Upload folder using huggingface_hub

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb8939248f6742205c74e4d0b02750aefa474ef978b7722b5b8fef38a0ec24ef
+size 369838470

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c782658b6ced9b40c3e420262e6075c80e0b89ca5f921b8c190b2670f4a36dc
+size 369838470

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f0fea615f521b0c0202f69379c81d14254f0c54443f27ac4d19b87e5cf04878
+size 369838470

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ca21aa117d912df1e4f50d72ef9daa534c4fa4f403b33b7d3c58723b8d89193
+size 369838470

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cca4e099b36d6863ceb5d302655f1b458a1076208614c9329ae93f7ca42334f3
+size 369837282

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round20.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ed831f732e62bb1a6e285fdec0028bbf35eea8378c5df59cadd26625eebe684
+size 369838470

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5b4c04db8c5afb525937c3c48687f5e38ea4058e4ced0910bdb026e8b75afad
+size 369837282

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_client_model_round7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08b5c9a1cd78c0d3c2fa9485b5fd227f0630306d390bc9b6efd95f07e8275c0f
+size 369837282

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/0_trainer_state.json

@@ -0,0 +1,378 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 97,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.020618556701030927,
+      "grad_norm": 1.7885549068450928,
+      "learning_rate": 2e-05,
+      "loss": 1.3158,
+      "step": 2
+    },
+    {
+      "epoch": 0.041237113402061855,
+      "grad_norm": 3.5465033054351807,
+      "learning_rate": 2e-05,
+      "loss": 1.4782,
+      "step": 4
+    },
+    {
+      "epoch": 0.061855670103092786,
+      "grad_norm": 2.387897491455078,
+      "learning_rate": 2e-05,
+      "loss": 1.4943,
+      "step": 6
+    },
+    {
+      "epoch": 0.08247422680412371,
+      "grad_norm": 1.9758349657058716,
+      "learning_rate": 2e-05,
+      "loss": 1.6686,
+      "step": 8
+    },
+    {
+      "epoch": 0.10309278350515463,
+      "grad_norm": 1.9711047410964966,
+      "learning_rate": 2e-05,
+      "loss": 0.8049,
+      "step": 10
+    },
+    {
+      "epoch": 0.12371134020618557,
+      "grad_norm": 1.0105559825897217,
+      "learning_rate": 2e-05,
+      "loss": 0.3966,
+      "step": 12
+    },
+    {
+      "epoch": 0.14432989690721648,
+      "grad_norm": 1.9358932971954346,
+      "learning_rate": 2e-05,
+      "loss": 1.779,
+      "step": 14
+    },
+    {
+      "epoch": 0.16494845360824742,
+      "grad_norm": 0.6600197553634644,
+      "learning_rate": 2e-05,
+      "loss": 1.0865,
+      "step": 16
+    },
+    {
+      "epoch": 0.18556701030927836,
+      "grad_norm": 2.0319929122924805,
+      "learning_rate": 2e-05,
+      "loss": 1.0487,
+      "step": 18
+    },
+    {
+      "epoch": 0.20618556701030927,
+      "grad_norm": 1.7845250368118286,
+      "learning_rate": 2e-05,
+      "loss": 1.0636,
+      "step": 20
+    },
+    {
+      "epoch": 0.2268041237113402,
+      "grad_norm": 2.1108617782592773,
+      "learning_rate": 2e-05,
+      "loss": 0.7817,
+      "step": 22
+    },
+    {
+      "epoch": 0.24742268041237114,
+      "grad_norm": 0.574131190776825,
+      "learning_rate": 2e-05,
+      "loss": 0.4896,
+      "step": 24
+    },
+    {
+      "epoch": 0.26804123711340205,
+      "grad_norm": 4.136038303375244,
+      "learning_rate": 2e-05,
+      "loss": 1.6166,
+      "step": 26
+    },
+    {
+      "epoch": 0.28865979381443296,
+      "grad_norm": 4.69835090637207,
+      "learning_rate": 2e-05,
+      "loss": 0.5229,
+      "step": 28
+    },
+    {
+      "epoch": 0.30927835051546393,
+      "grad_norm": 0.6893218755722046,
+      "learning_rate": 2e-05,
+      "loss": 0.166,
+      "step": 30
+    },
+    {
+      "epoch": 0.32989690721649484,
+      "grad_norm": 0.9169237613677979,
+      "learning_rate": 2e-05,
+      "loss": 1.2709,
+      "step": 32
+    },
+    {
+      "epoch": 0.35051546391752575,
+      "grad_norm": 2.487178325653076,
+      "learning_rate": 2e-05,
+      "loss": 1.0059,
+      "step": 34
+    },
+    {
+      "epoch": 0.3711340206185567,
+      "grad_norm": 1.6276919841766357,
+      "learning_rate": 2e-05,
+      "loss": 1.1054,
+      "step": 36
+    },
+    {
+      "epoch": 0.3917525773195876,
+      "grad_norm": 3.6395184993743896,
+      "learning_rate": 2e-05,
+      "loss": 1.5661,
+      "step": 38
+    },
+    {
+      "epoch": 0.41237113402061853,
+      "grad_norm": 1.6149156093597412,
+      "learning_rate": 2e-05,
+      "loss": 1.0574,
+      "step": 40
+    },
+    {
+      "epoch": 0.4329896907216495,
+      "grad_norm": 3.7386510372161865,
+      "learning_rate": 2e-05,
+      "loss": 0.7777,
+      "step": 42
+    },
+    {
+      "epoch": 0.4536082474226804,
+      "grad_norm": 3.9839937686920166,
+      "learning_rate": 2e-05,
+      "loss": 0.912,
+      "step": 44
+    },
+    {
+      "epoch": 0.4742268041237113,
+      "grad_norm": 5.850351810455322,
+      "learning_rate": 2e-05,
+      "loss": 1.7357,
+      "step": 46
+    },
+    {
+      "epoch": 0.4948453608247423,
+      "grad_norm": 3.660813093185425,
+      "learning_rate": 2e-05,
+      "loss": 1.0512,
+      "step": 48
+    },
+    {
+      "epoch": 0.5154639175257731,
