Upload exp_phase8_bridge_recipeABC_KD_204305/log.jsonl with huggingface_hub
Browse files
exp_phase8_bridge_recipeABC_KD_204305/log.jsonl
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"step": 0, "loss": 1.4725391864776611, "loss_mdlm": 1.2873576879501343, "loss_lm": 1.8518153429031372, "loss_ct": 0.0, "lr": 0.0, "gnorm": 0.5914597511291504, "bridge_gate_avg": 1.0, "bridge_out_proj_avg": 0.0, "elapsed_s": 4.052710771560669}
|
| 2 |
+
{"step": 100, "loss": 1.4585211277008057, "loss_mdlm": 1.3423362970352173, "loss_lm": 1.161848783493042, "loss_ct": 0.0, "lr": 2.5e-05, "gnorm": 0.6844421029090881, "bridge_gate_avg": 1.0009054243564606, "bridge_out_proj_avg": 0.000427968188887462, "elapsed_s": 50.265472650527954}
|
| 3 |
+
{"step": 200, "loss": 1.4572480916976929, "loss_mdlm": 1.3045012950897217, "loss_lm": 1.527467966079712, "loss_ct": 0.0, "lr": 5e-05, "gnorm": 0.8587601780891418, "bridge_gate_avg": 1.0033626556396484, "bridge_out_proj_avg": 0.0015961033350322396, "elapsed_s": 94.93989419937134}
|
| 4 |
+
{"step": 300, "loss": 1.3833705186843872, "loss_mdlm": 1.315598964691162, "loss_lm": 0.6777155995368958, "loss_ct": 0.0, "lr": 4.962019382530521e-05, "gnorm": 0.6005806922912598, "bridge_gate_avg": 1.0042824149131775, "bridge_out_proj_avg": 0.0018480518774595112, "elapsed_s": 139.2527949810028}
|
| 5 |
+
{"step": 400, "loss": 1.4433640241622925, "loss_mdlm": 1.3518251180648804, "loss_lm": 0.9153895378112793, "loss_ct": 0.0, "lr": 4.849231551964771e-05, "gnorm": 0.6869295835494995, "bridge_gate_avg": 1.0050093829631805, "bridge_out_proj_avg": 0.001974337617866695, "elapsed_s": 183.92998552322388}
|