172d9637438a445ceacb71d09dc5d02 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/requirements.txt @@ -0,0 +1,149 @@ +asttokens==3.0.1 +comm==0.2.3 +debugpy==1.8.21 +decorator==5.3.1 +executing==2.2.1 +nest-asyncio==1.6.0 +parso==0.8.7 +platformdirs==4.11.0 +psutil==7.2.2 +ptyprocess==0.7.0 +pure_eval==0.2.3 +Pygments==2.20.0 +pyzmq==27.1.0 +setuptools==83.0.0 +six==1.17.0 +tornado==6.5.7 +traitlets==5.15.0 +fsspec==2026.4.0 +wcwidth==0.8.2 +ipython_pygments_lexers==1.1.1 +jedi==0.20.0 +jupyter_core==5.9.1 +matplotlib-inline==0.2.2 +pexpect==4.9.0 +prompt_toolkit==3.0.53 +python-dateutil==2.9.0.post0 +stack_data==0.6.3 +wheel==0.47.0 +jupyter_client==8.9.1 +pip==26.1.2 +ipython==9.15.0 +ipykernel==7.2.0 +threadpoolctl==3.6.0 +pyparsing==3.3.2 +typing_extensions==4.15.0 +Jinja2==3.1.6 +narwhals==2.24.0 +kiwisolver==1.5.0 +joblib==1.5.3 +fonttools==4.63.0 +cycler==0.12.1 +scipy==1.17.1 +pandas==3.0.5 +contourpy==1.3.3 +scikit-learn==1.9.0 +matplotlib==3.11.1 +urllib3==2.7.0 +tqdm==4.70.0 +idna==3.18 +charset-normalizer==3.4.9 +certifi==2026.7.22 +requests==2.34.2 +seaborn==0.13.2 +uv==0.12.0 +shellingham==1.5.4 +mpmath==1.3.0 +attrs==26.1.0 +hf-xet==1.5.2 +nvidia-nccl-cu12==2.21.5 +MarkupSafe==3.0.3 +regex==2026.7.19 +importlib_metadata==9.0.0 +httpcore==1.0.9 +annotated-doc==0.0.5 +multidict==6.7.1 +aiohttp==3.14.3 +aiosignal==1.4.0 +xxhash==3.8.1 +aiohappyeyeballs==2.7.1 +mdurl==0.1.2 +cuda-toolkit==13.0.3.0 +networkx==3.6.1 +PyYAML==6.0.3 +nvidia-cufile==1.15.1.6 +typer==0.27.0 +torchaudio==2.6.0+cu124 +rich==15.0.0 +nvidia-cufft-cu12==11.2.1.3 +h11==0.16.0 +dill==0.4.1 +cuda-pathfinder==1.6.0 +filelock==3.29.0 +nvidia-nvtx-cu12==12.4.127 +httpx==0.28.1 +anyio==4.14.2 +numpy==2.4.4 +yarl==1.24.5 +click==8.4.2 +triton==3.2.0 +frozenlist==1.8.0 +zipp==4.1.0 +propcache==0.5.2 +tokenizers==0.22.2 +markdown-it-py==4.2.0 +nvidia-cuda-runtime==13.0.96 +cuda-bindings==13.3.1 +nvidia-cuda-cupti==13.0.85 +torch==2.6.0+cu124 +multiprocess==0.70.19 +pillow==12.2.0 +transformers==5.16.0.dev0 +wandb==0.28.1 +nvidia-curand==10.4.0.35 +sympy==1.13.1 +nvidia-cusparse==12.6.3.3 +nvidia-cuda-nvrtc==13.0.88 +typing-inspection==0.4.2 +nvidia-cusolver==12.0.4.66 +nvidia-cufft==12.0.0.61 +nvidia-cudnn-cu13==9.20.0.48 +nvidia-cublas==13.1.1.3 +pyarrow==25.0.0 +evaluate==0.4.6 +diffusers==0.39.0 +pydantic==2.13.4 +annotated-types==0.8.0 +protobuf==7.35.1 +sentry-sdk==2.66.1 +einops==0.8.2 +packaging==26.2 +nvidia-nvjitlink-cu12==12.4.127 +nvidia-curand-cu12==10.3.5.147 +nvidia-cusparselt-cu12==0.6.2 +nvidia-cusparse-cu12==12.3.1.170 +nvidia-cuda-runtime-cu12==12.4.127 +torchvision==0.21.0+cu124 +nvidia-cuda-nvrtc-cu12==12.4.127 +nvidia-cuda-cupti-cu12==12.4.127 +nvidia-cusolver-cu12==11.6.1.9 +nvidia-cublas-cu12==12.4.5.8 +nvidia-cudnn-cu12==9.1.0.70 +huggingface_hub==1.26.0 +datasets==5.0.1 +safetensors==0.8.0 +accelerate==1.14.0 +pydantic_core==2.46.4 +ninja==1.13.0 +autocommand==2.2.2 +backports.tarfile==1.2.0 +importlib_metadata==8.7.1 +jaraco.text==4.0.0 +jaraco.context==6.1.0 +jaraco.functools==4.4.0 +more-itertools==10.8.0 +packaging==26.0 +platformdirs==4.4.0 +tomli==2.4.0 +wheel==0.46.3 +zipp==3.23.0 diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/wandb-metadata.json b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/wandb-metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..9775cffd4ab8ca1f1b2d8a25324c201b5694db66 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/wandb-metadata.json @@ -0,0 +1,101 @@ +{ + "os": "Linux-5.15.0-126-generic-x86_64-with-glibc2.35", + "python": "CPython 3.11.15", + "startedAt": "2026-08-14T22:38:29.837436Z", + "args": [ + "--config", + "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml", + "--variants", + "glu-waleed10", + "glu-silu-waleed10", + "mlp-waleed10", + "mlp-silu-waleed10", + "glu-waleed", + "glu-situglu_low", + "glu-waleedglu_low" + ], + "program": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py", + "codePath": "sweep.py", + "codePathLocal": "sweep.py", + "git": { + "remote": "https://github.com/w-ahmad1a10/Activation.git", + "commit": "3fb47c6943d6927eafc39bd399a3d3a520bae74a" + }, + "email": "deepnevro@gmail.com", + "root": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation", + "host": "deeplens-k3s-node1", + "executable": "/mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python", + "cpu_count": 112, + "cpu_count_logical": 224, + "gpu": "NVIDIA H100 80GB HBM3", + "gpu_count": 8, + "disk": { + "/": { + "total": "1560765693952", + "used": "709102354432" + } + }, + "memory": { + "total": "2164089937920" + }, + "gpu_nvidia": [ + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-39c684a5-fde6-83d7-1663-0859795881ae" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-132944c4-b689-2b5f-89a4-d730401677ab" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bc6c3e3c-9b90-09ca-c034-774961847c54" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea" + } + ], + "cudaVersion": "12.4", + "writerId": "ld9jgt7guvta0zype38auuu3dri4f4ao" +} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..2b7b8916b1b170466ed9b352455df44740786c15 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/files/wandb-summary.json @@ -0,0 +1 @@ +{"_runtime":827,"eval/samples_per_second":981.954,"eval/loss":2.097445487976074,"train/loss":8.405266571044923,"_step":52,"train/learning_rate":0.001,"eval/runtime":9.7021,"train_steps_per_second":1.222,"_timestamp":1.7867479376740093e+09,"total_flos":5.420315836416e+15,"train/grad_norm":1.6953125,"train_runtime":818.1885,"train_loss":12.191291275024414,"train/global_step":1000,"train_samples_per_second":625.773,"train/epoch":0.5392289026691831,"eval/steps_per_second":7.73,"_wandb":{"runtime":827}} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug-core.log b/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug-core.log new file mode 100644 index 0000000000000000000000000000000000000000..588056df10bc8cabd5e2682287953b3e2b130ffe --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug-core.log @@ -0,0 +1,62 @@ +{"time":"2026-08-14T21:31:53.539152008Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp650y3sc2/port-1229423.txt","pid":1229423,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false} +{"time":"2026-08-14T21:31:53.540676876Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1229423} +{"time":"2026-08-14T21:31:53.540659035Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T21:31:53.717724383Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"} +{"time":"2026-08-14T21:31:53.802092564Z","level":"INFO","msg":"handleInformInit: received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:54.062397411Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:59.411924652Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.396635188Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.929834053Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.931620539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:48.932491197Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:49.464388506Z","level":"INFO","msg":"handleInformInit: received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:49.733853364Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:55.329792921Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.158537572Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.914569559Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.916156477Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:01.917534874Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:02.524868682Z","level":"INFO","msg":"handleInformInit: received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:02.786199875Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:08.192858634Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:53.643379161Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.173579992Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.17490129Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.175479186Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.694802218Z","level":"INFO","msg":"handleInformInit: received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:11:54.956642682Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:12:00.441973218Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:55.662446346Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.296327526Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.297810539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.29856977Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.82560196Z","level":"INFO","msg":"handleInformInit: received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:24:57.085772231Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:25:02.571913681Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:28.612623637Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.208727987Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.210106262Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.210705504Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.839594174Z","level":"INFO","msg":"handleInformInit: received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:30.099444184Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:35.620666929Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:17.704425638Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.35448297Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.355720686Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.356578583Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.899799881Z","level":"INFO","msg":"handleInformInit: received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:19.159494071Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:24.671083701Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.214192173Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.783107435Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.784410941Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:51.78516667Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088192277Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088229648Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088239651Z","level":"INFO","msg":"connection: closing","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088243873Z","level":"INFO","msg":"server: is shutting down"} +{"time":"2026-08-14T23:05:52.0882852Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088299095Z","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088345917Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088452051Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T23:05:52.088520571Z","level":"INFO","msg":"server: all connections closed"} diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug-internal.log new file mode 100644 index 0000000000000000000000000000000000000000..b84a4b6a35c4dd566185535934bdbbaa3e844350 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug-internal.log @@ -0,0 +1,127 @@ +{"time":"2026-08-14T22:38:29.839753806Z","level":"INFO","msg":"wandb-core"} +{"time":"2026-08-14T22:38:29.839881641Z","level":"INFO","msg":"stream: starting","core version":"0.28.1"} +{"time":"2026-08-14T22:38:30.099280653Z","level":"INFO","msg":"stream: created new stream","id":"uuv4h936"} +{"time":"2026-08-14T22:38:30.09934599Z","level":"INFO","msg":"handler: started"} +{"time":"2026-08-14T22:38:30.099436781Z","level":"INFO","msg":"stream: started"} +{"time":"2026-08-14T22:38:30.099446444Z","level":"INFO","msg":"writer: started","stream_id":"uuv4h936"} +{"time":"2026-08-14T22:38:30.099465321Z","level":"INFO","msg":"sender: started"} +{"time":"2026-08-14T22:38:30.622852598Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:38:30.727919945Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:38:45.623132075Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":0,"events_lines":1,"console_offset":0,"console_lines":1,"uploaded_len":2} +{"time":"2026-08-14T22:38:45.763643583Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:39:00.623186691Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":0,"history_lines":1,"events_offset":1,"events_lines":2,"console_offset":0,"console_lines":2} +{"time":"2026-08-14T22:39:00.73451107Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:39:15.62366638Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":1,"history_lines":1,"events_offset":3,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:39:15.750212811Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:39:30.623127087Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":2,"history_lines":1,"events_offset":5,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:39:30.743073481Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:39:45.623836933Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":3,"history_lines":1,"events_offset":7,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:39:45.741828658Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:40:00.622993325Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":4,"history_lines":1,"events_offset":9,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:40:00.748389964Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:40:15.623785947Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":5,"history_lines":1,"events_offset":11,"events_lines":2,"console_offset":2,"console_lines":10} +{"time":"2026-08-14T22:40:15.764323129Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:40:30.624711484Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":6,"history_lines":1,"events_offset":13,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:40:30.754962596Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:40:45.623161281Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":7,"history_lines":1,"events_offset":15,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:40:45.74181068Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:41:00.623387453Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":8,"history_lines":1,"events_offset":17,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:41:00.811009765Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:41:15.623668687Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":9,"history_lines":1,"events_offset":19,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:41:15.819698218Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:41:30.624034153Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":10,"history_lines":1,"events_offset":21,"events_lines":2,"console_offset":12,"console_lines":10} +{"time":"2026-08-14T22:41:30.744375294Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:41:45.623367722Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":1,"events_offset":23,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:41:45.733629686Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:42:00.623945693Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":12,"history_lines":1,"events_offset":25,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:42:00.817478108Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:42:15.623167074Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":27,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:42:15.776781027Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:42:30.623497458Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":1,"events_offset":29,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:42:30.750650809Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:42:45.623560543Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":14,"history_lines":1,"events_offset":31,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:42:45.741626512Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:43:00.623920271Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":15,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":22,"console_lines":10} +{"time":"2026-08-14T22:43:00.743461928Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:43:15.623946176Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":1,"events_offset":35,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:43:15.759830391Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:43:30.623120101Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":17,"history_lines":1,"events_offset":37,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:43:30.781567496Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:43:45.623179309Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":18,"history_lines":1,"events_offset":39,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:43:45.748600011Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:44:00.623702486Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:44:00.751946718Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:44:15.623242218Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":20,"history_lines":1,"events_offset":43,"events_lines":2,"console_offset":32,"console_lines":10} +{"time":"2026-08-14T22:44:15.77231813Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:44:30.623984617Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:44:30.751602635Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:44:45.623364818Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":22,"history_lines":1,"events_offset":47,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:44:45.741854808Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:45:00.623483905Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":23,"history_lines":1,"events_offset":49,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:45:00.800774527Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:45:15.623742903Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":1,"events_offset":51,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:45:15.786111552Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:45:30.623196754Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":25,"history_lines":1,"events_offset":53,"events_lines":2,"console_offset":42,"console_lines":10} +{"time":"2026-08-14T22:45:30.739426043Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:45:45.623413688Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":55,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:45:45.746796974Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:46:00.623004654Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":1,"events_offset":57,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:46:00.761911057Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:46:15.623185767Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":27,"history_lines":1,"events_offset":59,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:46:15.760034626Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:46:30.623209097Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":28,"history_lines":1,"events_offset":61,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:46:30.742448604Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:46:45.62383438Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":1,"events_offset":63,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:46:45.743003272Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:47:00.623709319Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":30,"history_lines":1,"events_offset":65,"events_lines":2,"console_offset":52,"console_lines":10} +{"time":"2026-08-14T22:47:00.765088087Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:47:15.623287238Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":1,"events_offset":67,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:47:15.755938233Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:47:30.6237545Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":32,"history_lines":1,"events_offset":69,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:47:30.754222743Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:47:45.623189066Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":1,"events_offset":71,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:47:45.778899685Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:48:00.622964382Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":34,"history_lines":1,"events_offset":73,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T22:48:00.750228932Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:48:15.623710633Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":1,"events_offset":75,"events_lines":2,"console_offset":62,"console_lines":10} +{"time":"2026-08-14T22:48:15.752612684Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:48:30.624001885Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":36,"history_lines":1,"events_offset":77,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:48:30.765758479Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:48:45.62306379Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":1,"events_offset":79,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:48:45.749591727Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:49:00.623663623Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":38,"history_lines":1,"events_offset":81,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:49:00.761729144Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:49:15.623488018Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":83,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:49:15.740487307Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:49:30.623278765Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":85,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T22:49:30.817638618Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:49:45.623878795Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":1,"events_offset":87,"events_lines":2,"console_offset":72,"console_lines":10} +{"time":"2026-08-14T22:49:45.738463974Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:50:00.623724299Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":41,"history_lines":1,"events_offset":