+      "grad_norm": 14.106647491455078,
+      "learning_rate": 2e-05,
+      "loss": 1.4082,
+      "step": 50
+    },
+    {
+      "epoch": 0.5360824742268041,
+      "grad_norm": 3.222626209259033,
+      "learning_rate": 2e-05,
+      "loss": 0.6463,
+      "step": 52
+    },
+    {
+      "epoch": 0.5567010309278351,
+      "grad_norm": 4.6796650886535645,
+      "learning_rate": 2e-05,
+      "loss": 1.3084,
+      "step": 54
+    },
+    {
+      "epoch": 0.5773195876288659,
+      "grad_norm": 2.90836238861084,
+      "learning_rate": 2e-05,
+      "loss": 1.3372,
+      "step": 56
+    },
+    {
+      "epoch": 0.5979381443298969,
+      "grad_norm": 3.991719961166382,
+      "learning_rate": 2e-05,
+      "loss": 1.3481,
+      "step": 58
+    },
+    {
+      "epoch": 0.6185567010309279,
+      "grad_norm": 1.9187226295471191,
+      "learning_rate": 2e-05,
+      "loss": 0.6076,
+      "step": 60
+    },
+    {
+      "epoch": 0.6391752577319587,
+      "grad_norm": 1.9427756071090698,
+      "learning_rate": 2e-05,
+      "loss": 0.6295,
+      "step": 62
+    },
+    {
+      "epoch": 0.6597938144329897,
+      "grad_norm": 2.391686201095581,
+      "learning_rate": 2e-05,
+      "loss": 1.135,
+      "step": 64
+    },
+    {
+      "epoch": 0.6804123711340206,
+      "grad_norm": 1.3331944942474365,
+      "learning_rate": 2e-05,
+      "loss": 1.3263,
+      "step": 66
+    },
+    {
+      "epoch": 0.7010309278350515,
+      "grad_norm": 0.4943288564682007,
+      "learning_rate": 2e-05,
+      "loss": 0.8955,
+      "step": 68
+    },
+    {
+      "epoch": 0.7216494845360825,
+      "grad_norm": 2.617217540740967,
+      "learning_rate": 2e-05,
+      "loss": 0.7329,
+      "step": 70
+    },
+    {
+      "epoch": 0.7422680412371134,
+      "grad_norm": 1.3934130668640137,
+      "learning_rate": 2e-05,
+      "loss": 1.1683,
+      "step": 72
+    },
+    {
+      "epoch": 0.7628865979381443,
+      "grad_norm": 0.945609450340271,
+      "learning_rate": 2e-05,
+      "loss": 0.6058,
+      "step": 74
+    },
+    {
+      "epoch": 0.7835051546391752,
+      "grad_norm": 1.7555233240127563,
+      "learning_rate": 2e-05,
+      "loss": 0.7444,
+      "step": 76
+    },
+    {
+      "epoch": 0.8041237113402062,
+      "grad_norm": 1.6663618087768555,
+      "learning_rate": 2e-05,
+      "loss": 0.6543,
+      "step": 78
+    },
+    {
+      "epoch": 0.8247422680412371,
+      "grad_norm": 2.9476871490478516,
+      "learning_rate": 2e-05,
+      "loss": 0.7212,
+      "step": 80
+    },
+    {
+      "epoch": 0.845360824742268,
+      "grad_norm": 4.171327590942383,
+      "learning_rate": 2e-05,
+      "loss": 1.0781,
+      "step": 82
+    },
+    {
+      "epoch": 0.865979381443299,
+      "grad_norm": 1.2571913003921509,
+      "learning_rate": 2e-05,
+      "loss": 1.1902,
+      "step": 84
+    },
+    {
+      "epoch": 0.8865979381443299,
+      "grad_norm": 2.324883222579956,
+      "learning_rate": 2e-05,
+      "loss": 1.4357,
+      "step": 86
+    },
+    {
+      "epoch": 0.9072164948453608,
+      "grad_norm": 1.769622564315796,
+      "learning_rate": 2e-05,
+      "loss": 0.6369,
+      "step": 88
+    },
+    {
+      "epoch": 0.9278350515463918,
+      "grad_norm": 1.4626708030700684,
+      "learning_rate": 2e-05,
+      "loss": 0.9478,
+      "step": 90
+    },
+    {
+      "epoch": 0.9484536082474226,
+      "grad_norm": 1.5452158451080322,
+      "learning_rate": 2e-05,
+      "loss": 1.5981,
+      "step": 92
+    },
+    {
+      "epoch": 0.9690721649484536,
+      "grad_norm": 1.4196434020996094,
+      "learning_rate": 2e-05,
+      "loss": 1.1368,
+      "step": 94
+    },
+    {
+      "epoch": 0.9896907216494846,
+      "grad_norm": 0.6402710676193237,
+      "learning_rate": 2e-05,
+      "loss": 0.5329,
+      "step": 96
+    },
+    {
+      "epoch": 1.0,
+      "step": 97,
+      "total_flos": 5756746026975232.0,
+      "train_loss": 1.0383215835413981,
+      "train_runtime": 283.7705,
+      "train_samples_per_second": 1.367,
+      "train_steps_per_second": 0.342
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 97,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5756746026975232.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:370b1db79ae79307d373e52fdcf5846624e6f6ad230ac369eea386c427fa22d5
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:786ed4479b1a809eb89d5a753dc89448dabc319918eb6df9d1f33dce7183d8bb
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e8c960fb7271256592ac9103b14201ea74c547f243dfaf14d27b6877499b29d
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:342f39308650a50e27c3384006998d380b4ee8782565da8a8c785cad37e94e9e
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e2933a0448457cc81da095354056445031a0197d4f32ede406420cdfff126ed
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round20.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f8621d8d5526426122df9113b89e3e07402e213bbca7ae98592879d054ce024
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d220fc720bc62776c37d47fd80f77cd33928fd8f26fecac46b6fb031eb4f03c
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_client_model_round7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1e2d3519b28a2f33deb11d7a2212f7fda010f1cad33e0a67eeb7ad4655b1db0
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/1_trainer_state.json

