diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..2a25eb6e5dea05fc6d7eeb1dc7401c51819256b2 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d7a4d0d6615d2510711f99c6b666e858a2598446d3263d37eed691a52c48cafa +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..7007a542a6962885544e725358788d2b80b3cc23 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8467983a8faf53ac385f579b43f293895165ec3553aa52dcb326de610583d6a0 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..6e26b4f18aa1d6ccd42701e16e0c091a53b8ab7a --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5bf18123b293dbbba941833be776184ede92ebbb3304e9365d5bf968ecc50b6 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..a7e8a1cf524b670a073bb33a53d7a9e3b48621b1 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35940a8c8eddd6b7454570d0e6c61016292d9d2e5197647f71eb2ddcdb093c33 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..e3664fadc0fb01371cfc56d33fe7a3fd02825b24 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea4a6fb5a88281d0acef0944c1cc603e502dfe0190a9f8f36296a7c9476d566a +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..466abf7f2099e9cd91771f863c16c66915f94dc0 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:468aaa38b2dbdf861f538b48b6969482f6d8913aeac15e921782031762a86c89 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..696fc4e855f5ed2ee1a109bfa10b4aa6f62542a2 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eed2f8b54154e62827fc7ccb130532f5449940421549b41b08b74e89ff8a3b4c +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..09f74252b193170acb7def6ee9f6376a1cd7caba --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:81889dbecd6a4ea0784812af1ca62a08caa07cf01b024f7335fd93dd392681c9 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..a8064438d0e2d384d8685d667a6ad109d0090679 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/0_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 1.4507830142974854, + "learning_rate": 2e-05, + "loss": 0.2925, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 3.2782809734344482, + "learning_rate": 2e-05, + "loss": 0.5226, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 3.6665804386138916, + "learning_rate": 2e-05, + "loss": 0.4469, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 6.7659735679626465, + "learning_rate": 2e-05, + "loss": 0.6084, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 2.461778163909912, + "learning_rate": 2e-05, + "loss": 0.6173, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 0.8903517723083496, + "learning_rate": 2e-05, + "loss": 0.095, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 4.321805000305176, + "learning_rate": 2e-05, + "loss": 0.4503, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 2.4119670391082764, + "learning_rate": 2e-05, + "loss": 0.3114, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 4.858500003814697, + "learning_rate": 2e-05, + "loss": 0.3571, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 3.7888195514678955, + "learning_rate": 2e-05, + "loss": 0.8946, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 0.9685637950897217, + "learning_rate": 2e-05, + "loss": 0.1218, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 6.1121015548706055, + "learning_rate": 2e-05, + "loss": 0.5919, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 5.337240219116211, + "learning_rate": 2e-05, + "loss": 0.366, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 4.253252983093262, + "learning_rate": 2e-05, + "loss": 0.4007, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 8.691818237304688, + "learning_rate": 2e-05, + "loss": 0.9219, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 4.221813201904297, + "learning_rate": 2e-05, + "loss": 0.3515, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 6.260042667388916, + "learning_rate": 2e-05, + "loss": 0.8002, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 11.410113334655762, + "learning_rate": 2e-05, + "loss": 1.451, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 3.6883702278137207, + "learning_rate": 2e-05, + "loss": 0.3644, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 2.033090353012085, + "learning_rate": 2e-05, + "loss": 0.2821, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 6.383613109588623, + "learning_rate": 2e-05, + "loss": 1.392, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 2.7196388244628906, + "learning_rate": 2e-05, + "loss": 0.1335, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 9.112273216247559, + "learning_rate": 2e-05, + "loss": 1.9183, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 4.017404556274414, + "learning_rate": 2e-05, + "loss": 0.5077, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 0.14292454719543457, + "learning_rate": 2e-05, + "loss": 0.323, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 2.870044469833374, + "learning_rate": 2e-05, + "loss": 0.679, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 7.838882923126221, + "learning_rate": 2e-05, + "loss": 0.5556, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 1.5267337560653687, + "learning_rate": 2e-05, + "loss": 0.6559, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 4.8004374504089355, + "learning_rate": 2e-05, + "loss": 1.1925, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 0.9164693355560303, + "learning_rate": 2e-05, + "loss": 0.3563, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 10.06078815460205, + "learning_rate": 2e-05, + "loss": 1.1676, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 4.685103416442871, + "learning_rate": 2e-05, + "loss": 0.4408, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 2.2955524921417236, + "learning_rate": 2e-05, + "loss": 0.1538, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 2.193854331970215, + "learning_rate": 2e-05, + "loss": 0.8281, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 0.9019300937652588, + "learning_rate": 2e-05, + "loss": 1.1189, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 10.440985679626465, + "learning_rate": 2e-05, + "loss": 2.1768, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 2.911909818649292, + "learning_rate": 2e-05, + "loss": 0.7279, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 0.03667999058961868, + "learning_rate": 2e-05, + "loss": 0.115, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 2.2721805572509766, + "learning_rate": 2e-05, + "loss": 0.1444, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 2.097175121307373, + "learning_rate": 2e-05, + "loss": 0.2108, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 0.7106596231460571, + "learning_rate": 2e-05, + "loss": 0.0699, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 2.563626766204834, + "learning_rate": 2e-05, + "loss": 0.1482, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 0.34837377071380615, + "learning_rate": 2e-05, + "loss": 0.0672, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 0.41047564148902893, + "learning_rate": 2e-05, + "loss": 0.3899, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 0.7772120833396912, + "learning_rate": 2e-05, + "loss": 0.0882, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 0.9385977387428284, + "learning_rate": 2e-05, + "loss": 1.0358, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 6.794859886169434, + "learning_rate": 2e-05, + "loss": 1.1388, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 8.23221492767334, + "learning_rate": 2e-05, + "loss": 0.3678, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 0.18338899314403534, + "learning_rate": 2e-05, + "loss": 0.2155, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 0.3215850591659546, + "learning_rate": 2e-05, + "loss": 0.0276, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 2053446886752256.0, + "train_loss": 0.5718875336647034, + "train_runtime": 100.3201, + "train_samples_per_second": 3.987, + "train_steps_per_second": 0.997 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 2053446886752256.