89,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:50:00.799989931Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:50:15.623752365Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":1,"events_offset":91,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:50:15.743875334Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:50:30.623494068Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":43,"history_lines":1,"events_offset":93,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:50:30.770133907Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:50:45.623820847Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":1,"events_offset":95,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T22:50:45.7508848Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:51:00.623390131Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":45,"history_lines":1,"events_offset":97,"events_lines":2,"console_offset":82,"console_lines":10} +{"time":"2026-08-14T22:51:00.745207319Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:51:15.623664471Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":99,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:51:15.736461778Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:51:30.623259339Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":1,"events_offset":101,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:51:30.818786791Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:51:45.623000769Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":48,"history_lines":1,"events_offset":103,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:51:45.765671758Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:52:00.623751572Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":1,"events_offset":105,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:52:00.790080962Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:52:15.623740502Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":50,"history_lines":2,"events_offset":107,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T22:52:15.784594602Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:52:18.133856062Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T22:52:18.134107904Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":52,"history_lines":1,"events_offset":109,"events_lines":1,"console_offset":92,"console_lines":18,"uploaded_len":3,"complete":true,"exit_code":0} +{"time":"2026-08-14T22:52:18.352188013Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:52:18.353320306Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T22:52:18.355743504Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T22:52:18.355755291Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T22:52:18.355885579Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T22:52:18.355890091Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug.log b/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..7c86cf0e863e6f3de652c9fc5451aa8a474eb00f --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug.log @@ -0,0 +1,24 @@ +2026-08-14 22:38:29,838 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():729] Logging user logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug.log +2026-08-14 22:38:29,838 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():730] Logging internal logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_223829-uuv4h936/logs/debug-internal.log +2026-08-14 22:38:29,838 INFO MainThread:1229423 [wandb_init.py:init():772] calling init triggers +2026-08-14 22:38:29,838 INFO MainThread:1229423 [wandb_init.py:init():777] wandb.init called with sweep_config: {} +config: {'_wandb': {}} +2026-08-14 22:38:29,838 INFO MainThread:1229423 [wandb_init.py:init():820] starting backend +2026-08-14 22:38:29,838 INFO MainThread:1229423 [wandb_init.py:init():835] sending inform_init request +2026-08-14 22:38:30,099 INFO MainThread:1229423 [wandb_init.py:init():840] backend started and connected +2026-08-14 22:38:30,101 INFO MainThread:1229423 [wandb_init.py:init():910] updated telemetry +2026-08-14 22:38:30,107 INFO MainThread:1229423 [wandb_init.py:init():933] communicating run to backend with 90.0 second timeout +2026-08-14 22:38:30,541 INFO MainThread:1229423 [wandb_init.py:init():978] starting run threads in backend +2026-08-14 22:38:30,617 INFO MainThread:1229423 [wandb_run.py:_console_start():2621] atexit reg +2026-08-14 22:38:30,618 INFO MainThread:1229423 [wandb_run.py:_redirect():2471] redirect: wrap_raw +2026-08-14 22:38:30,618 INFO MainThread:1229423 [wandb_run.py:_redirect():2540] Wrapping output streams. +2026-08-14 22:38:30,618 INFO MainThread:1229423 [wandb_run.py:_redirect():2563] Redirects installed. +2026-08-14 22:38:30,618 INFO MainThread:1229423 [wandb_init.py:init():1016] run started, returning control to user process +2026-08-14 22:38:30,619 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 21, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'glu', 'activation': 'situglu_low', 'waleed_beta': 10.0, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/glu-situglu_low-21L_run', 'per_device_train_batch_size': 128, 'num_train_epochs': 1, 'max_steps': 1000, 'learning_rate': 0.001, 'lr_scheduler_type': 'constant', 'lr_scheduler_kwargs': None, 'warmup_steps': 0, 'optim': 'adamw_torch', 'optim_args': None, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 4, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-glu-situglu_low-21L-4.0M-20260814-223829', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 60000, 'eval_delay': 0, 'per_device_eval_batch_size': 128, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 100, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/A-glu-situglu_low-21L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} +2026-08-14 22:38:30,621 INFO MainThread:1229423 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 3970432 - > +2026-08-14 22:38:30,621 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 3970432 None +2026-08-14 22:52:17,703 INFO MainThread:1229423 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research1/uuv4h936 +2026-08-14 22:52:17,704 INFO MainThread:1229423 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 22:52:17,704 INFO MainThread:1229423 [wandb_run.py:_restore():2570] restore +2026-08-14 22:52:17,704 INFO MainThread:1229423 [wandb_run.py:_restore():2576] restore done +2026-08-14 22:52:18,355 INFO MainThread:1229423 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_223829-uuv4h936/run-uuv4h936.wandb b/zain/Activation/wandb/run-20260814_223829-uuv4h936/run-uuv4h936.wandb new file mode 100644 index 0000000000000000000000000000000000000000..92610e938d8d0fedf7afbbce0f53643295ed1b8e --- /dev/null +++ b/zain/Activation/wandb/run-20260814_223829-uuv4h936/run-uuv4h936.wandb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:14389923a2e1db443bf3fa8eb594d811eba16747b2df40b54627962bdce417cb +size 584669 diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/config.yaml b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..24131c131174b02a1de0d8ab62cae51e2089c345 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/config.yaml @@ -0,0 +1,439 @@ +_name_or_path: + value: "" +_wandb: + value: + cli_version: 0.28.1 + e: + flcgglol3l78ow5s1xbn0jtcaidsdrfd: + args: + - --config + - /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml + - --variants + - glu-waleed10 + - glu-silu-waleed10 + - mlp-waleed10 + - mlp-silu-waleed10 + - glu-waleed + - glu-situglu_low + - glu-waleedglu_low + codePath: sweep.py + codePathLocal: sweep.py + cpu_count: 112 + cpu_count_logical: 224 + cudaVersion: "12.4" + disk: + /: + total: "1560765693952" + used: "709109972992" + email: deepnevro@gmail.com + executable: /mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python + git: + commit: 3fb47c6943d6927eafc39bd399a3d3a520bae74a + remote: https://github.com/w-ahmad1a10/Activation.git + gpu: NVIDIA H100 80GB HBM3 + gpu_count: 8 + gpu_nvidia: + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-39c684a5-fde6-83d7-1663-0859795881ae + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-132944c4-b689-2b5f-89a4-d730401677ab + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-bc6c3e3c-9b90-09ca-c034-774961847c54 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9 + - architecture: Hopper + cudaCores: 16896 + memoryTotal: "85520809984" + name: NVIDIA H100 80GB HBM3 + uuid: GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea + host: deeplens-k3s-node1 + memory: + total: "2164089937920" + os: Linux-5.15.0-126-generic-x86_64-with-glibc2.35 + program: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py + python: CPython 3.11.15 + root: /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation + startedAt: "2026-08-14T22:52:18.897712Z" + writerId: flcgglol3l78ow5s1xbn0jtcaidsdrfd + m: + - "1": train/global_step + "6": + - 3 + "7": [] + - "2": '*' + "5": 1 + "6": + - 1 + "7": [] + python_version: 3.11.15 + t: + "1": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "2": + - 1 + - 5 + - 11 + - 41 + - 49 + - 51 + - 53 + - 71 + "3": + - 2 + - 7 + - 13 + - 19 + - 62 + - 66 + "4": 3.11.15 + "5": 0.28.1 + "6": 5.16.0.dev0 + "9": + "1": transformers_trainer + "12": 0.28.1 + "13": linux-x86_64 +accelerator_config: + value: + dispatch_batches: null + even_batches: true + gradient_accumulation_kwargs: null + non_blocking: false + split_batches: false + use_seedable_sampler: true +activation: + value: waleedglu_low +adam_beta1: + value: 0.9 +adam_beta2: + value: 0.999 +adam_epsilon: + value: 1e-08 +architectures: + value: null +attention_bias: + value: false +attention_dropout: + value: 0 +auto_find_batch_size: + value: false +average_tokens_across_devices: + value: true +batch_eval_metrics: + value: false +bf16: + value: true +bf16_full_eval: + value: false +bos_token_id: + value: 1 +chunk_size_feed_forward: + value: 0 +data_seed: + value: 42 +dataloader_drop_last: + value: false +dataloader_in_order: + value: true +dataloader_multiprocessing_context: + value: null +dataloader_num_workers: + value: 0 +dataloader_persistent_workers: + value: false +dataloader_pin_memory: + value: true +dataloader_prefetch_factor: + value: null +ddp_backend: + value: null +ddp_broadcast_buffers: + value: null +ddp_bucket_cap_mb: + value: null +ddp_find_unused_parameters: + value: null +ddp_static_graph: + value: null +ddp_timeout: + value: 1800 +debug: + value: [] +deepspeed: + value: null +disable_tqdm: + value: false +do_eval: + value: true +do_predict: + value: false +do_train: + value: false +dtype: + value: null +enable_jit_checkpoint: + value: false +eos_token_id: + value: 2 +eval_accumulation_steps: + value: null +eval_delay: + value: 0 +eval_do_concat_batches: + value: true +eval_on_start: + value: false +eval_steps: + value: 60000 +eval_strategy: + value: steps +eval_use_gather_object: + value: false +fp16: + value: false +fp16_full_eval: + value: false +fsdp: + value: null +fsdp_config: + value: null +full_determinism: + value: false +gradient_accumulation_steps: + value: 4 +gradient_checkpointing: + value: false +gradient_checkpointing_kwargs: + value: null +greater_is_better: + value: null +head_dim: + value: 32 +hidden_act: + value: silu +hidden_size: + value: 128 +hub_always_push: + value: false +hub_model_id: + value: w-ahmad/A-glu-waleedglu_low-21L +hub_private_repo: + value: null +hub_revision: + value: null +hub_strategy: + value: every_save +hub_token: + value: +id2label: + value: + "0": LABEL_0 + "1": LABEL_1 +ignore_data_skip: + value: false +include_for_metrics: + value: [] +include_num_input_tokens_seen: + value: "no" +initializer_range: + value: 0.02 +intermediate_size: + value: 256 +is_encoder_decoder: + value: false +label_names: + value: null +label_smoothing_factor: + value: 0 +label2id: + value: + LABEL_0: 0 + LABEL_1: 1 +learning_rate: + value: 0.001 +length_column_name: + value: length +liger_kernel_config: + value: null +load_best_model_at_end: + value: false +local_rank: + value: -1 +log_level: + value: passive +log_level_replica: + value: warning +log_on_each_node: + value: true +logging_first_step: + value: false +logging_nan_inf_filter: + value: true +logging_steps: + value: 20 +logging_strategy: + value: steps +lr_scheduler_kwargs: + value: null +lr_scheduler_type: + value: constant +max_grad_norm: + value: 1 +max_position_embeddings: + value: 512 +max_steps: + value: 1000 +metric_for_best_model: + value: null +mlp_bias: + value: false +mlp_type: + value: glu +model/num_parameters: + value: 3970432 +model_type: + value: tiny_llama +neftune_noise_alpha: + value: null +num_attention_heads: + value: 4 +num_hidden_layers: + value: 21 +num_key_value_heads: + value: 4 +num_train_epochs: + value: 1 +optim: + value: adamw_torch +optim_args: + value: null +optim_target_modules: + value: null +output_attentions: + value: false +output_dir: + value: out/glu-waleedglu_low-21L_run +output_hidden_states: + value: false +pad_token_id: + value: 0 +parallelism_config: + value: null +per_device_eval_batch_size: + value: 128 +per_device_train_batch_size: + value: 128 +prediction_loss_only: + value: false +pretraining_tp: + value: 1 +problem_type: + value: null +project: + value: huggingface +push_to_hub: + value: false +remove_unused_columns: + value: false +report_to: + value: + - wandb +restore_callback_states_from_checkpoint: + value: false +resume_from_checkpoint: + value: null +return_dict: + value: true +rms_norm_eps: + value: 1e-06 +rope_parameters: + value: + rope_theta: 10000 + rope_type: default +run_name: + value: LM-glu-waleedglu_low-21L-4.0M-20260814-225218 +save_on_each_node: + value: false +save_only_model: + value: false +save_steps: + value: 100 +save_strategy: + value: steps +save_total_limit: + value: null +seed: + value: 42 +skip_memory_metrics: + value: true +tf32: + value: null +tie_word_embeddings: + value: true +tokenizer_name: + value: w-ahmad/tiny-stories-tokenizer +torch_compile: + value: false +torch_compile_backend: + value: null +torch_compile_mode: + value: null +torch_empty_cache_steps: + value: null +trackio_bucket_id: + value: null +trackio_space_id: + value: null +trackio_static_space_id: + value: null +train_sampling_strategy: + value: random +transformers_version: + value: 5.16.0.dev0 +use_cache: + value: false +use_cpu: + value: false +use_liger_kernel: + value: false +vocab_size: + value: 4096 +waleed_beta: + value: 10 +warmup_steps: + value: 0 +weight_decay: + value: 0.01 diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/output.log b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/output.log new file mode 100644 index 0000000000000000000000000000000000000000..bda71c7b0f8c30cd7e43d70c6c3c204db0f83c68 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/output.log @@ -0,0 +1,110 @@ + 10%|█ | 100/1000 [01:19<11:52, 1.26it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '28.88', 'grad_norm': '4', 'learning_rate': '0.001', 'epoch': '0.01078', 'train/total_time_seconds': '8.125', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '15.91', 'train/estimated_remaining_minutes': '6.636'} +{'loss': '24.06', 'grad_norm': '0.9219', 'learning_rate': '0.001', 'epoch': '0.02157', 'train/total_time_seconds': '16.25', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '31.75', 'train/estimated_remaining_minutes': '6.501'} +{'loss': '22.32', 'grad_norm': '1.414', 'learning_rate': '0.001', 'epoch': '0.03235', 'train/total_time_seconds': '24.36', 'train/time_per_step_avg': '0.406', 'train/epoch_time_elapsed': '47.58', 'train/estimated_remaining_minutes': '6.361'} +{'loss': '20.63', 'grad_norm': '2.484', 'learning_rate': '0.001', 'epoch': '0.04314', 'train/total_time_seconds': '32.45', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '63.21', 'train/estimated_remaining_minutes': '6.219'} +{'loss': '18.89', 'grad_norm': '1.852', 'learning_rate': '0.001', 'epoch': '0.05392', 'train/total_time_seconds': '40.54', 'train/time_per_step_avg': '0.4054', 'train/epoch_time_elapsed': '79.09', 'train/estimated_remaining_minutes': '6.08'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 74.14it/s] + 20%|██ | 200/1000 [02:38<10:28, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '17.51', 'grad_norm': '2.141', 'learning_rate': '0.001', 'epoch': '0.06471', 'train/total_time_seconds': '48.67', 'train/time_per_step_avg': '0.4054', 'train/epoch_time_elapsed': '94.94', 'train/estimated_remaining_minutes': '5.948'} +{'loss': '16.46', 'grad_norm': '2.672', 'learning_rate': '0.001', 'epoch': '0.07549', 'train/total_time_seconds': '56.78', 'train/time_per_step_avg': '0.4053', 'train/epoch_time_elapsed': '110.8', 'train/estimated_remaining_minutes': '5.813'} +{'loss': '15.62', 'grad_norm': '2.344', 'learning_rate': '0.001', 'epoch': '0.08628', 'train/total_time_seconds': '64.87', 'train/time_per_step_avg': '0.4051', 'train/epoch_time_elapsed': '126.6', 'train/estimated_remaining_minutes': '5.677'} +{'loss': '14.93', 'grad_norm': '2.125', 'learning_rate': '0.001', 'epoch': '0.09706', 'train/total_time_seconds': '73', 'train/time_per_step_avg': '0.4055', 'train/epoch_time_elapsed': '142.4', 'train/estimated_remaining_minutes': '5.543'} +{'loss': '14.34', 'grad_norm': '3.703', 'learning_rate': '0.001', 'epoch': '0.1078', 'train/total_time_seconds': '81.09', 'train/time_per_step_avg': '0.4055', 'train/epoch_time_elapsed': '158.1', 'train/estimated_remaining_minutes': '5.406'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 81.66it/s] + 30%|███ | 300/1000 [03:57<09:19, 1.25it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '13.89', 'grad_norm': '2.188', 'learning_rate': '0.001', 'epoch': '0.1186', 'train/total_time_seconds': '89.18', 'train/time_per_step_avg': '0.4051', 'train/epoch_time_elapsed': '173.9', 'train/estimated_remaining_minutes': '5.27'} +{'loss': '13.44', 'grad_norm': '2.938', 'learning_rate': '0.001', 'epoch': '0.1294', 'train/total_time_seconds': '97.44', 'train/time_per_step_avg': '0.4067', 'train/epoch_time_elapsed': '189.7', 'train/estimated_remaining_minutes': '5.143'} +{'loss': '13', 'grad_norm': '1.648', 'learning_rate': '0.001', 'epoch': '0.1402', 'train/total_time_seconds': '105.6', 'train/time_per_step_avg': '0.407', 'train/epoch_time_elapsed': '205.7', 'train/estimated_remaining_minutes': '5.008'} +{'loss': '12.63', 'grad_norm': '2.453', 'learning_rate': '0.001', 'epoch': '0.151', 'train/total_time_seconds': '113.7', 'train/time_per_step_avg': '0.4067', 'train/epoch_time_elapsed': '221.5', 'train/estimated_remaining_minutes': '4.872'} +{'loss': '12.29', 'grad_norm': '2.469', 'learning_rate': '0.001', 'epoch': '0.1618', 'train/total_time_seconds': '121.8', 'train/time_per_step_avg': '0.4069', 'train/epoch_time_elapsed': '237.3', 'train/estimated_remaining_minutes': '4.736'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 82.84it/s] + 40%|████ | 400/1000 [05:16<07:57, 1.26it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '11.97', 'grad_norm': '1.766', 'learning_rate': '0.001', 'epoch': '0.1726', 'train/total_time_seconds': '129.9', 'train/time_per_step_avg': '0.4069', 'train/epoch_time_elapsed': '253.2', 'train/estimated_remaining_minutes': '4.6'} +{'loss': '11.69', 'grad_norm': '2.359', 'learning_rate': '0.001', 'epoch': '0.1833', 'train/total_time_seconds': '138', 'train/time_per_step_avg': '0.4055', 'train/epoch_time_elapsed': '269', 'train/estimated_remaining_minutes': '4.465'} +{'loss': '11.43', 'grad_norm': '2.078', 'learning_rate': '0.001', 'epoch': '0.1941', 'train/total_time_seconds': '146.1', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '284.8', 'train/estimated_remaining_minutes': '4.33'} +{'loss': '11.16', 'grad_norm': '2', 'learning_rate': '0.001', 'epoch': '0.2049', 'train/total_time_seconds': '154.2', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '300.7', 'train/estimated_remaining_minutes': '4.194'} +{'loss': '10.96', 'grad_norm': '1.945', 'learning_rate': '0.001', 'epoch': '0.2157', 'train/total_time_seconds': '162.4', 'train/time_per_step_avg': '0.4059', 'train/epoch_time_elapsed': '316.6', 'train/estimated_remaining_minutes': '4.059'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 77.16it/s] + 50%|█████ | 500/1000 [06:35<06:34, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '10.74', 'grad_norm': '1.938', 'learning_rate': '0.001', 'epoch': '0.2265', 'train/total_time_seconds': '170.5', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '332.4', 'train/estimated_remaining_minutes': '3.924'} +{'loss': '10.61', 'grad_norm': '2.422', 'learning_rate': '0.001', 'epoch': '0.2373', 'train/total_time_seconds': '178.6', 'train/time_per_step_avg': '0.4062', 'train/epoch_time_elapsed': '348.3', 'train/estimated_remaining_minutes': '3.789'} +{'loss': '10.41', 'grad_norm': '2.375', 'learning_rate': '0.001', 'epoch': '0.248', 'train/total_time_seconds': '186.7', 'train/time_per_step_avg': '0.406', 'train/epoch_time_elapsed': '363.9', 'train/estimated_remaining_minutes': '3.653'} +{'loss': '10.24', 'grad_norm': '2.109', 'learning_rate': '0.001', 'epoch': '0.2588', 'train/total_time_seconds': '194.8', 'train/time_per_step_avg': '0.4059', 'train/epoch_time_elapsed': '379.8', 'train/estimated_remaining_minutes': '3.518'} +{'loss': '10.09', 'grad_norm': '1.953', 'learning_rate': '0.001', 'epoch': '0.2696', 'train/total_time_seconds': '203', 'train/time_per_step_avg': '0.4059', 'train/epoch_time_elapsed': '395.5', 'train/estimated_remaining_minutes': '3.383'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 60.21it/s] + 60%|██████ | 600/1000 [07:54<05:16, 1.26it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.966', 'grad_norm': '2.281', 'learning_rate': '0.001', 'epoch': '0.2804', 'train/total_time_seconds': '211.1', 'train/time_per_step_avg': '0.4055', 'train/epoch_time_elapsed': '411.5', 'train/estimated_remaining_minutes': '3.247'} +{'loss': '9.85', 'grad_norm': '2.297', 'learning_rate': '0.001', 'epoch': '0.2912', 'train/total_time_seconds': '219.1', 'train/time_per_step_avg': '0.4053', 'train/epoch_time_elapsed': '427.2', 