@@ -0,0 +1,378 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 97,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.020618556701030927,
+      "grad_norm": 1.8986397981643677,
+      "learning_rate": 2e-05,
+      "loss": 0.9165,
+      "step": 2
+    },
+    {
+      "epoch": 0.041237113402061855,
+      "grad_norm": 0.8982141017913818,
+      "learning_rate": 2e-05,
+      "loss": 0.7158,
+      "step": 4
+    },
+    {
+      "epoch": 0.061855670103092786,
+      "grad_norm": 1.1869165897369385,
+      "learning_rate": 2e-05,
+      "loss": 1.4353,
+      "step": 6
+    },
+    {
+      "epoch": 0.08247422680412371,
+      "grad_norm": 1.4538377523422241,
+      "learning_rate": 2e-05,
+      "loss": 0.4816,
+      "step": 8
+    },
+    {
+      "epoch": 0.10309278350515463,
+      "grad_norm": 1.2608749866485596,
+      "learning_rate": 2e-05,
+      "loss": 1.1197,
+      "step": 10
+    },
+    {
+      "epoch": 0.12371134020618557,
+      "grad_norm": 1.5531688928604126,
+      "learning_rate": 2e-05,
+      "loss": 1.1562,
+      "step": 12
+    },
+    {
+      "epoch": 0.14432989690721648,
+      "grad_norm": 2.4403159618377686,
+      "learning_rate": 2e-05,
+      "loss": 1.1279,
+      "step": 14
+    },
+    {
+      "epoch": 0.16494845360824742,
+      "grad_norm": 1.2706080675125122,
+      "learning_rate": 2e-05,
+      "loss": 1.3482,
+      "step": 16
+    },
+    {
+      "epoch": 0.18556701030927836,
+      "grad_norm": 0.8720521926879883,
+      "learning_rate": 2e-05,
+      "loss": 1.1394,
+      "step": 18
+    },
+    {
+      "epoch": 0.20618556701030927,
+      "grad_norm": 1.1683954000473022,
+      "learning_rate": 2e-05,
+      "loss": 0.8083,
+      "step": 20
+    },
+    {
+      "epoch": 0.2268041237113402,
+      "grad_norm": 1.2726197242736816,
+      "learning_rate": 2e-05,
+      "loss": 1.0113,
+      "step": 22
+    },
+    {
+      "epoch": 0.24742268041237114,
+      "grad_norm": 1.0261338949203491,
+      "learning_rate": 2e-05,
+      "loss": 0.6729,
+      "step": 24
+    },
+    {
+      "epoch": 0.26804123711340205,
+      "grad_norm": 5.61696195602417,
+      "learning_rate": 2e-05,
+      "loss": 1.2334,
+      "step": 26
+    },
+    {
+      "epoch": 0.28865979381443296,
+      "grad_norm": 1.71990168094635,
+      "learning_rate": 2e-05,
+      "loss": 0.9639,
+      "step": 28
+    },
+    {
+      "epoch": 0.30927835051546393,
+      "grad_norm": 2.458183526992798,
+      "learning_rate": 2e-05,
+      "loss": 1.0067,
+      "step": 30
+    },
+    {
+      "epoch": 0.32989690721649484,
+      "grad_norm": 2.6437785625457764,
+      "learning_rate": 2e-05,
+      "loss": 0.9413,
+      "step": 32
+    },
+    {
+      "epoch": 0.35051546391752575,
+      "grad_norm": 1.7311123609542847,
+      "learning_rate": 2e-05,
+      "loss": 0.5072,
+      "step": 34
+    },
+    {
+      "epoch": 0.3711340206185567,
+      "grad_norm": 1.6367141008377075,
+      "learning_rate": 2e-05,
+      "loss": 0.628,
+      "step": 36
+    },
+    {
+      "epoch": 0.3917525773195876,
+      "grad_norm": 0.9919836521148682,
+      "learning_rate": 2e-05,
+      "loss": 0.3894,
+      "step": 38
+    },
+    {
+      "epoch": 0.41237113402061853,
+      "grad_norm": 0.5254012942314148,
+      "learning_rate": 2e-05,
+      "loss": 0.5094,
+      "step": 40
+    },
+    {
+      "epoch": 0.4329896907216495,
+      "grad_norm": 1.6975247859954834,
+      "learning_rate": 2e-05,
+      "loss": 0.6632,
+      "step": 42
+    },
+    {
+      "epoch": 0.4536082474226804,
+      "grad_norm": 2.5024383068084717,
+      "learning_rate": 2e-05,
+      "loss": 1.181,
+      "step": 44
+    },
+    {
+      "epoch": 0.4742268041237113,
+      "grad_norm": 3.767259359359741,
+      "learning_rate": 2e-05,
+      "loss": 1.7095,
+      "step": 46
+    },
+    {
+      "epoch": 0.4948453608247423,
+      "grad_norm": 1.6667107343673706,
+      "learning_rate": 2e-05,
+      "loss": 1.007,
+      "step": 48
+    },
+    {
+      "epoch": 0.5154639175257731,
+      "grad_norm": 2.309662103652954,
+      "learning_rate": 2e-05,
+      "loss": 1.4642,
+      "step": 50
+    },
+    {
+      "epoch": 0.5360824742268041,
+      "grad_norm": 1.091732382774353,
+      "learning_rate": 2e-05,
+      "loss": 0.7542,
+      "step": 52
+    },
+    {
+      "epoch": 0.5567010309278351,
+      "grad_norm": 0.6196984648704529,
+      "learning_rate": 2e-05,
+      "loss": 0.5573,
+      "step": 54
+    },
+    {
+      "epoch": 0.5773195876288659,
+      "grad_norm": 0.402852863073349,
+      "learning_rate": 2e-05,
+      "loss": 0.6682,
+      "step": 56
+    },
+    {
+      "epoch": 0.5979381443298969,
+      "grad_norm": 0.7684454321861267,
+      "learning_rate": 2e-05,
+      "loss": 0.4199,
+      "step": 58
+    },
+    {
+      "epoch": 0.6185567010309279,
+      "grad_norm": 1.7286068201065063,
+      "learning_rate": 2e-05,
+      "loss": 1.2886,
+      "step": 60
+    },
+    {
+      "epoch": 0.6391752577319587,
+      "grad_norm": 1.2119078636169434,
+      "learning_rate": 2e-05,
+      "loss": 0.975,
+      "step": 62
+    },
+    {
+      "epoch": 0.6597938144329897,
+      "grad_norm": 1.0343464612960815,
+      "learning_rate": 2e-05,
+      "loss": 0.526,
+      "step": 64
+    },
+    {
+      "epoch": 0.6804123711340206,
+      "grad_norm": 3.151249885559082,
+      "learning_rate": 2e-05,
+      "loss": 0.9672,
+      "step": 66
+    },
+    {
+      "epoch": 0.7010309278350515,
+      "grad_norm": 2.050671339035034,
+      "learning_rate": 2e-05,
+      "loss": 1.3117,
+      "step": 68
+    },
+    {
+      "epoch": 0.7216494845360825,
+      "grad_norm": 2.2325823307037354,
+      "learning_rate": 2e-05,
+      "loss": 0.9194,
+      "step": 70
+    },
+    {
+      "epoch": 0.7422680412371134,
+      "grad_norm": 2.7051377296447754,
+      "learning_rate": 2e-05,
+      "loss": 1.7141,
+      "step": 72
+    },
+    {
+      "epoch": 0.7628865979381443,
+      "grad_norm": 0.7203123569488525,
+      "learning_rate": 2e-05,
+      "loss": 0.8371,
+      "step": 74
+    },
+    {
+      "epoch": 0.7835051546391752,
+      "grad_norm": 0.36563840508461,
+      "learning_rate": 2e-05,
+      "loss": 0.5306,
+      "step": 76
+    },
+    {
+      "epoch": 0.8041237113402062,
+      "grad_norm": 1.0923153162002563,
+      "learning_rate": 2e-05,
+      "loss": 0.6162,
+      "step": 78
+    },
+    {
+      "epoch": 0.8247422680412371,
+      "grad_norm": 1.4866068363189697,
+      "learning_rate": 2e-05,
+      "loss": 1.1174,
+      "step": 80
+    },
+    {
+      "epoch": 0.845360824742268,
+      "grad_norm": 1.3142611980438232,
+      "learning_rate": 2e-05,
+      "loss": 1.1617,
+      "step": 82
+    },
+    {
+      "epoch": 0.865979381443299,
+      "grad_norm": 0.786864161491394,
+      "learning_rate": 2e-05,
+      "loss": 1.0875,
+      "step": 84
+    },
+    {
+      "epoch": 0.8865979381443299,
+      "grad_norm": 5.453260898590088,
+      "learning_rate": 2e-05,
+      "loss": 2.6415,
+      "step": 86
+    },
+    {
+      "epoch": 0.9072164948453608,
+      "grad_norm": 2.76277494430542,
+      "learning_rate": 2e-05,
+      "loss": 1.6643,
+      "step": 88
+    },
+    {
+      "epoch": 0.9278350515463918,
+      "grad_norm": 3.1113247871398926,
+      "learning_rate": 2e-05,
+      "loss": 1.6608,
+      "step": 90
+    },
+    {
+      "epoch": 0.9484536082474226,
+      "grad_norm": 2.045165538787842,
+      "learning_rate": 2e-05,
+      "loss": 1.3748,
+      "step": 92
+    },
+    {
+      "epoch": 0.9690721649484536,
+      "grad_norm": 1.3238657712936401,
+      "learning_rate": 2e-05,
+      "loss": 0.6474,
+      "step": 94
+    },
+    {
+      "epoch": 0.9896907216494846,
+      "grad_norm": 1.4850274324417114,
+      "learning_rate": 2e-05,
+      "loss": 1.2191,
+      "step": 96
+    },
+    {
+      "epoch": 1.0,
+      "step": 97,
+      "total_flos": 1.357514687905792e+16,
+      "train_loss": 1.0192029107477247,
+      "train_runtime": 432.7985,
+      "train_samples_per_second": 0.896,
+      "train_steps_per_second": 0.224
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 97,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.357514687905792e+16,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10a34813c00d6ad8cdc583f8887781fecfb131343418ae94ec3447d7a3fe1bfc
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9db0446e3f8c533e3feae1f50596020e299611966289837d5806ffddf7b55f2e
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a49327d682e271891713c7c47e1696742b20027098eb662d19a166aa06442575
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d6f565813fc76f5b193fad10e0cfc5d676079a66425debe170b83c9cecef597
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6e1bbe1c61ba40ca48e147e953756de596d150add45ce56a952820f0a474fe7
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round20.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90abd315cf8aab026bf8f60c25bb92c532647720f2481a4be4d4a2828e93a7c5
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31a6eccd57cfc4753d17e9dbe0b1b2871b4163e9f9506aebfc91c68f77e059de
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_client_model_round7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8152affee06df72f20252a7a473eed4521a4660df6fce31047d2b6ebc30e069d
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/2_trainer_state.json