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..db00391011632f3c34fd71c5768004ad6dca60cf --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b9983a1b6c779a4d52cdc627f04f956d61efac456204bcc4dabada42d9a3dfdd +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..9c7ce530e08ef9fa45d27c63b4a86975a9894548 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f2de129e62beeb4971790b5e1f46c5a709dcf59503e9475b5787bf0cd581368e +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..71c0f272f86a4b635a965c583fa3caa041d8e662 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:24ed512b6dcd1a66d95d053edadd3e3fbdf255ba36752a7ade6a545f7a8c1654 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..63874013be2f365a5e55520b5a42ca8d204418c2 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4219da9ea85c1442677e975ed2d3f527a357245f0998ca76a59bd078777e3fd +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..b407ba54ccec5dd4f999f74bfac13fbf02f1aef2 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1ceea20bd9bb93c4ec0fd7ee4511cfb48fbb6ab8dac82e3d4240b59dba7c865 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..4aa8dcc6c2dfd85d1a5a0cb6371e30ce75f33649 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fac5c439ec99857d8e543ff416a2fd33e4648cf7a54164d4ff91a2487ca1202e +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..17760ac09d496470bcdcff8937233f395b1ef46d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ca3cc56d5d8ef6bf8ceb6673227c86c6fdfa5c11342ee85b31de747c9ee1296 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..2e60eeb3e0a4e2afc7cd0a9dfd726e39078ec954 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bf892a9fd738f4567b9929fca404c0a7a73c02558fbe6ed6dc4303c2e51f42b8 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..58ab4ae4356ad419590c10b59520125560f4f647 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/1_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 4.540980339050293, + "learning_rate": 2e-05, + "loss": 0.422, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 0.08287880569696426, + "learning_rate": 2e-05, + "loss": 0.0048, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 3.329127788543701, + "learning_rate": 2e-05, + "loss": 0.2607, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 0.11668016016483307, + "learning_rate": 2e-05, + "loss": 0.0061, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 1.266418218612671, + "learning_rate": 2e-05, + "loss": 0.0361, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 1.0858792066574097, + "learning_rate": 2e-05, + "loss": 0.0498, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 1.4955309629440308, + "learning_rate": 2e-05, + "loss": 0.0903, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 3.658975839614868, + "learning_rate": 2e-05, + "loss": 0.4419, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 0.029739608988165855, + "learning_rate": 2e-05, + "loss": 0.0057, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 0.04279458895325661, + "learning_rate": 2e-05, + "loss": 0.0027, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 1.1662931442260742, + "learning_rate": 2e-05, + "loss": 0.0642, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 1.0573233366012573, + "learning_rate": 2e-05, + "loss": 0.0731, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 4.016772747039795, + "learning_rate": 2e-05, + "loss": 0.3845, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 0.2739083766937256, + "learning_rate": 2e-05, + "loss": 0.0173, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 0.004112154711037874, + "learning_rate": 2e-05, + "loss": 0.5049, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 1.817938208580017, + "learning_rate": 2e-05, + "loss": 0.1084, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 0.040277149528265, + "learning_rate": 2e-05, + "loss": 0.0015, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 0.03433386608958244, + "learning_rate": 2e-05, + "loss": 0.0028, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 3.3714048862457275, + "learning_rate": 2e-05, + "loss": 0.1396, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 0.03902699798345566, + "learning_rate": 2e-05, + "loss": 0.0027, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 0.17097237706184387, + "learning_rate": 2e-05, + "loss": 0.006, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 0.004333522170782089, + "learning_rate": 2e-05, + "loss": 0.0081, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 0.13571485877037048, + "learning_rate": 2e-05, + "loss": 0.006, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 0.31877508759498596, + "learning_rate": 2e-05, + "loss": 0.0122, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 0.09019943326711655, + "learning_rate": 2e-05, + "loss": 0.0051, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 10.819501876831055, + "learning_rate": 2e-05, + "loss": 1.0651, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 0.5622828602790833, + "learning_rate": 2e-05, + "loss": 0.0183, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 0.3623190224170685, + "learning_rate": 2e-05, + "loss": 0.0462, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 0.40045422315597534, + "learning_rate": 2e-05, + "loss": 0.1874, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 0.0034316980745643377, + "learning_rate": 2e-05, + "loss": 0.0095, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 0.01208802405744791, + "learning_rate": 2e-05, + "loss": 0.0055, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 0.01345849223434925, + "learning_rate": 2e-05, + "loss": 0.0033, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 0.6658453941345215, + "learning_rate": 2e-05, + "loss": 0.0433, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 2.384672164916992, + "learning_rate": 2e-05, + "loss": 0.2239, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 0.07835771143436432, + "learning_rate": 2e-05, + "loss": 0.0073, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 1.4132821559906006, + "learning_rate": 2e-05, + "loss": 0.0383, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 0.9961208701133728, + "learning_rate": 2e-05, + "loss": 0.0747, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 0.9463048577308655, + "learning_rate": 2e-05, + "loss": 0.0433, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 0.01034417562186718, + "learning_rate": 2e-05, + "loss": 0.0328, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 0.0005425452254712582, + "learning_rate": 2e-05, + "loss": 0.3519, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 0.04632725194096565, + "learning_rate": 2e-05, + "loss": 0.0369, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 1.2183465957641602, + "learning_rate": 2e-05, + "loss": 0.0654, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 0.009153845719993114, + "learning_rate": 2e-05, + "loss": 0.0014, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 0.8759830594062805, + "learning_rate": 2e-05, + "loss": 0.0918, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 0.031464505940675735, + "learning_rate": 2e-05, + "loss": 0.0014, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 0.038530562072992325, + "learning_rate": 2e-05, + "loss": 0.0023, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 0.04050149768590927, + "learning_rate": 2e-05, + "loss": 0.0215, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 0.009601338766515255, + "learning_rate": 2e-05, + "loss": 0.001, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 0.02644410915672779, + "learning_rate": 2e-05, + "loss": 0.0018, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 13.793920516967773, + "learning_rate": 2e-05, + "loss": 0.8695, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 2069576665792512.0, + "train_loss": 0.11801103919744492, + "train_runtime": 100.2591, + "train_samples_per_second": 3.99, + "train_steps_per_second": 0.997 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 2069576665792512.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..9b55926545780630df5c5a6df43d0b6edac5541b --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03582a65c5f942f662a8eafdf27eccb06167fe68a235376a577c83b61464520f +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..3b5b7a65a6858eee7bbc9496e9f8c514a7617d09 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f25cc37e7b0ba30193079fe56dc4c3aa9b011c90f9b100c4edb8af121e24a47 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..eeeaa64d51373aeec1ee9cf0d28919d99c76fc77 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3e73eded9bd524f1dce2d82e0dcd45a995aec36460f360989c7f0a3633f9064b +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..6be278d7ba8b0aba5fd339f9f44927b149739960 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4f015a9439fe81810ee686e3def97f0a5628eb03eeb6647f2361e14acb54c6f1 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..4673182e016c82fc52f237cdf3c18e89d202bf1f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f1d2a8a90b6e9f82b7cd6b9d41f2426100a7a948c8be73b9cd03822077c3984 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..db1ab3234c2796852cc6366c38776e8e94645e46 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff05ff0a2fd64497a4dc87555b334386e1340bf51e0cb26b9563e5605c8b120e +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..b621b982feacfd65f69f5e6e741ca0f6259ef65f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:abb0d68b551454c4d09eebc6d9d9cedb828570c4d6224ed8d949afef34928fd5 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..4324bfdf465581b30b06907d5f743eb3a327bf50 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e31665b8c7d78dc560352ba149d087420a03bb9929818edc19a4016a5f1b8ba +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..741be070478d0eca042c3ab9ea9c2e2c1432007a --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/2_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 6.307283401489258, + "learning_rate": 2e-05, + "loss": 1.5954, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 1.8216760158538818, + "learning_rate": 2e-05, + "loss": 0.278, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 7.7118024826049805, + "learning_rate": 2e-05, + "loss": 1.2105, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 6.2828192710876465, + "learning_rate": 2e-05, + "loss": 1.0291, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 1.8280245065689087, + "learning_rate": 2e-05, + "loss": 0.3982, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 0.9868451356887817, + "learning_rate": 2e-05, + "loss": 1.2771, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 1.3482744693756104, + "learning_rate": 2e-05, + "loss": 0.6548, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 1.1375393867492676, + "learning_rate": 2e-05, + "loss": 0.6873, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 1.463499903678894, + "learning_rate": 2e-05, + "loss": 0.6526, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 5.554112911224365, + "learning_rate": 2e-05, + "loss": 1.2926, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 1.0732632875442505, + "learning_rate": 2e-05, + "loss": 0.461, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 2.5138278007507324, + "learning_rate": 2e-05, + "loss": 0.7348, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 2.534879446029663, + "learning_rate": 2e-05, + "loss": 0.8571, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 2.698505163192749, + "learning_rate": 2e-05, + "loss": 0.521, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 1.4479676485061646, + "learning_rate": 2e-05, + "loss": 0.3902, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 1.2816400527954102, + "learning_rate": 2e-05, + "loss": 0.4518, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 5.400059223175049, + "learning_rate": 2e-05, + "loss": 0.9999, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 2.5884220600128174, + "learning_rate": 2e-05, + "loss": 0.3547, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 2.3713040351867676, + "learning_rate": 2e-05, + "loss": 0.7269, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 0.25373563170433044, + "learning_rate": 2e-05, + "loss": 0.0746, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 4.246715068817139, + "learning_rate": 2e-05, + "loss": 0.6696, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 1.6761362552642822, + "learning_rate": 2e-05, + "loss": 0.6276, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 3.399017572402954, + "learning_rate": 2e-05, + "loss": 0.9067, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 0.692711591720581, + "learning_rate": 2e-05, + "loss": 0.1943, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 1.116909146308899, + "learning_rate": 2e-05, + "loss": 0.2336, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 1.456255555152893, + "learning_rate": 2e-05, + "loss": 0.1465, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 0.06800251454114914, + "learning_rate": 2e-05, + "loss": 0.2881, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 4.372195720672607, + "learning_rate": 2e-05, + "loss": 0.6529, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 4.281543254852295, + "learning_rate": 2e-05, + "loss": 1.313, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 3.5601418018341064, + "learning_rate": 2e-05, + "loss": 0.3216, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 1.4214577674865723, + "learning_rate": 2e-05, + "loss": 0.1358, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 2.7352545261383057, + "learning_rate": 2e-05, + "loss": 0.4594, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 3.6488208770751953, + "learning_rate": 2e-05, + "loss": 1.1297, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 0.18936391174793243, + "learning_rate": 2e-05, + "loss": 0.2297, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 0.24005119502544403, + "learning_rate": 2e-05, + "loss": 1.2341, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 0.6526658535003662, + "learning_rate": 2e-05, + "loss": 0.4384, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 1.2854243516921997, + "learning_rate": 2e-05, + "loss": 0.2658, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 1.8762297630310059, + "learning_rate": 2e-05, + "loss": 0.5192, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 8.563094139099121, + "learning_rate": 2e-05, + "loss": 0.3001, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 0.5608336329460144, + "learning_rate": 2e-05, + "loss": 0.2274, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 0.9091539978981018, + "learning_rate": 2e-05, + "loss": 0.2695, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 1.8625966310501099, + "learning_rate": 2e-05, + "loss": 0.2104, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 3.4447784423828125, + "learning_rate": 2e-05, + "loss": 0.5892, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 0.39084768295288086, + "learning_rate": 2e-05, + "loss": 0.2646, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 4.247977256774902, + "learning_rate": 2e-05, + "loss": 0.3688, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 0.8638454079627991, + "learning_rate": 2e-05, + "loss": 2.1514, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 0.26021599769592285, + "learning_rate": 2e-05, + "loss": 0.0341, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 2.017536163330078, + "learning_rate": 2e-05, + "loss": 0.2675, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 2.068211793899536, + "learning_rate": 2e-05, + "loss": 0.256, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 6.009552478790283, + "learning_rate": 2e-05, + "loss": 0.8954, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 4915713315700736.0, + "train_loss": 0.6049574661254883, + "train_runtime": 165.4178, + "train_samples_per_second": 2.418, + "train_steps_per_second": 0.605 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 4915713315700736.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..f6b564f1e8da8698f7fa71b0f2edba8a6aa6a8b4 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e8ec9a76aa2a365afb2403325260e12dec27ad4b984f60c9e7f5a9f76ebbdad1 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..89d1502d2eb031cf1b9a8d1881d3d936241401db --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0a23a2ed8e03fd3a4b21c1fea3e90f4eb1c0a3a99bedee884c388ca90b15211 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..b9d19d5c51d89ffdf104c3aa20730ba1bdca1dc1 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:70680ac43b83326240e7835655831bc4ceaf9f0de955840f5c6d0c1f4afcc2f1 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..5b32bb54daf509ff3da43ddc81f39aa53c76c647 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e4b81455da8eff19455dde4b74de4297a754f18d10b63e691a538c6b9946a3f +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..7ad536e1f1d32cf80188d53c49b1a9b75726b80c --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af9b733c0813536ef08508eb156501dcb5bd61a3091f43cb3f3f6010fe4b4917 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..a69789f0a2986a48364c5c0b5032c00eaf66ea18 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e03eca3be62c25b2a2aa30e419a67ca025067a04b2c6b04bf385daaae97ddbd +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..b1328682893c27be99f730d87faf59c37f65584c --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3787e3f8abd12f49c6d80d3cdcdede3a2f53808976bc36599bbc4428eea38f7 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..c5dd7859c40f0b2dd77323ff0a21d9550b95f129 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6adb60b55506bfdd98859d59018a85cefb09d6e024b8fa128b3beb888dc5315 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..28b491f5495906109fa063be7d79e2555540057d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/3_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 2.0921337604522705, + "learning_rate": 2e-05, + "loss": 0.6894, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 4.510809421539307, + "learning_rate": 2e-05, + "loss": 0.6862, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 4.531048774719238, + "learning_rate": 2e-05, + "loss": 1.2015, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 3.931663751602173, + "learning_rate": 2e-05, + "loss": 0.5979, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 6.292150020599365, + "learning_rate": 2e-05, + "loss": 1.058, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 4.606175422668457, + "learning_rate": 2e-05, + "loss": 0.8743, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 3.087507963180542, + "learning_rate": 2e-05, + "loss": 