'train/estimated_remaining_minutes': '3.111'} +{'loss': '9.731', 'grad_norm': '2.141', 'learning_rate': '0.001', 'epoch': '0.302', 'train/total_time_seconds': '227.2', 'train/time_per_step_avg': '0.4052', 'train/epoch_time_elapsed': '443', 'train/estimated_remaining_minutes': '2.976'} +{'loss': '9.615', 'grad_norm': '1.859', 'learning_rate': '0.001', 'epoch': '0.3128', 'train/total_time_seconds': '235.3', 'train/time_per_step_avg': '0.4051', 'train/epoch_time_elapsed': '458.7', 'train/estimated_remaining_minutes': '2.84'} +{'loss': '9.506', 'grad_norm': '2.297', 'learning_rate': '0.001', 'epoch': '0.3235', 'train/total_time_seconds': '243.5', 'train/time_per_step_avg': '0.405', 'train/epoch_time_elapsed': '474.8', 'train/estimated_remaining_minutes': '2.705'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 75.29it/s] + 70%|███████ | 700/1000 [09:14<03:57, 1.27it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '9.415', 'grad_norm': '2.047', 'learning_rate': '0.001', 'epoch': '0.3343', 'train/total_time_seconds': '251.6', 'train/time_per_step_avg': '0.4053', 'train/epoch_time_elapsed': '490.8', 'train/estimated_remaining_minutes': '2.57'} +{'loss': '9.329', 'grad_norm': '1.992', 'learning_rate': '0.001', 'epoch': '0.3451', 'train/total_time_seconds': '259.7', 'train/time_per_step_avg': '0.4057', 'train/epoch_time_elapsed': '506.8', 'train/estimated_remaining_minutes': '2.435'} +{'loss': '9.246', 'grad_norm': '1.961', 'learning_rate': '0.001', 'epoch': '0.3559', 'train/total_time_seconds': '267.8', 'train/time_per_step_avg': '0.4061', 'train/epoch_time_elapsed': '522.7', 'train/estimated_remaining_minutes': '2.3'} +{'loss': '9.161', 'grad_norm': '2.078', 'learning_rate': '0.001', 'epoch': '0.3667', 'train/total_time_seconds': '275.9', 'train/time_per_step_avg': '0.4061', 'train/epoch_time_elapsed': '538.6', 'train/estimated_remaining_minutes': '2.164'} +{'loss': '9.089', 'grad_norm': '1.594', 'learning_rate': '0.001', 'epoch': '0.3775', 'train/total_time_seconds': '284.1', 'train/time_per_step_avg': '0.406', 'train/epoch_time_elapsed': '554.3', 'train/estimated_remaining_minutes': '2.029'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 61.15it/s] + 80%|████████ | 800/1000 [10:33<02:35, 1.28it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.999', 'grad_norm': '1.922', 'learning_rate': '0.001', 'epoch': '0.3882', 'train/total_time_seconds': '292.2', 'train/time_per_step_avg': '0.4062', 'train/epoch_time_elapsed': '570.6', 'train/estimated_remaining_minutes': '1.894'} +{'loss': '8.962', 'grad_norm': '1.852', 'learning_rate': '0.001', 'epoch': '0.399', 'train/total_time_seconds': '300.3', 'train/time_per_step_avg': '0.4063', 'train/epoch_time_elapsed': '586.4', 'train/estimated_remaining_minutes': '1.759'} +{'loss': '8.894', 'grad_norm': '1.781', 'learning_rate': '0.001', 'epoch': '0.4098', 'train/total_time_seconds': '308.5', 'train/time_per_step_avg': '0.4061', 'train/epoch_time_elapsed': '602.3', 'train/estimated_remaining_minutes': '1.623'} +{'loss': '8.858', 'grad_norm': '1.906', 'learning_rate': '0.001', 'epoch': '0.4206', 'train/total_time_seconds': '316.6', 'train/time_per_step_avg': '0.4062', 'train/epoch_time_elapsed': '618', 'train/estimated_remaining_minutes': '1.488'} +{'loss': '8.755', 'grad_norm': '1.836', 'learning_rate': '0.001', 'epoch': '0.4314', 'train/total_time_seconds': '324.7', 'train/time_per_step_avg': '0.406', 'train/epoch_time_elapsed': '633.7', 'train/estimated_remaining_minutes': '1.353'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 80.39it/s] + 90%|█████████ | 900/1000 [11:52<01:17, 1.28it/s][transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.733', 'grad_norm': '1.797', 'learning_rate': '0.001', 'epoch': '0.4422', 'train/total_time_seconds': '332.8', 'train/time_per_step_avg': '0.4054', 'train/epoch_time_elapsed': '649.4', 'train/estimated_remaining_minutes': '1.217'} +{'loss': '8.692', 'grad_norm': '1.82', 'learning_rate': '0.001', 'epoch': '0.453', 'train/total_time_seconds': '340.9', 'train/time_per_step_avg': '0.4052', 'train/epoch_time_elapsed': '665.3', 'train/estimated_remaining_minutes': '1.082'} +{'loss': '8.623', 'grad_norm': '1.922', 'learning_rate': '0.001', 'epoch': '0.4637', 'train/total_time_seconds': '349', 'train/time_per_step_avg': '0.405', 'train/epoch_time_elapsed': '681.2', 'train/estimated_remaining_minutes': '0.9468'} +{'loss': '8.586', 'grad_norm': '1.844', 'learning_rate': '0.001', 'epoch': '0.4745', 'train/total_time_seconds': '357', 'train/time_per_step_avg': '0.4047', 'train/epoch_time_elapsed': '696.9', 'train/estimated_remaining_minutes': '0.8114'} +{'loss': '8.535', 'grad_norm': '1.945', 'learning_rate': '0.001', 'epoch': '0.4853', 'train/total_time_seconds': '365.1', 'train/time_per_step_avg': '0.4045', 'train/epoch_time_elapsed': '712.4', 'train/estimated_remaining_minutes': '0.6761'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 79.59it/s] +100%|██████████| 1000/1000 [13:21<00:00, 1.26i[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. +{'loss': '8.493', 'grad_norm': '1.875', 'learning_rate': '0.001', 'epoch': '0.4961', 'train/total_time_seconds': '373.3', 'train/time_per_step_avg': '0.4052', 'train/epoch_time_elapsed': '728.5', 'train/estimated_remaining_minutes': '0.541'} +{'loss': '8.475', 'grad_norm': '1.641', 'learning_rate': '0.001', 'epoch': '0.5069', 'train/total_time_seconds': '381.4', 'train/time_per_step_avg': '0.4056', 'train/epoch_time_elapsed': '744.5', 'train/estimated_remaining_minutes': '0.4058'} +{'loss': '8.457', 'grad_norm': '1.789', 'learning_rate': '0.001', 'epoch': '0.5177', 'train/total_time_seconds': '389.5', 'train/time_per_step_avg': '0.4055', 'train/epoch_time_elapsed': '760.2', 'train/estimated_remaining_minutes': '0.2705'} +{'loss': '8.418', 'grad_norm': '1.695', 'learning_rate': '0.001', 'epoch': '0.5284', 'train/total_time_seconds': '397.6', 'train/time_per_step_avg': '0.406', 'train/epoch_time_elapsed': '776', 'train/estimated_remaining_minutes': '0.1353'} +{'loss': '8.385', 'grad_norm': '1.703', 'learning_rate': '0.001', 'epoch': '0.5392', 'train/total_time_seconds': '405.8', 'train/time_per_step_avg': '0.4068', 'train/epoch_time_elapsed': '792', 'train/estimated_remaining_minutes': '0'} + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes +{'eval_loss': '2.096', 'eval_runtime': '9.686', 'eval_samples_per_second': '983.6', 'eval_steps_per_second': '7.743', 'epoch': '0.5392', 'train/total_time_seconds': '405.8', 'train/time_per_step_avg': '0.4068', 'train/epoch_time_elapsed': '801.7', 'train/estimated_remaining_minutes': '0'} + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 82.55it/s] +100%|██████████| 1000/1000 [13:21<00:00, 1.25it/s], ?it/s] +{'train_runtime': '802.5', 'train_samples_per_second': '638', 'train_steps_per_second': '1.246', 'train_loss': '11.86', 'epoch': '0.5392', 'train/total_time_seconds': '405.8', 'train/time_per_step_avg': '0.4068', 'train/epoch_time_elapsed': '801.7', 'train/estimated_remaining_minutes': '0'} +100%|██████████| 75/75 [00:09<00:00, 7.80it/s] +[transformers] TinyLlamaForCausalLM has generative capabilities, as `prepare_inputs_for_generation` is explicitly defined. However, it doesn't directly inherit from `GenerationMixin`. From 👉v4.50👈 onwards, `PreTrainedModel` will NOT inherit from `GenerationMixin`, and this model will lose the ability to call `generate` and other related functions. + - If you're using `trust_remote_code=True`, you can get rid of this warning by loading the model with an auto class. See https://huggingface.co/docs/transformers/en/model_doc/auto#auto-classes + - If you are the owner of the model architecture code, please modify your model class such that it inherits from `GenerationMixin` (after `PreTrainedModel`, otherwise you'll get an exception). + - If you are not the owner of the model architecture class, please contact the model code owner to update it. +Writing model shards: 100%|██████████| 1/1 [00:00<00:00, 78.79it/s] +>>> FINISHED glu-waleedglu_low-21L successfully diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/requirements.txt b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/requirements.txt new file mode 100644 index 0000000000000000000000000000000000000000..f36b19895172d9637438a445ceacb71d09dc5d02 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/requirements.txt @@ -0,0 +1,149 @@ +asttokens==3.0.1 +comm==0.2.3 +debugpy==1.8.21 +decorator==5.3.1 +executing==2.2.1 +nest-asyncio==1.6.0 +parso==0.8.7 +platformdirs==4.11.0 +psutil==7.2.2 +ptyprocess==0.7.0 +pure_eval==0.2.3 +Pygments==2.20.0 +pyzmq==27.1.0 +setuptools==83.0.0 +six==1.17.0 +tornado==6.5.7 +traitlets==5.15.0 +fsspec==2026.4.0 +wcwidth==0.8.2 +ipython_pygments_lexers==1.1.1 +jedi==0.20.0 +jupyter_core==5.9.1 +matplotlib-inline==0.2.2 +pexpect==4.9.0 +prompt_toolkit==3.0.53 +python-dateutil==2.9.0.post0 +stack_data==0.6.3 +wheel==0.47.0 +jupyter_client==8.9.1 +pip==26.1.2 +ipython==9.15.0 +ipykernel==7.2.0 +threadpoolctl==3.6.0 +pyparsing==3.3.2 +typing_extensions==4.15.0 +Jinja2==3.1.6 +narwhals==2.24.0 +kiwisolver==1.5.0 +joblib==1.5.3 +fonttools==4.63.0 +cycler==0.12.1 +scipy==1.17.1 +pandas==3.0.5 +contourpy==1.3.3 +scikit-learn==1.9.0 +matplotlib==3.11.1 +urllib3==2.7.0 +tqdm==4.70.0 +idna==3.18 +charset-normalizer==3.4.9 +certifi==2026.7.22 +requests==2.34.2 +seaborn==0.13.2 +uv==0.12.0 +shellingham==1.5.4 +mpmath==1.3.0 +attrs==26.1.0 +hf-xet==1.5.2 +nvidia-nccl-cu12==2.21.5 +MarkupSafe==3.0.3 +regex==2026.7.19 +importlib_metadata==9.0.0 +httpcore==1.0.9 +annotated-doc==0.0.5 +multidict==6.7.1 +aiohttp==3.14.3 +aiosignal==1.4.0 +xxhash==3.8.1 +aiohappyeyeballs==2.7.1 +mdurl==0.1.2 +cuda-toolkit==13.0.3.0 +networkx==3.6.1 +PyYAML==6.0.3 +nvidia-cufile==1.15.1.6 +typer==0.27.0 +torchaudio==2.6.0+cu124 +rich==15.0.0 +nvidia-cufft-cu12==11.2.1.3 +h11==0.16.0 +dill==0.4.1 +cuda-pathfinder==1.6.0 +filelock==3.29.0 +nvidia-nvtx-cu12==12.4.127 +httpx==0.28.1 +anyio==4.14.2 +numpy==2.4.4 +yarl==1.24.5 +click==8.4.2 +triton==3.2.0 +frozenlist==1.8.0 +zipp==4.1.0 +propcache==0.5.2 +tokenizers==0.22.2 +markdown-it-py==4.2.0 +nvidia-cuda-runtime==13.0.96 +cuda-bindings==13.3.1 +nvidia-cuda-cupti==13.0.85 +torch==2.6.0+cu124 +multiprocess==0.70.19 +pillow==12.2.0 +transformers==5.16.0.dev0 +wandb==0.28.1 +nvidia-curand==10.4.0.35 +sympy==1.13.1 +nvidia-cusparse==12.6.3.3 +nvidia-cuda-nvrtc==13.0.88 +typing-inspection==0.4.2 +nvidia-cusolver==12.0.4.66 +nvidia-cufft==12.0.0.61 +nvidia-cudnn-cu13==9.20.0.48 +nvidia-cublas==13.1.1.3 +pyarrow==25.0.0 +evaluate==0.4.6 +diffusers==0.39.0 +pydantic==2.13.4 +annotated-types==0.8.0 +protobuf==7.35.1 +sentry-sdk==2.66.1 +einops==0.8.2 +packaging==26.2 +nvidia-nvjitlink-cu12==12.4.127 +nvidia-curand-cu12==10.3.5.147 +nvidia-cusparselt-cu12==0.6.2 +nvidia-cusparse-cu12==12.3.1.170 +nvidia-cuda-runtime-cu12==12.4.127 +torchvision==0.21.0+cu124 +nvidia-cuda-nvrtc-cu12==12.4.127 +nvidia-cuda-cupti-cu12==12.4.127 +nvidia-cusolver-cu12==11.6.1.9 +nvidia-cublas-cu12==12.4.5.8 +nvidia-cudnn-cu12==9.1.0.70 +huggingface_hub==1.26.0 +datasets==5.0.1 +safetensors==0.8.0 +accelerate==1.14.0 +pydantic_core==2.46.4 +ninja==1.13.0 +autocommand==2.2.2 +backports.tarfile==1.2.0 +importlib_metadata==8.7.1 +jaraco.text==4.0.0 +jaraco.context==6.1.0 +jaraco.functools==4.4.0 +more-itertools==10.8.0 +packaging==26.0 +platformdirs==4.4.0 +tomli==2.4.0 +wheel==0.46.3 +zipp==3.23.0 diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/wandb-metadata.json b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/wandb-metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..dabd1bf1cbe6ce934b1bbef7f9611882f08a1c14 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/wandb-metadata.json @@ -0,0 +1,101 @@ +{ + "os": "Linux-5.15.0-126-generic-x86_64-with-glibc2.35", + "python": "CPython 3.11.15", + "startedAt": "2026-08-14T22:52:18.897712Z", + "args": [ + "--config", + "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/configs/baseline.yaml", + "--variants", + "glu-waleed10", + "glu-silu-waleed10", + "mlp-waleed10", + "mlp-silu-waleed10", + "glu-waleed", + "glu-situglu_low", + "glu-waleedglu_low" + ], + "program": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/sweep.py", + "codePath": "sweep.py", + "codePathLocal": "sweep.py", + "git": { + "remote": "https://github.com/w-ahmad1a10/Activation.git", + "commit": "3fb47c6943d6927eafc39bd399a3d3a520bae74a" + }, + "email": "deepnevro@gmail.com", + "root": "/mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation", + "host": "deeplens-k3s-node1", + "executable": "/mnt/data/zainulabideen/zain-exp/notebooks/my_env/bin/python", + "cpu_count": 112, + "cpu_count_logical": 224, + "gpu": "NVIDIA H100 80GB HBM3", + "gpu_count": 8, + "disk": { + "/": { + "total": "1560765693952", + "used": "709109972992" + } + }, + "memory": { + "total": "2164089937920" + }, + "gpu_nvidia": [ + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-39c684a5-fde6-83d7-1663-0859795881ae" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-68012e5a-38b6-b643-0ca6-62fb66720bf3" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-132944c4-b689-2b5f-89a4-d730401677ab" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-2df386cc-6d26-d0e2-7a2d-a057b0d95864" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bfa16575-1d94-1aa2-4537-2c93433f42ef" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-bc6c3e3c-9b90-09ca-c034-774961847c54" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-00a441e1-7c95-e7d6-4c35-43d6b291aea9" + }, + { + "name": "NVIDIA H100 80GB HBM3", + "memoryTotal": "85520809984", + "cudaCores": 16896, + "architecture": "Hopper", + "uuid": "GPU-1c4d29a2-4647-6fce-d8fc-0c5ecfbbd6ea" + } + ], + "cudaVersion": "12.4", + "writerId": "flcgglol3l78ow5s1xbn0jtcaidsdrfd" +} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/wandb-summary.json b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/wandb-summary.json new file mode 100644 index 0000000000000000000000000000000000000000..d03c558ee0d25f4a00844b7fe782453574d70767 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/files/wandb-summary.json @@ -0,0 +1 @@ +{"eval/runtime":9.8163,"total_flos":5.420315836416e+15,"train/learning_rate":0.001,"_timestamp":1.7867487511890867e+09,"train/grad_norm":1.703125,"train_runtime":802.5252,"eval/loss":2.096169948577881,"train/loss":8.385391235351562,"train_loss":11.859544631958007,"train/epoch":0.5392289026691831,"_runtime":811,"train_steps_per_second":1.246,"eval/samples_per_second":970.533,"eval/steps_per_second":7.64,"_wandb":{"runtime":811},"_step":52,"train/global_step":1000,"train_samples_per_second":637.986} \ No newline at end of file diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug-core.log b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug-core.log new file mode 100644 index 0000000000000000000000000000000000000000..588056df10bc8cabd5e2682287953b3e2b130ffe --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug-core.log @@ -0,0 +1,62 @@ +{"time":"2026-08-14T21:31:53.539152008Z","level":"INFO","msg":"main: starting server","port-filename":"/tmp/tmp650y3sc2/port-1229423.txt","pid":1229423,"detached":false,"idle-timeout":600000000000,"log-level":0,"disable-analytics":false,"shutdown-on-parent-exit":false,"enable-dcgm-profiling":false} +{"time":"2026-08-14T21:31:53.540676876Z","level":"INFO","msg":"server: will exit if parent process dies","ppid":1229423} +{"time":"2026-08-14T21:31:53.540659035Z","level":"INFO","msg":"server: accepting connections","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T21:31:53.717724383Z","level":"INFO","msg":"connection: ManageConnectionData: new connection created","id":"1(@)"} +{"time":"2026-08-14T21:31:53.802092564Z","level":"INFO","msg":"handleInformInit: received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:54.062397411Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:31:59.411924652Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.396635188Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.929834053Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"32s1adqk73rr"} +{"time":"2026-08-14T21:45:48.931620539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:48.932491197Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"q9vcmje5","id":"1(@)"} +{"time":"2026-08-14T21:45:49.464388506Z","level":"INFO","msg":"handleInformInit: received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:49.733853364Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:45:55.329792921Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.158537572Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.914569559Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"dr322rnt7xgx"} +{"time":"2026-08-14T21:59:01.916156477Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:01.917534874Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"k24y90ws","id":"1(@)"} +{"time":"2026-08-14T21:59:02.524868682Z","level":"INFO","msg":"handleInformInit: received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:02.786199875Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T21:59:08.192858634Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:53.643379161Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.173579992Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2g99yy5pussw"} +{"time":"2026-08-14T22:11:54.17490129Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.175479186Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"npap5k3g","id":"1(@)"} +{"time":"2026-08-14T22:11:54.694802218Z","level":"INFO","msg":"handleInformInit: received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:11:54.956642682Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:12:00.441973218Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:55.662446346Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.296327526Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"raaqqqmajsfs"} +{"time":"2026-08-14T22:24:56.297810539Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.29856977Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"a3607q94","id":"1(@)"} +{"time":"2026-08-14T22:24:56.82560196Z","level":"INFO","msg":"handleInformInit: received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:24:57.085772231Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:25:02.571913681Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:28.612623637Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.208727987Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"2nb8wusqu4fo"} +{"time":"2026-08-14T22:38:29.210106262Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.210705504Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"wcvy5vyw","id":"1(@)"} +{"time":"2026-08-14T22:38:29.839594174Z","level":"INFO","msg":"handleInformInit: received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:30.099444184Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:38:35.620666929Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:17.704425638Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.35448297Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"r0aouorh27ls"} +{"time":"2026-08-14T22:52:18.355720686Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.356578583Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"uuv4h936","id":"1(@)"} +{"time":"2026-08-14T22:52:18.899799881Z","level":"INFO","msg":"handleInformInit: received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:19.159494071Z","level":"INFO","msg":"handleInformInit: stream started","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T22:52:24.671083701Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.214192173Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.783107435Z","level":"INFO","msg":"connection: cancelling request","id":"1(@)","requestId":"hn7lqjhi3ekb"} +{"time":"2026-08-14T23:05:51.784410941Z","level":"INFO","msg":"handleInformFinish: finish message received","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:51.78516667Z","level":"INFO","msg":"handleInformFinish: stream closed","streamId":"x87hwrtd","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088192277Z","level":"INFO","msg":"handleInformTeardown: server teardown initiated","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088229648Z","level":"INFO","msg":"handleInformTeardown: server shutdown complete","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088239651Z","level":"INFO","msg":"connection: closing","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088243873Z","level":"INFO","msg":"server: is shutting down"} +{"time":"2026-08-14T23:05:52.0882852Z","level":"INFO","msg":"connection: closed successfully","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088299095Z","level":"INFO","msg":"processOutgoingData: finished","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088345917Z","level":"INFO","msg":"connection: ManageConnectionData: connection closed","id":"1(@)"} +{"time":"2026-08-14T23:05:52.088452051Z","level":"INFO","msg":"server: listener closed","addr":{"Name":"/tmp/wandb-1229423-1230766-2604858008/socket","Net":"unix"}} +{"time":"2026-08-14T23:05:52.088520571Z","level":"INFO","msg":"server: all connections closed"} diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug-internal.log b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug-internal.log new file mode 100644 index 0000000000000000000000000000000000000000..f0d6a5145da181eff8e09db1cf162d7d9c3fe892 --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug-internal.log @@ -0,0 +1,125 @@ +{"time":"2026-08-14T22:52:18.899935224Z","level":"INFO","msg":"wandb-core"} +{"time":"2026-08-14T22:52:18.900059726Z","level":"INFO","msg":"stream: starting","core version":"0.28.1"} +{"time":"2026-08-14T22:52:19.159323457Z","level":"INFO","msg":"stream: created new stream","id":"x87hwrtd"} +{"time":"2026-08-14T22:52:19.159397893Z","level":"INFO","msg":"handler: started"} +{"time":"2026-08-14T22:52:19.159488091Z","level":"INFO","msg":"stream: started"} +{"time":"2026-08-14T22:52:19.159498739Z","level":"INFO","msg":"writer: started","stream_id":"x87hwrtd"} +{"time":"2026-08-14T22:52:19.159517459Z","level":"INFO","msg":"sender: started"} +{"time":"2026-08-14T22:52:19.639774365Z","level":"INFO","msg":"filestream: sending request","total_files":1,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:52:19.747545916Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:52:34.63989996Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":0,"events_lines":1,"console_offset":0,"console_lines":1,"uploaded_len":2} +{"time":"2026-08-14T22:52:34.753206901Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:52:49.640161126Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":0,"history_lines":1,"events_offset":1,"events_lines":2,"console_offset":0,"console_lines":2} +{"time":"2026-08-14T22:52:49.753656239Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:53:04.64061303Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":1,"history_lines":1,"events_offset":3,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:53:05.12850246Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:53:19.640773246Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":2,"history_lines":1,"events_offset":5,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:53:19.784076666Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:53:34.64072698Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":3,"history_lines":1,"events_offset":7,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:53:34.790272192Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:53:49.640061838Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":4,"history_lines":1,"events_offset":9,"events_lines":2,"console_offset":0,"console_lines":1} +{"time":"2026-08-14T22:53:49.766974735Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:54:04.640138525Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":5,"history_lines":1,"events_offset":11,"events_lines":2,"console_offset":2,"console_lines":10} +{"time":"2026-08-14T22:54:04.819143226Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:54:19.640674048Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":6,"history_lines":1,"events_offset":13,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:54:19.781153852Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:54:34.64021132Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":7,"history_lines":1,"events_offset":15,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:54:34.759141516Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:54:49.640114934Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":8,"history_lines":1,"events_offset":17,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:54:49.774319392Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:55:04.640020451Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":9,"history_lines":1,"events_offset":19,"events_lines":2,"console_offset":10,"console_lines":1} +{"time":"2026-08-14T22:55:04.821670501Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:55:19.640218533Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":10,"history_lines":1,"events_offset":21,"events_lines":2,"console_offset":12,"console_lines":10} +{"time":"2026-08-14T22:55:19.781167787Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:55:34.639954436Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":11,"history_lines":1,"events_offset":23,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:55:34.756231989Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:55:49.640866912Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":12,"history_lines":1,"events_offset":25,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:55:49.773763299Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:56:04.640699027Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":13,"history_lines":1,"events_offset":27,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:56:04.832391223Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:56:19.640778711Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":14,"history_lines":1,"events_offset":29,"events_lines":2,"console_offset":20,"console_lines":1} +{"time":"2026-08-14T22:56:19.792824806Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:56:34.640618055Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":15,"history_lines":1,"events_offset":31,"events_lines":2,"console_offset":22,"console_lines":10} +{"time":"2026-08-14T22:56:34.765952009Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:56:49.640749335Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":16,"history_lines":1,"events_offset":33,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:56:49.777390871Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:57:04.640765379Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":17,"history_lines":1,"events_offset":35,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:57:04.78237223Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:57:19.639838114Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":37,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:57:19.761700546Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:57:34.640153773Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":18,"history_lines":1,"events_offset":39,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:57:34.774227677Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:57:49.640071669Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":19,"history_lines":1,"events_offset":41,"events_lines":2,"console_offset":30,"console_lines":1} +{"time":"2026-08-14T22:57:49.784561715Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:58:04.640369162Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":20,"history_lines":1,"events_offset":43,"events_lines":2,"console_offset":32,"console_lines":10} +{"time":"2026-08-14T22:58:04.822086475Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:58:19.640455468Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":21,"history_lines":1,"events_offset":45,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:58:19.774003293Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:58:34.640258375Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":22,"history_lines":1,"events_offset":47,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:58:34.768594022Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:58:49.640664677Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":23,"history_lines":1,"events_offset":49,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:58:49.797977928Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:59:04.640408194Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":24,"history_lines":1,"events_offset":51,"events_lines":2,"console_offset":40,"console_lines":1} +{"time":"2026-08-14T22:59:04.78466187Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:59:19.639900545Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":25,"history_lines":1,"events_offset":53,"events_lines":2,"console_offset":42,"console_lines":10} +{"time":"2026-08-14T22:59:19.814854155Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:59:34.640141393Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":26,"history_lines":1,"events_offset":55,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:59:34.764394275Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T22:59:49.64017283Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":27,"history_lines":1,"events_offset":57,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T22:59:49.790137488Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:00:04.640230589Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":28,"history_lines":1,"events_offset":59,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T23:00:04.856602179Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:00:19.639863667Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":29,"history_lines":1,"events_offset":61,"events_lines":2,"console_offset":50,"console_lines":1} +{"time":"2026-08-14T23:00:19.784597592Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:00:34.640280511Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":30,"history_lines":1,"events_offset":63,"events_lines":2,"console_offset":52,"console_lines":10} +{"time":"2026-08-14T23:00:34.755114248Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:00:49.64027529Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":31,"history_lines":1,"events_offset":65,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T23:00:49.801645223Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:01:04.640368598Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":32,"history_lines":1,"events_offset":67,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T23:01:04.780184969Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:01:19.64059479Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":33,"history_lines":1,"events_offset":69,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T23:01:19.932445296Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:01:34.640822178Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":34,"history_lines":1,"events_offset":71,"events_lines":2,"console_offset":60,"console_lines":1} +{"time":"2026-08-14T23:01:34.778763495Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:01:49.640735236Z","level":"INFO","msg":"filestream: sending