@@ -0,0 +1,378 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 97,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.020618556701030927,
+      "grad_norm": 0.9275849461555481,
+      "learning_rate": 2e-05,
+      "loss": 1.2235,
+      "step": 2
+    },
+    {
+      "epoch": 0.041237113402061855,
+      "grad_norm": 3.5949947834014893,
+      "learning_rate": 2e-05,
+      "loss": 0.6944,
+      "step": 4
+    },
+    {
+      "epoch": 0.061855670103092786,
+      "grad_norm": 2.8406832218170166,
+      "learning_rate": 2e-05,
+      "loss": 0.5023,
+      "step": 6
+    },
+    {
+      "epoch": 0.08247422680412371,
+      "grad_norm": 0.5248422622680664,
+      "learning_rate": 2e-05,
+      "loss": 0.4287,
+      "step": 8
+    },
+    {
+      "epoch": 0.10309278350515463,
+      "grad_norm": 3.409879684448242,
+      "learning_rate": 2e-05,
+      "loss": 1.3373,
+      "step": 10
+    },
+    {
+      "epoch": 0.12371134020618557,
+      "grad_norm": 1.4362713098526,
+      "learning_rate": 2e-05,
+      "loss": 1.4465,
+      "step": 12
+    },
+    {
+      "epoch": 0.14432989690721648,
+      "grad_norm": 1.8497318029403687,
+      "learning_rate": 2e-05,
+      "loss": 0.8547,
+      "step": 14
+    },
+    {
+      "epoch": 0.16494845360824742,
+      "grad_norm": 1.0694129467010498,
+      "learning_rate": 2e-05,
+      "loss": 0.8493,
+      "step": 16
+    },
+    {
+      "epoch": 0.18556701030927836,
+      "grad_norm": 1.5638664960861206,
+      "learning_rate": 2e-05,
+      "loss": 0.505,
+      "step": 18
+    },
+    {
+      "epoch": 0.20618556701030927,
+      "grad_norm": 1.5820540189743042,
+      "learning_rate": 2e-05,
+      "loss": 0.6212,
+      "step": 20
+    },
+    {
+      "epoch": 0.2268041237113402,
+      "grad_norm": 4.486837863922119,
+      "learning_rate": 2e-05,
+      "loss": 0.9427,
+      "step": 22
+    },
+    {
+      "epoch": 0.24742268041237114,
+      "grad_norm": 3.43084979057312,
+      "learning_rate": 2e-05,
+      "loss": 1.3259,
+      "step": 24
+    },
+    {
+      "epoch": 0.26804123711340205,
+      "grad_norm": 1.7316274642944336,
+      "learning_rate": 2e-05,
+      "loss": 1.0513,
+      "step": 26
+    },
+    {
+      "epoch": 0.28865979381443296,
+      "grad_norm": 1.9279751777648926,
+      "learning_rate": 2e-05,
+      "loss": 1.3768,
+      "step": 28
+    },
+    {
+      "epoch": 0.30927835051546393,
+      "grad_norm": 1.6935036182403564,
+      "learning_rate": 2e-05,
+      "loss": 1.3627,
+      "step": 30
+    },
+    {
+      "epoch": 0.32989690721649484,
+      "grad_norm": 2.276658296585083,
+      "learning_rate": 2e-05,
+      "loss": 0.7526,
+      "step": 32
+    },
+    {
+      "epoch": 0.35051546391752575,
+      "grad_norm": 1.0096880197525024,
+      "learning_rate": 2e-05,
+      "loss": 0.3399,
+      "step": 34
+    },
+    {
+      "epoch": 0.3711340206185567,
+      "grad_norm": 6.736175537109375,
+      "learning_rate": 2e-05,
+      "loss": 1.2885,
+      "step": 36
+    },
+    {
+      "epoch": 0.3917525773195876,
+      "grad_norm": 5.622121810913086,
+      "learning_rate": 2e-05,
+      "loss": 1.1467,
+      "step": 38
+    },
+    {
+      "epoch": 0.41237113402061853,
+      "grad_norm": 0.8184213638305664,
+      "learning_rate": 2e-05,
+      "loss": 0.444,
+      "step": 40
+    },
+    {
+      "epoch": 0.4329896907216495,
+      "grad_norm": 2.9587604999542236,
+      "learning_rate": 2e-05,
+      "loss": 1.0432,
+      "step": 42
+    },
+    {
+      "epoch": 0.4536082474226804,
+      "grad_norm": 1.1915258169174194,
+      "learning_rate": 2e-05,
+      "loss": 1.5732,
+      "step": 44
+    },
+    {
+      "epoch": 0.4742268041237113,
+      "grad_norm": 1.614179253578186,
+      "learning_rate": 2e-05,
+      "loss": 0.6172,
+      "step": 46
+    },
+    {
+      "epoch": 0.4948453608247423,
+      "grad_norm": 0.6889798045158386,
+      "learning_rate": 2e-05,
+      "loss": 0.745,
+      "step": 48
+    },
+    {
+      "epoch": 0.5154639175257731,
+      "grad_norm": 1.6385116577148438,
+      "learning_rate": 2e-05,
+      "loss": 0.5239,
+      "step": 50
+    },
+    {
+      "epoch": 0.5360824742268041,
+      "grad_norm": 2.3733930587768555,
+      "learning_rate": 2e-05,
+      "loss": 1.1497,
+      "step": 52
+    },
+    {
+      "epoch": 0.5567010309278351,
+      "grad_norm": 6.621945381164551,
+      "learning_rate": 2e-05,
+      "loss": 1.9435,
+      "step": 54
+    },
+    {
+      "epoch": 0.5773195876288659,
+      "grad_norm": 3.033151626586914,
+      "learning_rate": 2e-05,
+      "loss": 1.2542,
+      "step": 56
+    },
+    {
+      "epoch": 0.5979381443298969,