0.6243, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 8.593318939208984, + "learning_rate": 2e-05, + "loss": 0.4402, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 5.68522834777832, + "learning_rate": 2e-05, + "loss": 1.3279, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 1.9175809621810913, + "learning_rate": 2e-05, + "loss": 0.7962, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 8.656201362609863, + "learning_rate": 2e-05, + "loss": 2.0415, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 1.99509859085083, + "learning_rate": 2e-05, + "loss": 0.2472, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 7.396963119506836, + "learning_rate": 2e-05, + "loss": 0.8892, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 5.109688758850098, + "learning_rate": 2e-05, + "loss": 0.4696, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 3.3529930114746094, + "learning_rate": 2e-05, + "loss": 0.6545, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 1.4711360931396484, + "learning_rate": 2e-05, + "loss": 0.5218, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 4.621761322021484, + "learning_rate": 2e-05, + "loss": 0.9679, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 3.575629472732544, + "learning_rate": 2e-05, + "loss": 0.5973, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 9.701947212219238, + "learning_rate": 2e-05, + "loss": 1.4922, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 18.55413818359375, + "learning_rate": 2e-05, + "loss": 1.1465, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 7.169260025024414, + "learning_rate": 2e-05, + "loss": 0.9151, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 7.758701324462891, + "learning_rate": 2e-05, + "loss": 1.3213, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 7.772892475128174, + "learning_rate": 2e-05, + "loss": 0.4431, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 3.169400215148926, + "learning_rate": 2e-05, + "loss": 0.9307, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 6.446079730987549, + "learning_rate": 2e-05, + "loss": 1.4162, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 4.463912487030029, + "learning_rate": 2e-05, + "loss": 0.8901, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 7.208362102508545, + "learning_rate": 2e-05, + "loss": 1.4832, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 11.086746215820312, + "learning_rate": 2e-05, + "loss": 2.0512, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 6.340665340423584, + "learning_rate": 2e-05, + "loss": 1.0799, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 2.8643109798431396, + "learning_rate": 2e-05, + "loss": 0.5789, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 7.802653789520264, + "learning_rate": 2e-05, + "loss": 1.8293, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 4.388473033905029, + "learning_rate": 2e-05, + "loss": 1.0891, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 4.431455612182617, + "learning_rate": 2e-05, + "loss": 0.9788, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 0.8305065035820007, + "learning_rate": 2e-05, + "loss": 0.6982, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 14.678071975708008, + "learning_rate": 2e-05, + "loss": 1.9248, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 2.8914413452148438, + "learning_rate": 2e-05, + "loss": 0.5398, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 2.2980284690856934, + "learning_rate": 2e-05, + "loss": 1.6722, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 4.637665271759033, + "learning_rate": 2e-05, + "loss": 1.3349, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 2.5008809566497803, + "learning_rate": 2e-05, + "loss": 1.134, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 2.6805694103240967, + "learning_rate": 2e-05, + "loss": 0.935, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 7.456270694732666, + "learning_rate": 2e-05, + "loss": 1.5703, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 3.0763790607452393, + "learning_rate": 2e-05, + "loss": 0.3613, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 8.599203109741211, + "learning_rate": 2e-05, + "loss": 0.9858, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 7.865464687347412, + "learning_rate": 2e-05, + "loss": 1.5265, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 7.330792427062988, + "learning_rate": 2e-05, + "loss": 1.392, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 5.383084774017334, + "learning_rate": 2e-05, + "loss": 1.248, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 6.0959649085998535, + "learning_rate": 2e-05, + "loss": 1.9657, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 2.155146598815918, + "learning_rate": 2e-05, + "loss": 0.8844, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 5.614543437957764, + "learning_rate": 2e-05, + "loss": 1.4389, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 8.140727043151855, + "learning_rate": 2e-05, + "loss": 1.3601, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 2098644891205632.0, + "train_loss": 1.066049919128418, + "train_runtime": 99.4099, + "train_samples_per_second": 4.024, + "train_steps_per_second": 1.006 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 2098644891205632.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..cca68ac750f3fb876304e59db97a8fead41cef72 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3506ffc3670759125f1d1c08b6420c1bf38b2f186b977f9cbb77ec8fcbeea88e +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..cd542f2749721f634870290c46b61a4c343cb827 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7be4b6efdf8aeda1d55eab73770ec3ff6a9a14501a2aa60e833ec969ed086533 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..98204b2e8bde4cafe44d7d8a726be93572fa2f2d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:90585364faf7ecfbee025db3a036cc013caf849e1bed9b8c5d1c4570c6065777 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..c65dafb3b84b43828a0680ca4a0a1aea61747ad8 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9236b8e15ebe274a58931ff42727de3f819bbb0f8493426399ff78877af9de2b +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..774759b0f41f8d74ea2a844f0ff4cba63f826604 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a161db119e3dd4b0ad54e2dc5dae4212100d532c0ec87749b3eb4d37f395bf48 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..b9053a693c629b555763dc3263ce1bec78ebb336 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16b3859505775fae6ab29fd2053ab49c87ca8a84e54cda0d5304f7c72b066afd +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..fd82af6f3acfca1d4d6bf49cd3665672ba75fec4 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:05e363377b0c5548371d96debaaf6fff6fa705a7ca1321530510e181d3556b7e +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..1e6006349891f46181fcd1bfbf3ea91d91f7c2a5 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9cd95080fae073060aa659888b415d2135edd62af172653cfed0f99ab94464f6 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..da4b45dcc2a46689cd87705eb3e729064cf17107 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/4_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 3.6383607387542725, + "learning_rate": 2e-05, + "loss": 0.9771, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 1.4260591268539429, + "learning_rate": 2e-05, + "loss": 0.5006, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 1.6014289855957031, + "learning_rate": 2e-05, + "loss": 0.782, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 2.321310043334961, + "learning_rate": 2e-05, + "loss": 0.3546, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 1.937113881111145, + "learning_rate": 2e-05, + "loss": 1.2421, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 2.4271795749664307, + "learning_rate": 2e-05, + "loss": 1.005, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 0.9630257487297058, + "learning_rate": 2e-05, + "loss": 0.635, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 2.273303985595703, + "learning_rate": 2e-05, + "loss": 0.9471, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 1.193511724472046, + "learning_rate": 2e-05, + "loss": 0.6021, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 2.983036518096924, + "learning_rate": 2e-05, + "loss": 0.6765, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 1.626178503036499, + "learning_rate": 2e-05, + "loss": 1.0259, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 1.8325883150100708, + "learning_rate": 2e-05, + "loss": 0.5989, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 2.876636266708374, + "learning_rate": 2e-05, + "loss": 0.8092, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 2.8380074501037598, + "learning_rate": 2e-05, + "loss": 0.5194, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 2.227553129196167, + "learning_rate": 2e-05, + "loss": 