Upload exp_phase8_bridge_recipeABC_KDfinal_210539/log.jsonl with huggingface_hub
Browse files
exp_phase8_bridge_recipeABC_KDfinal_210539/log.jsonl
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"step": 0, "loss": 1.4550122022628784, "loss_mdlm": 1.259800672531128, "loss_lm": 1.952115535736084, "loss_ct": 0.0, "lr": 0.0, "gnorm": 3.7382724285125732, "bridge_gate_avg": 1.0, "bridge_out_proj_avg": 0.0, "elapsed_s": 3.3938510417938232}
|
| 2 |
+
{"step": 100, "loss": 1.4721546173095703, "loss_mdlm": 1.3542838096618652, "loss_lm": 1.178707480430603, "loss_ct": 0.0, "lr": 2.5e-05, "gnorm": 0.7069711685180664, "bridge_gate_avg": 1.0006183683872223, "bridge_out_proj_avg": 0.0002895164871006273, "elapsed_s": 39.042622566223145}
|
| 3 |
+
{"step": 200, "loss": 1.4553706645965576, "loss_mdlm": 1.306047797203064, "loss_lm": 1.493228554725647, "loss_ct": 0.0, "lr": 5e-05, "gnorm": 0.8207499980926514, "bridge_gate_avg": 1.0025786757469177, "bridge_out_proj_avg": 0.0011746301024686545, "elapsed_s": 72.44151282310486}
|
| 4 |
+
{"step": 300, "loss": 1.3855949640274048, "loss_mdlm": 1.3179843425750732, "loss_lm": 0.676105797290802, "loss_ct": 0.0, "lr": 4.962019382530521e-05, "gnorm": 0.5946103930473328, "bridge_gate_avg": 1.0040066242218018, "bridge_out_proj_avg": 0.0016325409524142742, "elapsed_s": 106.13035106658936}
|
| 5 |
+
{"step": 400, "loss": 1.4263664484024048, "loss_mdlm": 1.3361878395080566, "loss_lm": 0.9017863273620605, "loss_ct": 0.0, "lr": 4.849231551964771e-05, "gnorm": 0.6501176953315735, "bridge_gate_avg": 1.0049960911273956, "bridge_out_proj_avg": 0.0018213355797342956, "elapsed_s": 148.903493642807}
|
| 6 |
+
{"step": 500, "loss": 1.4045137166976929, "loss_mdlm": 1.3280352354049683, "loss_lm": 0.7647848725318909, "loss_ct": 0.0, "lr": 4.665063509461097e-05, "gnorm": 0.6396908760070801, "bridge_gate_avg": 1.0064693093299866, "bridge_out_proj_avg": 0.00198435943457298, "elapsed_s": 192.2466812133789}
|
| 7 |
+
{"step": 600, "loss": 1.4320340156555176, "loss_mdlm": 1.361405611038208, "loss_lm": 0.7062844038009644, "loss_ct": 0.0, "lr": 4.415111107797445e-05, "gnorm": 0.4718630313873291, "bridge_gate_avg": 1.0077454149723053, "bridge_out_proj_avg": 0.002063257503323257, "elapsed_s": 235.4988136291504}
|
| 8 |
+
{"step": 700, "loss": 1.3952058553695679, "loss_mdlm": 1.3229749202728271, "loss_lm": 0.7223093509674072, "loss_ct": 0.0, "lr": 4.1069690242163484e-05, "gnorm": 0.5057083368301392, "bridge_gate_avg": 1.008663684129715, "bridge_out_proj_avg": 0.002154247194994241, "elapsed_s": 279.5959782600403}
|
| 9 |
+
{"step": 800, "loss": 1.3961926698684692, "loss_mdlm": 1.2931430339813232, "loss_lm": 1.0304961204528809, "loss_ct": 0.0, "lr": 3.7500000000000003e-05, "gnorm": 0.6546099185943604, "bridge_gate_avg": 1.0095482468605042, "bridge_out_proj_avg": 0.00224525696830824, "elapsed_s": 323.0146646499634}
|
| 10 |
+
{"step": 900, "loss": 1.3548227548599243, "loss_mdlm": 1.2822661399841309, "loss_lm": 0.7255656719207764, "loss_ct": 0.0, "lr": 3.355050358314172e-05, "gnorm": 0.55722975730896, "bridge_gate_avg": 1.009833812713623, "bridge_out_proj_avg": 0.002278179454151541, "elapsed_s": 366.97983288764954}