+      "grad_norm": 1.3858976364135742,
+      "learning_rate": 2e-05,
+      "loss": 1.1066,
+      "step": 58
+    },
+    {
+      "epoch": 0.6185567010309279,
+      "grad_norm": 1.2492917776107788,
+      "learning_rate": 2e-05,
+      "loss": 0.6909,
+      "step": 60
+    },
+    {
+      "epoch": 0.6391752577319587,
+      "grad_norm": 1.3139721155166626,
+      "learning_rate": 2e-05,
+      "loss": 1.3802,
+      "step": 62
+    },
+    {
+      "epoch": 0.6597938144329897,
+      "grad_norm": 0.7570559978485107,
+      "learning_rate": 2e-05,
+      "loss": 0.6115,
+      "step": 64
+    },
+    {
+      "epoch": 0.6804123711340206,
+      "grad_norm": 2.8449275493621826,
+      "learning_rate": 2e-05,
+      "loss": 0.8201,
+      "step": 66
+    },
+    {
+      "epoch": 0.7010309278350515,
+      "grad_norm": 1.3498934507369995,
+      "learning_rate": 2e-05,
+      "loss": 0.6134,
+      "step": 68
+    },
+    {
+      "epoch": 0.7216494845360825,
+      "grad_norm": 0.9374634027481079,
+      "learning_rate": 2e-05,
+      "loss": 0.896,
+      "step": 70
+    },
+    {
+      "epoch": 0.7422680412371134,
+      "grad_norm": 0.9612107872962952,
+      "learning_rate": 2e-05,
+      "loss": 0.7274,
+      "step": 72
+    },
+    {
+      "epoch": 0.7628865979381443,
+      "grad_norm": 1.954071044921875,
+      "learning_rate": 2e-05,
+      "loss": 0.9464,
+      "step": 74
+    },
+    {
+      "epoch": 0.7835051546391752,
+      "grad_norm": 1.1018635034561157,
+      "learning_rate": 2e-05,
+      "loss": 0.2764,
+      "step": 76
+    },
+    {
+      "epoch": 0.8041237113402062,
+      "grad_norm": 1.9673211574554443,
+      "learning_rate": 2e-05,
+      "loss": 0.7883,
+      "step": 78
+    },
+    {
+      "epoch": 0.8247422680412371,
+      "grad_norm": 1.7728729248046875,
+      "learning_rate": 2e-05,
+      "loss": 0.9561,
+      "step": 80
+    },
+    {
+      "epoch": 0.845360824742268,
+      "grad_norm": 0.83953857421875,
+      "learning_rate": 2e-05,
+      "loss": 0.3297,
+      "step": 82
+    },
+    {
+      "epoch": 0.865979381443299,
+      "grad_norm": 1.0060991048812866,
+      "learning_rate": 2e-05,
+      "loss": 0.5352,
+      "step": 84
+    },
+    {
+      "epoch": 0.8865979381443299,
+      "grad_norm": 2.0772080421447754,
+      "learning_rate": 2e-05,
+      "loss": 0.1712,
+      "step": 86
+    },
+    {
+      "epoch": 0.9072164948453608,
+      "grad_norm": 0.6878994703292847,
+      "learning_rate": 2e-05,
+      "loss": 0.84,
+      "step": 88
+    },
+    {
+      "epoch": 0.9278350515463918,
+      "grad_norm": 2.3218917846679688,
+      "learning_rate": 2e-05,
+      "loss": 0.9899,
+      "step": 90
+    },
+    {
+      "epoch": 0.9484536082474226,
+      "grad_norm": 1.3924381732940674,
+      "learning_rate": 2e-05,
+      "loss": 0.4655,
+      "step": 92
+    },
+    {
+      "epoch": 0.9690721649484536,
+      "grad_norm": 0.6978415846824646,
+      "learning_rate": 2e-05,
+      "loss": 0.4733,
+      "step": 94
+    },
+    {
+      "epoch": 0.9896907216494846,
+      "grad_norm": 1.8935520648956299,
+      "learning_rate": 2e-05,
+      "loss": 0.6506,
+      "step": 96
+    },
+    {
+      "epoch": 1.0,
+      "step": 97,
+      "total_flos": 1.549123199172608e+16,
+      "train_loss": 0.8593464733399067,
+      "train_runtime": 470.9244,
+      "train_samples_per_second": 0.824,
+      "train_steps_per_second": 0.206
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 97,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.549123199172608e+16,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ff896230f57e27e7efb28914991eb6d19a9aec05291e19038cdb761e2ef5f6
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:080373777ac3c4300e13d3fd2e643161a026404ca2b58cf8b03d824c241bed6e
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0819ec21ef683084401b7b83249b2199b0896526a18b20cd8bd4b69774c7678a
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e6f125278992e54fb10a7579a5b264da2311e679cdf452181e06ef1d1f2811f
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c908af596f0443c233324475fdbe5601d5df6b5600b93c3e742622ca3ee6519
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round20.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86bfa4042903b9f2ca17b83f019f4cbf0153327097c41c3f88184fc545403474
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9558c665e69591d78ecc3a8312b4ebe20d4baa013d0ddb21d6d69f77ab76bc4
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_client_model_round7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cfdf0b1c8e3ca781e7795f6201dae68670ec0b64fe67c316dd74300d3e2bfe2b
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/3_trainer_state.json