0.6971, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 2.7015953063964844, + "learning_rate": 2e-05, + "loss": 0.6297, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 3.0312774181365967, + "learning_rate": 2e-05, + "loss": 1.3067, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 2.7130300998687744, + "learning_rate": 2e-05, + "loss": 0.702, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 0.3681621849536896, + "learning_rate": 2e-05, + "loss": 0.1352, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 0.8578549027442932, + "learning_rate": 2e-05, + "loss": 0.1746, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 1.2145777940750122, + "learning_rate": 2e-05, + "loss": 0.62, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 0.8366961479187012, + "learning_rate": 2e-05, + "loss": 0.237, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 3.8259220123291016, + "learning_rate": 2e-05, + "loss": 1.1652, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 2.451016426086426, + "learning_rate": 2e-05, + "loss": 0.2307, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 7.296032905578613, + "learning_rate": 2e-05, + "loss": 0.6684, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 2.7998406887054443, + "learning_rate": 2e-05, + "loss": 0.5629, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 1.9942561388015747, + "learning_rate": 2e-05, + "loss": 0.4002, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 3.880875587463379, + "learning_rate": 2e-05, + "loss": 0.5188, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 5.010158538818359, + "learning_rate": 2e-05, + "loss": 0.6704, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 6.04047155380249, + "learning_rate": 2e-05, + "loss": 0.8029, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 3.4900665283203125, + "learning_rate": 2e-05, + "loss": 0.9319, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 1.4386566877365112, + "learning_rate": 2e-05, + "loss": 0.392, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 2.075533866882324, + "learning_rate": 2e-05, + "loss": 1.0757, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 5.302359104156494, + "learning_rate": 2e-05, + "loss": 1.017, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 3.4823062419891357, + "learning_rate": 2e-05, + "loss": 1.1879, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 0.5839654803276062, + "learning_rate": 2e-05, + "loss": 0.1357, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 3.582629442214966, + "learning_rate": 2e-05, + "loss": 1.5566, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 1.8023282289505005, + "learning_rate": 2e-05, + "loss": 0.513, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 0.7795921564102173, + "learning_rate": 2e-05, + "loss": 0.8818, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 2.052372932434082, + "learning_rate": 2e-05, + "loss": 0.6126, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 1.606740117073059, + "learning_rate": 2e-05, + "loss": 0.5911, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 1.955881953239441, + "learning_rate": 2e-05, + "loss": 0.6584, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 2.623976469039917, + "learning_rate": 2e-05, + "loss": 0.5529, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 2.0439250469207764, + "learning_rate": 2e-05, + "loss": 0.9415, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 0.6250275373458862, + "learning_rate": 2e-05, + "loss": 0.1703, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 2.0065391063690186, + "learning_rate": 2e-05, + "loss": 0.6163, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 3.4686758518218994, + "learning_rate": 2e-05, + "loss": 1.1005, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 2.2346303462982178, + "learning_rate": 2e-05, + "loss": 0.8502, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 3.8489227294921875, + "learning_rate": 2e-05, + "loss": 1.3295, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 2.55889892578125, + "learning_rate": 2e-05, + "loss": 0.634, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 5692822418096128.0, + "train_loss": 0.71892333984375, + "train_runtime": 165.0653, + "train_samples_per_second": 2.423, + "train_steps_per_second": 0.606 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 5692822418096128.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..0a98a4acd9385398d99f1f74f8fa6a037a71ea95 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:04c033bed86fd20d1bf5f45a017a7235285089bde6b845487e1d24f22528cb78 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..2a10cadd1e2812997b77789aa4b6a73d2b232546 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3aa0ede5f53ed6efb703e4de0c88d026168e8444ff6938342da4a8d4b540d5df +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..250766c333e4f4534186ec0734989cd858a4337d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c172229e49be46b6d28ec7d39d0dec151b6a3955e703aca7df6e3bab423f73b +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..0502f5c36de6939732dbd93355d7b5e943e9cd75 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1781b39f8ac95a428751a6de041f2ef6a523a42fcaff84daefa48b7c5e687690 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..397fc4b33215b9af3368ebb861eadfb18dda12cc --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:204e4484268b051e95249cd0b0d80d5fc1c8b309c758d33fca781ee24da8090f +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..6b958dc77c2a6fcc6636c9ab54e7c3fd3329a3ee --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d2201c7622da8c7055538ec209eb33e571a2cb40b39a265916d1f3a25804cef7 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..eaf20c1f523936811c33a3e1dd2e9baf482083f3 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:329c4d6662a81e263fd214896528ab8116d33464d8580754dfa2da77fee106ed +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..63f73b13cba73d3c877943939fe6c60beb94b61a --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:859880a9c0877a27f55851e8fe5f9ab2db56f0ecd6d9ec783800802643baf81c +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..38cd825af8c69d1c5bc094dc5b113a945f22ba87 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/5_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 8.129776000976562, + "learning_rate": 2e-05, + "loss": 1.4123, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 7.267589569091797, + "learning_rate": 2e-05, + "loss": 1.555, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 1.5580347776412964, + "learning_rate": 2e-05, + "loss": 0.0764, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 3.456258773803711, + "learning_rate": 2e-05, + "loss": 0.1247, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 2.054414749145508, + "learning_rate": 2e-05, + "loss": 0.2661, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 1.959441065788269, + "learning_rate": 2e-05, + "loss": 0.1191, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 1.5240881443023682, + "learning_rate": 2e-05, + "loss": 1.3779, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 1.9770129919052124, + "learning_rate": 2e-05, + "loss": 0.369, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 1.607049584388733, + "learning_rate": 2e-05, + "loss": 0.225, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 2.911104440689087, + "learning_rate": 2e-05, + "loss": 0.5344, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 0.14911149442195892, + "learning_rate": 2e-05, + "loss": 0.2073, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 0.3970499336719513, + "learning_rate": 2e-05, + "loss": 0.0627, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 6.610349655151367, + "learning_rate": 2e-05, + "loss": 1.4124, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 6.448593616485596, + "learning_rate": 2e-05, + "loss": 1.7906, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 0.1567811667919159, + "learning_rate": 2e-05, + "loss": 0.1779, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 3.029139995574951, + "learning_rate": 2e-05, + "loss": 0.4042, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 0.1624116450548172, + "learning_rate": 2e-05, + "loss": 0.1239, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 1.771087408065796, + "learning_rate": 2e-05, + "loss": 0.1957, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 2.283267021179199, + "learning_rate": 2e-05, + "loss": 0.3572, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 2.6100361347198486, + "learning_rate": 2e-05, + "loss": 0.7945, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 0.4068808853626251, + "learning_rate": 2e-05, + "loss": 0.1732, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 0.47189369797706604, + "learning_rate": 2e-05, + "loss": 0.2693, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 1.0255151987075806, + "learning_rate": 2e-05, + "loss": 