request","total_files":2,"events_offset":73,"events_lines":2,"console_offset":62,"console_lines":9} +{"time":"2026-08-14T23:01:49.796279172Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:02:04.640680739Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":35,"history_lines":1,"events_offset":75,"events_lines":2,"console_offset":70,"console_lines":2} +{"time":"2026-08-14T23:02:04.818107653Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:02:19.640102427Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":36,"history_lines":1,"events_offset":77,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T23:02:19.762193316Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:02:34.640139877Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":37,"history_lines":1,"events_offset":79,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T23:02:34.77673409Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:02:49.64022817Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":38,"history_lines":1,"events_offset":81,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T23:02:49.754709121Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:03:04.640066603Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":39,"history_lines":1,"events_offset":83,"events_lines":2,"console_offset":70,"console_lines":1} +{"time":"2026-08-14T23:03:04.766355544Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:03:19.64047442Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":40,"history_lines":1,"events_offset":85,"events_lines":2,"console_offset":72,"console_lines":10} +{"time":"2026-08-14T23:03:19.766918789Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:03:34.641007741Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":41,"history_lines":1,"events_offset":87,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T23:03:34.756058889Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:03:49.639971734Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":42,"history_lines":1,"events_offset":89,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T23:03:49.855593366Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:04:04.64049672Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":43,"history_lines":1,"events_offset":91,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T23:04:04.826373715Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:04:19.640075875Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":44,"history_lines":1,"events_offset":93,"events_lines":2,"console_offset":80,"console_lines":1} +{"time":"2026-08-14T23:04:19.777401573Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:04:34.640677385Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":45,"history_lines":1,"events_offset":95,"events_lines":2,"console_offset":82,"console_lines":10} +{"time":"2026-08-14T23:04:34.75097421Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:04:49.640209408Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":46,"history_lines":1,"events_offset":97,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T23:04:49.771265805Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:05:04.640260707Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":47,"history_lines":1,"events_offset":99,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T23:05:04.814218179Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:05:19.640605006Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":48,"history_lines":1,"events_offset":101,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T23:05:19.789343315Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:05:34.640257812Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":49,"history_lines":1,"events_offset":103,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T23:05:34.762414876Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:05:49.640633384Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":50,"history_lines":2,"events_offset":105,"events_lines":2,"console_offset":90,"console_lines":1} +{"time":"2026-08-14T23:05:49.792158266Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:05:51.648767082Z","level":"INFO","msg":"fileTransfer: Close: file transfer manager closed"} +{"time":"2026-08-14T23:05:51.64896578Z","level":"INFO","msg":"filestream: sending request","total_files":4,"history_offset":52,"history_lines":1,"events_offset":107,"events_lines":1,"console_offset":92,"console_lines":18,"uploaded_len":3,"complete":true,"exit_code":0} +{"time":"2026-08-14T23:05:51.780983716Z","level":"INFO","msg":"filestream: request sent","status":"200 OK"} +{"time":"2026-08-14T23:05:51.782090216Z","level":"INFO","msg":"handler: operation stats","stats":{}} +{"time":"2026-08-14T23:05:51.784439708Z","level":"INFO","msg":"stream: finishing up"} +{"time":"2026-08-14T23:05:51.784450919Z","level":"INFO","msg":"handler: closed"} +{"time":"2026-08-14T23:05:51.784580115Z","level":"INFO","msg":"sender: closed"} +{"time":"2026-08-14T23:05:51.784585218Z","level":"INFO","msg":"stream: all finished"} diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug.log b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug.log new file mode 100644 index 0000000000000000000000000000000000000000..260ba09071346d6486053be4fd7d8462c229df2f --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug.log @@ -0,0 +1,24 @@ +2026-08-14 22:52:18,898 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():729] Logging user logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug.log +2026-08-14 22:52:18,898 INFO MainThread:1229423 [wandb_init.py:setup_run_log_directory():730] Logging internal logs to /mnt/data/zainulabideen/zain-exp/notebooks/zain/Activation/wandb/run-20260814_225218-x87hwrtd/logs/debug-internal.log +2026-08-14 22:52:18,898 INFO MainThread:1229423 [wandb_init.py:init():772] calling init triggers +2026-08-14 22:52:18,898 INFO MainThread:1229423 [wandb_init.py:init():777] wandb.init called with sweep_config: {} +config: {'_wandb': {}} +2026-08-14 22:52:18,898 INFO MainThread:1229423 [wandb_init.py:init():820] starting backend +2026-08-14 22:52:18,898 INFO MainThread:1229423 [wandb_init.py:init():835] sending inform_init request +2026-08-14 22:52:19,159 INFO MainThread:1229423 [wandb_init.py:init():840] backend started and connected +2026-08-14 22:52:19,161 INFO MainThread:1229423 [wandb_init.py:init():910] updated telemetry +2026-08-14 22:52:19,167 INFO MainThread:1229423 [wandb_init.py:init():933] communicating run to backend with 90.0 second timeout +2026-08-14 22:52:19,559 INFO MainThread:1229423 [wandb_init.py:init():978] starting run threads in backend +2026-08-14 22:52:19,634 INFO MainThread:1229423 [wandb_run.py:_console_start():2621] atexit reg +2026-08-14 22:52:19,634 INFO MainThread:1229423 [wandb_run.py:_redirect():2471] redirect: wrap_raw +2026-08-14 22:52:19,635 INFO MainThread:1229423 [wandb_run.py:_redirect():2540] Wrapping output streams. +2026-08-14 22:52:19,635 INFO MainThread:1229423 [wandb_run.py:_redirect():2563] Redirects installed. +2026-08-14 22:52:19,635 INFO MainThread:1229423 [wandb_init.py:init():1016] run started, returning control to user process +2026-08-14 22:52:19,636 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb None None {'transformers_version': '5.16.0.dev0', 'architectures': None, 'output_hidden_states': False, 'return_dict': True, 'dtype': None, 'chunk_size_feed_forward': 0, 'is_encoder_decoder': False, 'id2label': {0: 'LABEL_0', 1: 'LABEL_1'}, 'label2id': {'LABEL_0': 0, 'LABEL_1': 1}, 'problem_type': None, 'vocab_size': 4096, 'hidden_size': 128, 'intermediate_size': 256, 'num_hidden_layers': 21, 'num_attention_heads': 4, 'num_key_value_heads': 4, 'hidden_act': 'silu', 'max_position_embeddings': 512, 'initializer_range': 0.02, 'rms_norm_eps': 1e-06, 'use_cache': False, 'pad_token_id': 0, 'bos_token_id': 1, 'eos_token_id': 2, 'pretraining_tp': 1, 'tie_word_embeddings': True, 'rope_parameters': {'rope_theta': 10000.0, 'rope_type': 'default'}, 'attention_bias': False, 'attention_dropout': 0.0, 'mlp_bias': False, 'head_dim': 32, '_name_or_path': '', 'tokenizer_name': 'w-ahmad/tiny-stories-tokenizer', 'mlp_type': 'glu', 'activation': 'waleedglu_low', 'waleed_beta': 10.0, 'model_type': 'tiny_llama', 'output_attentions': False, 'output_dir': 'out/glu-waleedglu_low-21L_run', 'per_device_train_batch_size': 128, 'num_train_epochs': 1, 'max_steps': 1000, 'learning_rate': 0.001, 'lr_scheduler_type': 'constant', 'lr_scheduler_kwargs': None, 'warmup_steps': 0, 'optim': 'adamw_torch', 'optim_args': None, 'weight_decay': 0.01, 'adam_beta1': 0.9, 'adam_beta2': 0.999, 'adam_epsilon': 1e-08, 'optim_target_modules': None, 'gradient_accumulation_steps': 4, 'average_tokens_across_devices': True, 'max_grad_norm': 1.0, 'label_smoothing_factor': 0.0, 'bf16': True, 'fp16': False, 'bf16_full_eval': False, 'fp16_full_eval': False, 'tf32': None, 'gradient_checkpointing': False, 'gradient_checkpointing_kwargs': None, 'torch_compile': False, 'torch_compile_backend': None, 'torch_compile_mode': None, 'use_liger_kernel': False, 'liger_kernel_config': None, 'neftune_noise_alpha': None, 'torch_empty_cache_steps': None, 'auto_find_batch_size': False, 'logging_strategy': 'steps', 'logging_steps': 20, 'logging_first_step': False, 'log_on_each_node': True, 'logging_nan_inf_filter': True, 'include_num_input_tokens_seen': 'no', 'log_level': 'passive', 'log_level_replica': 'warning', 'disable_tqdm': False, 'report_to': ['wandb'], 'run_name': 'LM-glu-waleedglu_low-21L-4.0M-20260814-225218', 'project': 'huggingface', 'trackio_space_id': None, 'trackio_bucket_id': None, 'trackio_static_space_id': None, 'eval_strategy': 'steps', 'eval_steps': 60000, 'eval_delay': 0, 'per_device_eval_batch_size': 128, 'prediction_loss_only': False, 'eval_on_start': False, 'eval_do_concat_batches': True, 'eval_use_gather_object': False, 'eval_accumulation_steps': None, 'include_for_metrics': [], 'batch_eval_metrics': False, 'save_only_model': False, 'save_strategy': 'steps', 'save_steps': 100, 'save_on_each_node': False, 'save_total_limit': None, 'enable_jit_checkpoint': False, 'push_to_hub': False, 'hub_token': '', 'hub_private_repo': None, 'hub_model_id': 'w-ahmad/A-glu-waleedglu_low-21L', 'hub_strategy': 'every_save', 'hub_always_push': False, 'hub_revision': None, 'load_best_model_at_end': False, 'metric_for_best_model': None, 'greater_is_better': None, 'ignore_data_skip': False, 'restore_callback_states_from_checkpoint': False, 'full_determinism': False, 'seed': 42, 'data_seed': 42, 'use_cpu': False, 'accelerator_config': {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}, 'parallelism_config': None, 'dataloader_drop_last': False, 'dataloader_num_workers': 0, 'dataloader_pin_memory': True, 'dataloader_persistent_workers': False, 'dataloader_prefetch_factor': None, 'dataloader_multiprocessing_context': None, 'dataloader_in_order': True, 'remove_unused_columns': False, 'label_names': None, 'train_sampling_strategy': 'random', 'length_column_name': 'length', 'ddp_find_unused_parameters': None, 'ddp_bucket_cap_mb': None, 'ddp_broadcast_buffers': None, 'ddp_static_graph': None, 'ddp_backend': None, 'ddp_timeout': 1800, 'fsdp': None, 'fsdp_config': None, 'deepspeed': None, 'debug': [], 'skip_memory_metrics': True, 'do_train': False, 'do_eval': True, 'do_predict': False, 'resume_from_checkpoint': None, 'local_rank': -1} +2026-08-14 22:52:19,638 INFO MainThread:1229423 [wandb_config.py:__setitem__():155] [no run ID] config set model/num_parameters = 3970432 - > +2026-08-14 22:52:19,638 INFO MainThread:1229423 [wandb_run.py:_config_callback():1346] config_cb model/num_parameters 3970432 None +2026-08-14 23:05:51,213 INFO MainThread:1229423 [wandb_run.py:_finish():2383] finishing run deepnevro-deepnevro/research1/x87hwrtd +2026-08-14 23:05:51,213 INFO MainThread:1229423 [wandb_run.py:_atexit_cleanup():2588] got exitcode: 0 +2026-08-14 23:05:51,214 INFO MainThread:1229423 [wandb_run.py:_restore():2570] restore +2026-08-14 23:05:51,214 INFO MainThread:1229423 [wandb_run.py:_restore():2576] restore done +2026-08-14 23:05:51,783 INFO MainThread:1229423 [wandb_run.py:_footer_sync_info():3993] logging synced files diff --git a/zain/Activation/wandb/run-20260814_225218-x87hwrtd/run-x87hwrtd.wandb b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/run-x87hwrtd.wandb new file mode 100644 index 0000000000000000000000000000000000000000..ca137123dbc75ea6a50192e9e4a5a47e8d90711e --- /dev/null +++ b/zain/Activation/wandb/run-20260814_225218-x87hwrtd/run-x87hwrtd.wandb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0d4bf270604d1c66ad37ae16d99f9e4fe1cf61a755592ff70a0fdb4604b7f9e2 +size 581657