@@ -0,0 +1,378 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 97,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.020618556701030927,
+      "grad_norm": 5.465381622314453,
+      "learning_rate": 2e-05,
+      "loss": 0.8918,
+      "step": 2
+    },
+    {
+      "epoch": 0.041237113402061855,
+      "grad_norm": 1.7941179275512695,
+      "learning_rate": 2e-05,
+      "loss": 0.7903,
+      "step": 4
+    },
+    {
+      "epoch": 0.061855670103092786,
+      "grad_norm": 1.6007481813430786,
+      "learning_rate": 2e-05,
+      "loss": 0.8676,
+      "step": 6
+    },
+    {
+      "epoch": 0.08247422680412371,
+      "grad_norm": 0.9969778656959534,
+      "learning_rate": 2e-05,
+      "loss": 0.4928,
+      "step": 8
+    },
+    {
+      "epoch": 0.10309278350515463,
+      "grad_norm": 0.4684109389781952,
+      "learning_rate": 2e-05,
+      "loss": 0.2868,
+      "step": 10
+    },
+    {
+      "epoch": 0.12371134020618557,
+      "grad_norm": 0.28731948137283325,
+      "learning_rate": 2e-05,
+      "loss": 0.2312,
+      "step": 12
+    },
+    {
+      "epoch": 0.14432989690721648,
+      "grad_norm": 1.3333783149719238,
+      "learning_rate": 2e-05,
+      "loss": 0.8662,
+      "step": 14
+    },
+    {
+      "epoch": 0.16494845360824742,
+      "grad_norm": 1.9330962896347046,
+      "learning_rate": 2e-05,
+      "loss": 0.8947,
+      "step": 16
+    },
+    {
+      "epoch": 0.18556701030927836,
+      "grad_norm": 2.568071126937866,
+      "learning_rate": 2e-05,
+      "loss": 1.0742,
+      "step": 18
+    },
+    {
+      "epoch": 0.20618556701030927,
+      "grad_norm": 1.4679968357086182,
+      "learning_rate": 2e-05,
+      "loss": 1.5971,
+      "step": 20
+    },
+    {
+      "epoch": 0.2268041237113402,
+      "grad_norm": 0.4212726950645447,
+      "learning_rate": 2e-05,
+      "loss": 0.6416,
+      "step": 22
+    },
+    {
+      "epoch": 0.24742268041237114,
+      "grad_norm": 1.8013954162597656,
+      "learning_rate": 2e-05,
+      "loss": 0.8423,
+      "step": 24
+    },
+    {
+      "epoch": 0.26804123711340205,
+      "grad_norm": 1.1382532119750977,
+      "learning_rate": 2e-05,
+      "loss": 0.7341,
+      "step": 26
+    },
+    {
+      "epoch": 0.28865979381443296,
+      "grad_norm": 1.0272934436798096,
+      "learning_rate": 2e-05,
+      "loss": 0.5432,
+      "step": 28
+    },
+    {
+      "epoch": 0.30927835051546393,
+      "grad_norm": 0.6234714984893799,
+      "learning_rate": 2e-05,
+      "loss": 0.2671,
+      "step": 30
+    },
+    {
+      "epoch": 0.32989690721649484,
+      "grad_norm": 1.1389981508255005,
+      "learning_rate": 2e-05,
+      "loss": 0.7717,
+      "step": 32
+    },
+    {
+      "epoch": 0.35051546391752575,
+      "grad_norm": 5.075913906097412,
+      "learning_rate": 2e-05,
+      "loss": 1.3037,
+      "step": 34
+    },
+    {
+      "epoch": 0.3711340206185567,
+      "grad_norm": 1.1098531484603882,
+      "learning_rate": 2e-05,
+      "loss": 0.8804,
+      "step": 36
+    },
+    {
+      "epoch": 0.3917525773195876,
+      "grad_norm": 1.1887625455856323,
+      "learning_rate": 2e-05,
+      "loss": 0.7009,
+      "step": 38
+    },
+    {
+      "epoch": 0.41237113402061853,
+      "grad_norm": 0.27039554715156555,
+      "learning_rate": 2e-05,
+      "loss": 0.4646,
+      "step": 40
+    },
+    {
+      "epoch": 0.4329896907216495,
+      "grad_norm": 1.437915325164795,
+      "learning_rate": 2e-05,
+      "loss": 0.3613,
+      "step": 42
+    },
+    {
+      "epoch": 0.4536082474226804,
+      "grad_norm": 1.6871410608291626,
+      "learning_rate": 2e-05,
+      "loss": 0.4272,
+      "step": 44
+    },
+    {
+      "epoch": 0.4742268041237113,
+      "grad_norm": 4.205522060394287,
+      "learning_rate": 2e-05,
+      "loss": 1.3826,
+      "step": 46
+    },
+    {
+      "epoch": 0.4948453608247423,
+      "grad_norm": 1.2980058193206787,
+      "learning_rate": 2e-05,
+      "loss": 0.8113,
+      "step": 48
+    },
+    {
+      "epoch": 0.5154639175257731,
+      "grad_norm": 0.826991856098175,
+      "learning_rate": 2e-05,
+      "loss": 0.3983,
+      "step": 50
+    },
+    {
+      "epoch": 0.5360824742268041,
+      "grad_norm": 1.9655853509902954,
+      "learning_rate": 2e-05,
+      "loss": 1.2521,
+      "step": 52
+    },
+    {
+      "epoch": 0.5567010309278351,
+      "grad_norm": 1.085252046585083,
+      "learning_rate": 2e-05,
+      "loss": 0.636,
+      "step": 54
+    },
+    {
+      "epoch": 0.5773195876288659,
+      "grad_norm": 1.889613389968872,
+      "learning_rate": 2e-05,
+      "loss": 0.6425,
+      "step": 56
+    },
+    {