0.1383, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 0.07138078659772873, + "learning_rate": 2e-05, + "loss": 0.8398, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 0.021043304353952408, + "learning_rate": 2e-05, + "loss": 0.0042, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 0.32961738109588623, + "learning_rate": 2e-05, + "loss": 0.0314, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 3.3906283378601074, + "learning_rate": 2e-05, + "loss": 0.1685, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 0.9188004732131958, + "learning_rate": 2e-05, + "loss": 0.4201, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 0.7056655287742615, + "learning_rate": 2e-05, + "loss": 0.1079, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 0.8054582476615906, + "learning_rate": 2e-05, + "loss": 0.0658, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 5.253702640533447, + "learning_rate": 2e-05, + "loss": 1.7994, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 1.857879877090454, + "learning_rate": 2e-05, + "loss": 0.7767, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 0.037640877068042755, + "learning_rate": 2e-05, + "loss": 0.0685, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 0.6612502336502075, + "learning_rate": 2e-05, + "loss": 0.0833, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 0.6361874938011169, + "learning_rate": 2e-05, + "loss": 0.0205, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 0.1627129316329956, + "learning_rate": 2e-05, + "loss": 0.031, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 1.4068979024887085, + "learning_rate": 2e-05, + "loss": 0.1573, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 4.893678188323975, + "learning_rate": 2e-05, + "loss": 0.9358, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 1.9647676944732666, + "learning_rate": 2e-05, + "loss": 0.1192, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 0.21145711839199066, + "learning_rate": 2e-05, + "loss": 0.0185, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 0.061686694622039795, + "learning_rate": 2e-05, + "loss": 0.2625, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 0.19522836804389954, + "learning_rate": 2e-05, + "loss": 0.0139, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 0.095303013920784, + "learning_rate": 2e-05, + "loss": 0.0361, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 0.7677666544914246, + "learning_rate": 2e-05, + "loss": 0.0939, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 0.13143938779830933, + "learning_rate": 2e-05, + "loss": 0.5931, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 1.7637028694152832, + "learning_rate": 2e-05, + "loss": 1.2071, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 3.308182954788208, + "learning_rate": 2e-05, + "loss": 0.7625, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 0.0632585808634758, + "learning_rate": 2e-05, + "loss": 1.0905, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 1.9383233785629272, + "learning_rate": 2e-05, + "loss": 0.161, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 10.960673332214355, + "learning_rate": 2e-05, + "loss": 1.1678, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 4977468595240960.0, + "train_loss": 0.4640686821937561, + "train_runtime": 164.7987, + "train_samples_per_second": 2.427, + "train_steps_per_second": 0.607 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 4977468595240960.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..0a4b4811ed5501aef7f274afc3db28410920fd4f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4879eeea57bb6b6bbf271a44d5b847f3115d5817decbbc03e8a509912f82e237 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..d00d6c9cbfa1d3fa131246aeae79bd158ba92853 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e100169baaa0c65402fd8bb5719aba208659ff22bd5b1431cca0ceda9f4f427 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..8936f0497ae71ca20213a9a905b5b0628c96530e --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a95bd923380a4a4f23b03a25e2d46b3025fdf3d40b0d452a8a36910ce6e93c49 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..b97d4da43587e4bc02b6ac674997a17d51bf714d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9dc795a5376bf0701aeee317ee00ef1548d507bafdf6c53c96de67e846b6f622 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..8702a9bedee0ccd6959cf38fdc8be95e22d4fe0d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5c52dad7a6ea7123e93c1f4ce1c74cf5480fe767785255c74c4816d5f83e11cc +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..e50f5cef090b6e5b59f801087f23efc487f3ac7d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e8190eb5f966a068018838fe363f9a0fe2e267e493e20f9dccd1f03d893cb1d +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..935806e5e64f784e26c8bb7387e4cf96282c65cc --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fbef087f7e2062221abd44b7a9d2611bab1fc79d5876ee576b7a7af70ef2ca7d +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..cc3e11658b9c4dbe68b3fff24ea627e37c511243 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f353627efde8f7de1aaa333d87fcdaced3ac8931b06a2ed7651e3aa47229989e +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..d398d76312b2e88fe21c6966750822b1db0eba1d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/6_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 2.538498878479004, + "learning_rate": 2e-05, + "loss": 0.5055, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 2.919354200363159, + "learning_rate": 2e-05, + "loss": 0.8872, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 5.324807167053223, + "learning_rate": 2e-05, + "loss": 1.0824, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 3.248241424560547, + "learning_rate": 2e-05, + "loss": 0.8132, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 2.0372238159179688, + "learning_rate": 2e-05, + "loss": 0.7887, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 4.135166645050049, + "learning_rate": 2e-05, + "loss": 0.5965, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 7.198287487030029, + "learning_rate": 2e-05, + "loss": 0.8933, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 4.858378887176514, + "learning_rate": 2e-05, + "loss": 1.3957, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 4.046154499053955, + "learning_rate": 2e-05, + "loss": 0.9707, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 5.124589443206787, + "learning_rate": 2e-05, + "loss": 1.6885, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 4.039063930511475, + "learning_rate": 2e-05, + "loss": 2.2162, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 4.183140754699707, + "learning_rate": 2e-05, + "loss": 0.7296, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 3.109577178955078, + "learning_rate": 2e-05, + "loss": 0.9091, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 0.9415032267570496, + "learning_rate": 2e-05, + "loss": 0.6753, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 3.8953185081481934, + "learning_rate": 2e-05, + "loss": 0.6345, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 4.256652355194092, + "learning_rate": 2e-05, + "loss": 0.4792, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 2.1587517261505127, + "learning_rate": 2e-05, + "loss": 0.688, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 4.906612873077393, + "learning_rate": 2e-05, + "loss": 1.7607, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 3.246847152709961, + "learning_rate": 2e-05, + "loss": 0.8009, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 1.7671501636505127, + "learning_rate": 2e-05, + "loss": 0.4656, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 1.8953907489776611, + "learning_rate": 2e-05, + "loss": 0.6538, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 1.432029128074646, + "learning_rate": 2e-05, + "loss": 0.8037, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 1.767372965812683, + "learning_rate": 2e-05, + "loss": 0.6649, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 1.7011256217956543, + "learning_rate": 2e-05, + "loss": 0.3354, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 6.432901859283447, + "learning_rate": 2e-05, + "loss": 2.1649, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 2.6365115642547607, + "learning_rate": 2e-05, + "loss": 0.5956, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 2.40716552734375, + "learning_rate": 2e-05, + "loss": 0.3832, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 4.074473857879639, + "learning_rate": 2e-05, + "loss": 0.8111, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 1.9980592727661133, + "learning_rate": 2e-05, + "loss": 0.4799, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 3.5517821311950684, + "learning_rate": 2e-05, + "loss": 0.3543, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 3.1833691596984863, + "learning_rate": 