Upload exp_phase8_bridge_recipeABC_KDvector_205921/log.jsonl with huggingface_hub
Browse files
exp_phase8_bridge_recipeABC_KDvector_205921/log.jsonl
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"step": 0, "loss": 1.4397666454315186, "loss_mdlm": 1.2537955045700073, "loss_lm": 1.8597111701965332, "loss_ct": 0.0, "lr": 0.0, "gnorm": 0.6087887287139893, "bridge_gate_avg": 1.0, "bridge_out_proj_avg": 0.0, "elapsed_s": 3.439493179321289}
|
| 2 |
+
{"step": 100, "loss": 1.5048388242721558, "loss_mdlm": 1.3898241519927979, "loss_lm": 1.1501470804214478, "loss_ct": 0.0, "lr": 2.5e-05, "gnorm": 1.3593474626541138, "bridge_gate_avg": 1.000905990600586, "bridge_out_proj_avg": 0.00041494897595839575, "elapsed_s": 39.62070178985596}
|
| 3 |
+
{"step": 200, "loss": 1.4586104154586792, "loss_mdlm": 1.307114601135254, "loss_lm": 1.5149580240249634, "loss_ct": 0.0, "lr": 5e-05, "gnorm": 0.8405117392539978, "bridge_gate_avg": 1.0032425820827484, "bridge_out_proj_avg": 0.0014885156997479498, "elapsed_s": 72.97754859924316}
|
| 4 |
+
{"step": 300, "loss": 1.3889762163162231, "loss_mdlm": 1.3210264444351196, "loss_lm": 0.679497241973877, "loss_ct": 0.0, "lr": 4.962019382530521e-05, "gnorm": 0.5548238754272461, "bridge_gate_avg": 1.0042250752449036, "bridge_out_proj_avg": 0.0017994331719819456, "elapsed_s": 105.82692909240723}
|
| 5 |
+
{"step": 400, "loss": 1.4189480543136597, "loss_mdlm": 1.3294824361801147, "loss_lm": 0.8946556448936462, "loss_ct": 0.0, "lr": 4.849231551964771e-05, "gnorm": 0.6611032485961914, "bridge_gate_avg": 1.0050831139087677, "bridge_out_proj_avg": 0.0019464924989733845, "elapsed_s": 139.0864233970642}
|
| 6 |
+
{"step": 500, "loss": 1.3946219682693481, "loss_mdlm": 1.318877935409546, "loss_lm": 0.7574405670166016, "loss_ct": 0.0, "lr": 4.665063509461097e-05, "gnorm": 0.622032642364502, "bridge_gate_avg": 1.0066613256931305, "bridge_out_proj_avg": 0.0020702381734736264, "elapsed_s": 170.70620012283325}
|
| 7 |
+
{"step": 600, "loss": 1.4287285804748535, "loss_mdlm": 1.3591159582138062, "loss_lm": 0.696125864982605, "loss_ct": 0.0, "lr": 4.415111107797445e-05, "gnorm": 0.45830243825912476, "bridge_gate_avg": 1.0077978074550629, "bridge_out_proj_avg": 0.0021373408962972462, "elapsed_s": 202.25112318992615}
|
| 8 |
+
{"step": 700, "loss": 1.3978749513626099, "loss_mdlm": 1.3242653608322144, "loss_lm": 0.7360957264900208, "loss_ct": 0.0, "lr": 4.1069690242163484e-05, "gnorm": 0.5344671607017517, "bridge_gate_avg": 1.008602261543274, "bridge_out_proj_avg": 0.0022139695938676596, "elapsed_s": 234.09968161582947}
|
| 9 |
+
{"step": 800, "loss": 1.3974812030792236, "loss_mdlm": 1.2946234941482544, "loss_lm": 1.028577208518982, "loss_ct": 0.0, "lr": 3.7500000000000003e-05, "gnorm": 0.6743119359016418, "bridge_gate_avg": 1.0092793405056, "bridge_out_proj_avg": 0.002260442590340972, "elapsed_s": 265.6796271800995}
|
| 10 |
+
{"step": 900, "loss": 1.408442497253418, "loss_mdlm": 1.3355846405029297, "loss_lm": 0.728578507900238, "loss_ct": 0.0, "lr": 3.355050358314172e-05, "gnorm": 0.5627123117446899, "bridge_gate_avg": 1.0098463594913483, "bridge_out_proj_avg": 0.0022965074749663472, "elapsed_s": 297.5393133163452}
|
| 11 |
+
{"step": 1000, "loss": 1.410537600517273, "loss_mdlm": 1.3456944227218628, "loss_lm": 0.6484313011169434, "loss_ct": 0.0, "lr": 2.9341204441673266e-05, "gnorm": 0.4210379123687744, "bridge_gate_avg": 1.010658323764801, "bridge_out_proj_avg": 0.002353915711864829, "elapsed_s": 329.41717314720154}
|