+      "epoch": 0.5979381443298969,
+      "grad_norm": 2.9438934326171875,
+      "learning_rate": 2e-05,
+      "loss": 0.6987,
+      "step": 58
+    },
+    {
+      "epoch": 0.6185567010309279,
+      "grad_norm": 0.5686755180358887,
+      "learning_rate": 2e-05,
+      "loss": 0.267,
+      "step": 60
+    },
+    {
+      "epoch": 0.6391752577319587,
+      "grad_norm": 0.5724007487297058,
+      "learning_rate": 2e-05,
+      "loss": 0.3444,
+      "step": 62
+    },
+    {
+      "epoch": 0.6597938144329897,
+      "grad_norm": 1.3709577322006226,
+      "learning_rate": 2e-05,
+      "loss": 0.6371,
+      "step": 64
+    },
+    {
+      "epoch": 0.6804123711340206,
+      "grad_norm": 1.0185165405273438,
+      "learning_rate": 2e-05,
+      "loss": 0.8028,
+      "step": 66
+    },
+    {
+      "epoch": 0.7010309278350515,
+      "grad_norm": 1.3173937797546387,
+      "learning_rate": 2e-05,
+      "loss": 0.6559,
+      "step": 68
+    },
+    {
+      "epoch": 0.7216494845360825,
+      "grad_norm": 0.3681342601776123,
+      "learning_rate": 2e-05,
+      "loss": 0.1988,
+      "step": 70
+    },
+    {
+      "epoch": 0.7422680412371134,
+      "grad_norm": 0.9298673272132874,
+      "learning_rate": 2e-05,
+      "loss": 0.814,
+      "step": 72
+    },
+    {
+      "epoch": 0.7628865979381443,
+      "grad_norm": 1.2699919939041138,
+      "learning_rate": 2e-05,
+      "loss": 0.5768,
+      "step": 74
+    },
+    {
+      "epoch": 0.7835051546391752,
+      "grad_norm": 2.6978492736816406,
+      "learning_rate": 2e-05,
+      "loss": 1.0957,
+      "step": 76
+    },
+    {
+      "epoch": 0.8041237113402062,
+      "grad_norm": 1.3678396940231323,
+      "learning_rate": 2e-05,
+      "loss": 0.8172,
+      "step": 78
+    },
+    {
+      "epoch": 0.8247422680412371,
+      "grad_norm": 5.189119815826416,
+      "learning_rate": 2e-05,
+      "loss": 1.1335,
+      "step": 80
+    },
+    {
+      "epoch": 0.845360824742268,
+      "grad_norm": 2.356233835220337,
+      "learning_rate": 2e-05,
+      "loss": 1.126,
+      "step": 82
+    },
+    {
+      "epoch": 0.865979381443299,
+      "grad_norm": 1.0532094240188599,
+      "learning_rate": 2e-05,
+      "loss": 0.8347,
+      "step": 84
+    },
+    {
+      "epoch": 0.8865979381443299,
+      "grad_norm": 1.6790062189102173,
+      "learning_rate": 2e-05,
+      "loss": 0.6408,
+      "step": 86
+    },
+    {
+      "epoch": 0.9072164948453608,
+      "grad_norm": 4.085683345794678,
+      "learning_rate": 2e-05,
+      "loss": 0.595,
+      "step": 88
+    },
+    {
+      "epoch": 0.9278350515463918,
+      "grad_norm": 1.189156174659729,
+      "learning_rate": 2e-05,
+      "loss": 0.5897,
+      "step": 90
+    },
+    {
+      "epoch": 0.9484536082474226,
+      "grad_norm": 0.6069275736808777,
+      "learning_rate": 2e-05,
+      "loss": 0.9475,
+      "step": 92
+    },
+    {
+      "epoch": 0.9690721649484536,
+      "grad_norm": 3.666023015975952,
+      "learning_rate": 2e-05,
+      "loss": 0.8139,
+      "step": 94
+    },
+    {
+      "epoch": 0.9896907216494846,
+      "grad_norm": 0.7568903565406799,
+      "learning_rate": 2e-05,
+      "loss": 0.9121,
+      "step": 96
+    },
+    {
+      "epoch": 1.0,
+      "step": 97,
+      "total_flos": 2.07833438093312e+16,
+      "train_loss": 0.7429608340115891,
+      "train_runtime": 455.0442,
+      "train_samples_per_second": 0.853,
+      "train_steps_per_second": 0.213
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 97,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.07833438093312e+16,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98ae67cc7881795e67b6b2b24fff7b8372ee549b84f1cae4ab8b30d72b55a8e4
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26ba73bae8933df7f2d31878fa576a49fe213349cefbc417808909fcf7658148
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e290980a99d775d0947e011fdc7eb7166501c54469affe28e434f19f8c81aca5
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:343b3641c285b2619787a13de4844305fb426038dfa5499950938cf558a1e0f3
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d3ca7c1aba23a71baaefbbd626c8ed37c1beaf5e02b7459cea8d2f4d7527fa5
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round20.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90ff72f18aefb6d0946800b285fe88bd5a4b105c202f638c2524d9faf80b227b
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6441cb6bb9cb2a114399447e3474b20f41a3917f358ef38fe90739cef6d7396e
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_client_model_round7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77bdd5373faef04e2ffd2c1529d9059c48384f3eeb54d52b9c84555745b54f78
+size 794706058