2e-05, + "loss": 0.72, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 2.168567657470703, + "learning_rate": 2e-05, + "loss": 0.4164, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 3.072348117828369, + "learning_rate": 2e-05, + "loss": 0.8195, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 3.523315191268921, + "learning_rate": 2e-05, + "loss": 0.6332, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 3.4730019569396973, + "learning_rate": 2e-05, + "loss": 0.9315, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 5.45216178894043, + "learning_rate": 2e-05, + "loss": 1.1873, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 8.291824340820312, + "learning_rate": 2e-05, + "loss": 1.7552, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 0.6789868474006653, + "learning_rate": 2e-05, + "loss": 0.172, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 3.0612618923187256, + "learning_rate": 2e-05, + "loss": 1.0881, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 3.230428695678711, + "learning_rate": 2e-05, + "loss": 1.1719, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 4.128942012786865, + "learning_rate": 2e-05, + "loss": 0.6366, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 1.5044084787368774, + "learning_rate": 2e-05, + "loss": 0.4254, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 2.1075439453125, + "learning_rate": 2e-05, + "loss": 1.2789, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 5.075677394866943, + "learning_rate": 2e-05, + "loss": 1.4246, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 4.089530944824219, + "learning_rate": 2e-05, + "loss": 1.3205, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 2.1596262454986572, + "learning_rate": 2e-05, + "loss": 0.3942, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 2.69488787651062, + "learning_rate": 2e-05, + "loss": 0.7403, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 2.6293485164642334, + "learning_rate": 2e-05, + "loss": 1.3489, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 1.3719948530197144, + "learning_rate": 2e-05, + "loss": 0.8554, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 1.8209477663040161, + "learning_rate": 2e-05, + "loss": 1.199, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 9851938422128640.0, + "train_loss": 0.8949293899536133, + "train_runtime": 169.8238, + "train_samples_per_second": 2.355, + "train_steps_per_second": 0.589 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 9851938422128640.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..fa1ab6f19dd5a664c7a1961482e4c15a03c3f654 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a754b8922bf2fee3d6d354fb15ef778f0318152371be3b920d3101b35e7a3e0a +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..83c26bf4b1c6b57edd00f5b7574ad63a5b51eb9d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6df54917b2d096fa38872f704c2de8de40845c8cc02b4f472619b2468b6a76a4 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..78eb89455827470063348624f9d88f84afffec74 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:285185e6778712128d72891d6f8da4afc21deff653cd3ad4f0273b9f3d983ca3 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..30e6e296e23ac01b36eeb70f242d5e0084e40395 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d57f173b76d36aff9a4f4340b64dd3512199fc3990b10b3ed702fa0fa22b413a +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..2ad7ae7c8cae5a649e49cdff48adb6eabe6c65b0 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ed462fca2f93ef303a166a242600e33e5cd6339568f69e740158f6051a240a0 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..44ad72d434351490082b89705d72487724217425 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d5b65fa366a1f44b1d1380111f53b67252d41c46ee4dbbc68a0d3b06215b65b9 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..5097cbc958cc4978a628a63529aa3d59a452e7fb --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:89bc1972137b8157248c063bf240695292b988a59404371e47acaa5d3cdd97d9 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..5698e525caf871aaaa09a2289b8cff3abf384c2e --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:091e8d9852ac78cae63fb3d4feb4166f25cb3b9ab15b68bf8e355ca755d3bba6 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..40c4cc9cf5716ada438cf87a820a1fd7b5ff382b --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/7_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 0.18270663917064667, + "learning_rate": 2e-05, + "loss": 1.3016, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 0.5449135899543762, + "learning_rate": 2e-05, + "loss": 0.4386, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 0.009561922401189804, + "learning_rate": 2e-05, + "loss": 0.4707, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 0.7884088754653931, + "learning_rate": 2e-05, + "loss": 1.5395, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 3.279318332672119, + "learning_rate": 2e-05, + "loss": 0.1594, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 3.92635440826416, + "learning_rate": 2e-05, + "loss": 0.8371, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 0.5402039289474487, + "learning_rate": 2e-05, + "loss": 0.0554, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 0.22995802760124207, + "learning_rate": 2e-05, + "loss": 0.0489, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 0.3144186735153198, + "learning_rate": 2e-05, + "loss": 0.1712, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 7.272878170013428, + "learning_rate": 2e-05, + "loss": 0.8405, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 0.6919670701026917, + "learning_rate": 2e-05, + "loss": 0.3617, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 7.183122634887695, + "learning_rate": 2e-05, + "loss": 1.5823, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 2.401442289352417, + "learning_rate": 2e-05, + "loss": 0.3284, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 3.0500175952911377, + "learning_rate": 2e-05, + "loss": 1.0448, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 1.429979681968689, + "learning_rate": 2e-05, + "loss": 0.3418, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 1.1483200788497925, + "learning_rate": 2e-05, + "loss": 0.554, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 3.196305513381958, + "learning_rate": 2e-05, + "loss": 1.2177, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 0.7277846932411194, + "learning_rate": 2e-05, + "loss": 0.3271, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 6.5030107498168945, + "learning_rate": 2e-05, + "loss": 1.0504, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 1.683670997619629, + "learning_rate": 2e-05, + "loss": 0.746, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 0.3026253283023834, + "learning_rate": 2e-05, + "loss": 0.1643, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 1.1621618270874023, + "learning_rate": 2e-05, + "loss": 0.1844, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 1.30198335647583, + "learning_rate": 2e-05, + "loss": 0.2549, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 1.2834713459014893, + "learning_rate": 2e-05, + "loss": 0.7611, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 1.8870837688446045, + "learning_rate": 2e-05, + "loss": 0.3573, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 1.6936038732528687, + "learning_rate": 2e-05, + "loss": 0.3437, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 0.1798095554113388, + "learning_rate": 2e-05, + "loss": 0.208, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 0.8046261668205261, + "learning_rate": 2e-05, + "loss": 0.1425, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 3.687539577484131, + "learning_rate": 2e-05, + "loss": 0.3281, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 0.3803871273994446, + "learning_rate": 2e-05, + "loss": 0.1108, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 0.381933331489563, + "learning_rate": 2e-05, + "loss": 0.1543, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 1.2285979986190796, + "learning_rate": 2e-05, + "loss": 0.3204, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 1.6024672985076904, + "learning_rate": 2e-05, + "loss": 0.1438, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 7.325067043304443, + "learning_rate": 2e-05, + "loss": 1.7731, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 0.1541583091020584, + "learning_rate": 2e-05, + "loss": 0.2939, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 0.19075414538383484, + "learning_rate": 2e-05, + "loss": 0.0823, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 0.3542366325855255, + "learning_rate": 2e-05, + "loss": 0.0293, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 1.414999008178711, + "learning_rate": 2e-05, + "loss": 0.2212, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 6.296550750732422, + "learning_rate": 