|
| 11 |
+
{"step": 1000, "loss": 1.3783096075057983, "loss_mdlm": 1.3135958909988403, "loss_lm": 0.6471367478370667, "loss_ct": 0.0, "lr": 2.9341204441673266e-05, "gnorm": 0.41373664140701294, "bridge_gate_avg": 1.010467916727066, "bridge_out_proj_avg": 0.0023215481778606772, "elapsed_s": 411.454332113266}
|
| 12 |
+
{"step": 1100, "loss": 1.3672008514404297, "loss_mdlm": 1.3333079814910889, "loss_lm": 0.3389289379119873, "loss_ct": 0.0, "lr": 2.5e-05, "gnorm": 0.33943483233451843, "bridge_gate_avg": 1.0109125077724457, "bridge_out_proj_avg": 0.00235831766622141, "elapsed_s": 455.5775465965271}
|
| 13 |
+
{"step": 1200, "loss": 1.413881778717041, "loss_mdlm": 1.343287706375122, "loss_lm": 0.7059412598609924, "loss_ct": 0.0, "lr": 2.0658795558326743e-05, "gnorm": 0.46543946862220764, "bridge_gate_avg": 1.0112093389034271, "bridge_out_proj_avg": 0.0023823960800655186, "elapsed_s": 499.6888575553894}
|
| 14 |
+
{"step": 1300, "loss": 1.3803119659423828, "loss_mdlm": 1.3173803091049194, "loss_lm": 0.6293163299560547, "loss_ct": 0.0, "lr": 1.6449496416858284e-05, "gnorm": 0.5648322105407715, "bridge_gate_avg": 1.0114585757255554, "bridge_out_proj_avg": 0.002394759561866522, "elapsed_s": 543.8969376087189}
|
| 15 |
+
{"step": 1400, "loss": 1.3459510803222656, "loss_mdlm": 1.2746572494506836, "loss_lm": 0.7129377126693726, "loss_ct": 0.0, "lr": 1.2500000000000006e-05, "gnorm": 0.6428865790367126, "bridge_gate_avg": 1.0115884840488434, "bridge_out_proj_avg": 0.00240002857754007, "elapsed_s": 587.9394738674164}
|
| 16 |
+
{"step": 1500, "loss": 1.3825302124023438, "loss_mdlm": 1.3018296957015991, "loss_lm": 0.8070047497749329, "loss_ct": 0.0, "lr": 8.930309757836517e-06, "gnorm": 0.5085216760635376, "bridge_gate_avg": 1.0118375420570374, "bridge_out_proj_avg": 0.0024107499048113823, "elapsed_s": 632.1493453979492}
|
| 17 |
+
{"step": 1600, "loss": 1.4188456535339355, "loss_mdlm": 1.337924599647522, "loss_lm": 0.8092102408409119, "loss_ct": 0.0, "lr": 5.848888922025553e-06, "gnorm": 0.5302265882492065, "bridge_gate_avg": 1.0120702981948853, "bridge_out_proj_avg": 0.002423252852167934, "elapsed_s": 676.0981588363647}
|
| 18 |
+
{"step": 1700, "loss": 1.3620613813400269, "loss_mdlm": 1.3197686672210693, "loss_lm": 0.42292746901512146, "loss_ct": 0.0, "lr": 3.3493649053890326e-06, "gnorm": 0.4531289339065552, "bridge_gate_avg": 1.0121068358421326, "bridge_out_proj_avg": 0.0024279048084281385, "elapsed_s": 720.1140241622925}
|
| 19 |
+
{"step": 1800, "loss": 1.403516173362732, "loss_mdlm": 1.3171329498291016, "loss_lm": 0.8638321757316589, "loss_ct": 0.0, "lr": 1.5076844803522922e-06, "gnorm": 0.6531834006309509, "bridge_gate_avg": 1.0121425688266754, "bridge_out_proj_avg": 0.002430240041576326, "elapsed_s": 764.1136481761932}
|
| 20 |
+
{"step": 1900, "loss": 1.405068039894104, "loss_mdlm": 1.3349440097808838, "loss_lm": 0.701240062713623, "loss_ct": 0.0, "lr": 3.7980617469479953e-07, "gnorm": 0.5972647070884705, "bridge_gate_avg": 1.0121615827083588, "bridge_out_proj_avg": 0.002431624219752848, "elapsed_s": 808.3364908695221}
|