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/4_trainer_state.json

@@ -0,0 +1,378 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 97,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.020618556701030927,
+      "grad_norm": 0.9477136731147766,
+      "learning_rate": 2e-05,
+      "loss": 0.551,
+      "step": 2
+    },
+    {
+      "epoch": 0.041237113402061855,
+      "grad_norm": 3.8200013637542725,
+      "learning_rate": 2e-05,
+      "loss": 1.1141,
+      "step": 4
+    },
+    {
+      "epoch": 0.061855670103092786,
+      "grad_norm": 0.16243354976177216,
+      "learning_rate": 2e-05,
+      "loss": 0.376,
+      "step": 6
+    },
+    {
+      "epoch": 0.08247422680412371,
+      "grad_norm": 1.3112159967422485,
+      "learning_rate": 2e-05,
+      "loss": 0.7677,
+      "step": 8
+    },
+    {
+      "epoch": 0.10309278350515463,
+      "grad_norm": 2.1373825073242188,
+      "learning_rate": 2e-05,
+      "loss": 1.256,
+      "step": 10
+    },
+    {
+      "epoch": 0.12371134020618557,
+      "grad_norm": 2.1234893798828125,
+      "learning_rate": 2e-05,
+      "loss": 0.8886,
+      "step": 12
+    },
+    {
+      "epoch": 0.14432989690721648,
+      "grad_norm": 0.04594890773296356,
+      "learning_rate": 2e-05,
+      "loss": 0.7081,
+      "step": 14
+    },
+    {
+      "epoch": 0.16494845360824742,
+      "grad_norm": 1.3787109851837158,
+      "learning_rate": 2e-05,
+      "loss": 0.828,
+      "step": 16
+    },
+    {
+      "epoch": 0.18556701030927836,
+      "grad_norm": 0.6793889999389648,
+      "learning_rate": 2e-05,
+      "loss": 1.0967,
+      "step": 18
+    },
+    {
+      "epoch": 0.20618556701030927,
+      "grad_norm": 0.771903932094574,
+      "learning_rate": 2e-05,
+      "loss": 1.6246,
+      "step": 20
+    },
+    {
+      "epoch": 0.2268041237113402,
+      "grad_norm": 1.1218531131744385,
+      "learning_rate": 2e-05,
+      "loss": 0.8997,
+      "step": 22
+    },
+    {
+      "epoch": 0.24742268041237114,
+      "grad_norm": 1.01219642162323,
+      "learning_rate": 2e-05,
+      "loss": 0.8123,
+      "step": 24
+    },
+    {
+      "epoch": 0.26804123711340205,
+      "grad_norm": 1.0736132860183716,
+      "learning_rate": 2e-05,
+      "loss": 0.8224,
+      "step": 26
+    },
+    {
+      "epoch": 0.28865979381443296,
+      "grad_norm": 1.2845109701156616,
+      "learning_rate": 2e-05,
+      "loss": 1.2574,
+      "step": 28
+    },
+    {
+      "epoch": 0.30927835051546393,
+      "grad_norm": 1.4137431383132935,
+      "learning_rate": 2e-05,
+      "loss": 1.6976,
+      "step": 30
+    },
+    {
+      "epoch": 0.32989690721649484,
+      "grad_norm": 1.9815335273742676,
+      "learning_rate": 2e-05,
+      "loss": 1.0386,
+      "step": 32
+    },
+    {
+      "epoch": 0.35051546391752575,
+      "grad_norm": 2.5250496864318848,
+      "learning_rate": 2e-05,
+      "loss": 1.5427,
+      "step": 34
+    },
+    {
+      "epoch": 0.3711340206185567,
+      "grad_norm": 0.8500065207481384,
+      "learning_rate": 2e-05,
+      "loss": 0.9216,
+      "step": 36
+    },
+    {
+      "epoch": 0.3917525773195876,
+      "grad_norm": 1.1719383001327515,
+      "learning_rate": 2e-05,
+      "loss": 1.0847,
+      "step": 38
+    },
+    {
+      "epoch": 0.41237113402061853,
+      "grad_norm": 3.1414475440979004,
+      "learning_rate": 2e-05,
+      "loss": 1.0832,
+      "step": 40
+    },
+    {
+      "epoch": 0.4329896907216495,
+      "grad_norm": 2.8507297039031982,
+      "learning_rate": 2e-05,
+      "loss": 1.0086,
+      "step": 42
+    },
+    {
+      "epoch": 0.4536082474226804,
+      "grad_norm": 2.59656023979187,
+      "learning_rate": 2e-05,
+      "loss": 1.4416,
+      "step": 44
+    },
+    {
+      "epoch": 0.4742268041237113,
+      "grad_norm": 1.6737186908721924,
+      "learning_rate": 2e-05,
+      "loss": 1.2538,
+      "step": 46
+    },
+    {
+      "epoch": 0.4948453608247423,
+      "grad_norm": 0.5178043246269226,
+      "learning_rate": 2e-05,
+      "loss": 0.2702,
+      "step": 48
+    },
+    {
+      "epoch": 0.5154639175257731,
+      "grad_norm": 1.748957633972168,
+      "learning_rate": 2e-05,
+      "loss": 1.0557,
+      "step": 50
+    },
+    {
+      "epoch": 0.5360824742268041,
+      "grad_norm": 1.7886401414871216,
+      "learning_rate": 2e-05,
+      "loss": 0.6472,
+      "step": 52
+    },
+    {
+      "epoch": 0.5567010309278351,
+      "grad_norm": 1.393855094909668,
+      "learning_rate": 2e-05,
+      "loss": 0.791,
+      "step": 54
+    },
+    {
+      "epoch": 0.5773195876288659,
+      "grad_norm": 1.5556002855300903,
+      "learning_rate": 2e-05,
+      "loss": 1.1448,
+      "step": 56
+    },
+    {
+      "epoch": 0.5979381443298969,
+      "grad_norm": 2.578552484512329,
+      "learning_rate": 2e-05,
+      "loss": 1.3115,
+      "step": 58
+    },
+    {
+      "epoch": 0.6185567010309279,
+      "grad_norm": 1.4991388320922852,
+      "learning_rate": 2e-05,
+      "loss": 0.8843,
+      "step": 60
+    },
+    {
+      "epoch": 0.6391752577319587,
+      "grad_norm": 1.8561006784439087,
+      "learning_rate": 2e-05,
+      "loss": 0.5295,
+      "step": 62
+    },
+    {
+      "epoch": 0.6597938144329897,
+      "grad_norm": 1.5938780307769775,
+      "learning_rate": 2e-05,
+      "loss": 0.5507,
+      "step": 64
+    },
+    {
+      "epoch": 0.6804123711340206,
+      "grad_norm": 2.841743230819702,
+      "learning_rate": 2e-05,
+      "loss": 1.1595,
+      "step": 66
+    },
+    {
+      "epoch": 0.7010309278350515,
+      "grad_norm": 1.491766095161438,
+      "learning_rate": 2e-05,
+      "loss": 0.6423,
+      "step": 68
+    },
+    {
+      "epoch": 0.7216494845360825,
+      "grad_norm": 1.3314822912216187,
+      "learning_rate": 2e-05,
+      "loss": 1.0332,
+      "step": 70
+    },
+    {
+      "epoch": 0.7422680412371134,
+      "grad_norm": 2.0549938678741455,
+      "learning_rate": 2e-05,
+      "loss": 2.0647,
+      "step": 72
+    },
+    {
+      "epoch": 0.7628865979381443,
+      "grad_norm": 1.329912781715393,
+      "learning_rate": 2e-05,
+      "loss": 0.5477,
+      "step": 74
+    },
+    {
+      "epoch": 0.7835051546391752,
+      "grad_norm": 0.7718127369880676,
+      "learning_rate": 2e-05,
+      "loss": 0.5856,
+      "step": 76
+    },
+    {
+      "epoch": 0.8041237113402062,
+      "grad_norm": 2.3758749961853027,
+      "learning_rate": 2e-05,
+      "loss": 0.7453,
+      "step": 78
+    },
+    {
+      "epoch": 0.8247422680412371,
+      "grad_norm": 3.360313892364502,
+      "learning_rate": 2e-05,
+      "loss": 1.9905,
+      "step": 80
+    },
+    {
+      "epoch": 0.845360824742268,
+      "grad_norm": 0.17600414156913757,
+      "learning_rate": 2e-05,
+      "loss": 0.5286,
+      "step": 82
+    },
+    {
+      "epoch": 0.865979381443299,
+      "grad_norm": 1.8083326816558838,
+      "learning_rate": 2e-05,
+      "loss": 0.6053,
+      "step": 84
+    },
+    {
+      "epoch": 0.8865979381443299,
+      "grad_norm": 0.21756744384765625,
+      "learning_rate": 2e-05,
+      "loss": 0.0233,
+      "step": 86
+    },
+    {
+      "epoch": 0.9072164948453608,
+      "grad_norm": 1.2804031372070312,
+      "learning_rate": 2e-05,
+      "loss": 1.3146,
+      "step": 88
+    },
+    {
+      "epoch": 0.9278350515463918,
+      "grad_norm": 0.4804908037185669,
+      "learning_rate": 2e-05,
+      "loss": 0.3251,
+      "step": 90
+    },
+    {
+      "epoch": 0.9484536082474226,
+      "grad_norm": 2.686302900314331,
+      "learning_rate": 2e-05,
+      "loss": 1.2593,
+      "step": 92
+    },
+    {
+      "epoch": 0.9690721649484536,
+      "grad_norm": 0.10516992211341858,
+      "learning_rate": 2e-05,
+      "loss": 0.943,
+      "step": 94
+    },
+    {
+      "epoch": 0.9896907216494846,
+      "grad_norm": 3.6283299922943115,
+      "learning_rate": 2e-05,
+      "loss": 0.768,
+      "step": 96
+    },
+    {
+      "epoch": 1.0,
+      "step": 97,
+      "total_flos": 2.055998584704205e+16,
+      "train_loss": 0.9511987253562691,
+      "train_runtime": 446.2144,
+      "train_samples_per_second": 0.87,
+      "train_steps_per_second": 0.217
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 97,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.055998584704205e+16,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/5_client_model_round10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:855cbde8acac797b45d2371a9a3a0617b6d33f8d3ac9f8faf754a91f4aa333ab
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/5_client_model_round12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b282f874d4137ecb39abe65b92973331e82f116025a6173ce936ffac843e21a
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/5_client_model_round15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c2207f2441529cebfb2c6a7d55a8c4712bd0e01d8715dd171287deefbd34fe9
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/5_client_model_round17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63d312918f1bfb9e1258851b1c99e9c3086b5c321c9b716922766e69510cec70
+size 794708086

client_states_feddualMultipqfull_homoAgg_moe2_NEWAensureOrth_Bcca_T05_freq10_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr97_T0125_decay099_SEED2/5_client_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22ee8edfb9eaa1c31e1e1cb53250581d93bee9375df9ce52701adf79d658b652
+size 794706058