2e-05, + "loss": 0.8679, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 5.9345293045043945, + "learning_rate": 2e-05, + "loss": 0.7815, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 4.660360813140869, + "learning_rate": 2e-05, + "loss": 1.0491, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 0.11966075003147125, + "learning_rate": 2e-05, + "loss": 0.0489, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 0.5875673890113831, + "learning_rate": 2e-05, + "loss": 0.2149, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 1.3151464462280273, + "learning_rate": 2e-05, + "loss": 0.9811, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 0.2438245564699173, + "learning_rate": 2e-05, + "loss": 0.3944, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 0.5219161510467529, + "learning_rate": 2e-05, + "loss": 0.1049, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 8.12782096862793, + "learning_rate": 2e-05, + "loss": 1.1725, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 1.9610928297042847, + "learning_rate": 2e-05, + "loss": 0.5585, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 0.580781102180481, + "learning_rate": 2e-05, + "loss": 0.1237, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 0.3809000551700592, + "learning_rate": 2e-05, + "loss": 0.438, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 5166712907366400.0, + "train_loss": 0.5205245929956436, + "train_runtime": 165.82, + "train_samples_per_second": 2.412, + "train_steps_per_second": 0.603 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 5166712907366400.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..4086aeede8986b84ed1437084cff0e361f1194b7 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dcce618c70dac23f2655dedd65ace2673a1927b05dddd65f52aa548573b3e00b +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..93ceadd3720e6b31b49abe21a065e3171dccece0 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3accc7b61e8f3c996879acde8bb8fdc0f9b1ae6a3e2890ee15642a006df5534f +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..850aebdbd397d747a90b7fb7051d69a38bdb7e52 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:347fe57a37c53176f65a46deb64e0215e9e0cb42e211674adcf75a903a57935a +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..22d667540a9186a9b26f31b318e3dd31fd209b6c --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7f3d3b1def1f788b8dd534ad6e3e6c069704e87287830cf49ec0cefbf91f8e55 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..b0107b134bc5105689478d896d914a7745f6b3f4 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0771f2b3a9139b7f5e375da667b3dc699f78a498a4304ed9436ee2327208d1ee +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..d30462a7c16c9f06797a5c5c369f56c3de609734 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:65f92847792ba6372156a2424a90a07630ca23c2d1fc2c33118fd4bfefa8f2ec +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..4ee9c3bbb6dc361d2a03a1feb8bcc8b9d7cfe1d9 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0fb096697cb8c0f71f43ccdcc019751ced8f6a81ec95dec6571b51a26212383e +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..48480d1dcc6d3d16ba0da4d05b760bbd1d8800c2 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d49dd1ef3bb0e078781fab4f63036f94a838b44da0fc4affc543c8b18a805459 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6809535a0c4527dae59c89b8ce43ccd89887a561 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_NOCONT_bs4_saveoptim_lr2e-5_5e-5_sc315_4tasks_5rounds_fixitr100_T0125_decay099_SEED2/8_trainer_state.json @@ -0,0 +1,392 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 100, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.02, + "grad_norm": 0.5958298444747925, + "learning_rate": 2e-05, + "loss": 0.427, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 0.5032282471656799, + "learning_rate": 2e-05, + "loss": 0.0559, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 5.3274030685424805, + "learning_rate": 2e-05, + "loss": 1.311, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 0.13229762017726898, + "learning_rate": 2e-05, + "loss": 0.2214, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 5.794295310974121, + "learning_rate": 2e-05, + "loss": 0.375, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 0.26891660690307617, + "learning_rate": 2e-05, + "loss": 0.0349, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 0.06987325847148895, + "learning_rate": 2e-05, + "loss": 0.0205, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 0.17918087542057037, + "learning_rate": 2e-05, + "loss": 0.036, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 0.01345374621450901, + "learning_rate": 2e-05, + "loss": 0.0091, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 3.414367914199829, + "learning_rate": 2e-05, + "loss": 0.313, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 10.32875919342041, + "learning_rate": 2e-05, + "loss": 0.9422, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 1.7462822198867798, + "learning_rate": 2e-05, + "loss": 0.8388, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 2.7958827018737793, + "learning_rate": 2e-05, + "loss": 0.392, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 0.11593950539827347, + "learning_rate": 2e-05, + "loss": 0.0415, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 0.07363761961460114, + "learning_rate": 2e-05, + "loss": 0.2254, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 3.063910722732544, + "learning_rate": 2e-05, + "loss": 0.4197, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 0.10028170794248581, + "learning_rate": 2e-05, + "loss": 0.0928, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 0.7315466403961182, + "learning_rate": 2e-05, + "loss": 0.1009, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 0.7060317993164062, + "learning_rate": 2e-05, + "loss": 0.0774, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 0.337223082780838, + "learning_rate": 2e-05, + "loss": 0.7106, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 0.007600266020745039, + "learning_rate": 2e-05, + "loss": 0.3585, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 3.6515843868255615, + "learning_rate": 2e-05, + "loss": 1.0281, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 0.063631571829319, + "learning_rate": 2e-05, + "loss": 0.3464, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 0.0748390257358551, + "learning_rate": 2e-05, + "loss": 0.055, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 0.05504520982503891, + "learning_rate": 2e-05, + "loss": 0.0381, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 3.143303632736206, + "learning_rate": 2e-05, + "loss": 1.2647, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 0.8141433596611023, + "learning_rate": 2e-05, + "loss": 1.3926, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 3.7388546466827393, + "learning_rate": 2e-05, + "loss": 1.0186, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 0.2085130363702774, + "learning_rate": 2e-05, + "loss": 0.0152, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 5.296192169189453, + "learning_rate": 2e-05, + "loss": 0.2494, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 2.7831602096557617, + "learning_rate": 2e-05, + "loss": 0.6257, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 0.6261029839515686, + "learning_rate": 2e-05, + "loss": 0.145, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 2.655792713165283, + "learning_rate": 2e-05, + "loss": 0.3986, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 3.6786677837371826, + "learning_rate": 2e-05, + "loss": 0.6878, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 5.979826927185059, + "learning_rate": 2e-05, + "loss": 1.0474, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 1.0922977924346924, + "learning_rate": 2e-05, + "loss": 0.5536, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 0.024878626689314842, + "learning_rate": 2e-05, + "loss": 0.5604, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 0.6924214959144592, + "learning_rate": 2e-05, + "loss": 0.1295, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 0.07321670651435852, + "learning_rate": 2e-05, + "loss": 0.0156, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 0.7287451028823853, + "learning_rate": 2e-05, + "loss": 0.3552, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 2.961714506149292, + "learning_rate": 2e-05, + "loss": 0.3924, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 1.496504306793213, + "learning_rate": 2e-05, + "loss": 0.25, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 0.296202689409256, + "learning_rate": 2e-05, + "loss": 0.0996, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 0.5026251077651978, + "learning_rate": 2e-05, + "loss": 0.0705, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 2.435487747192383, + "learning_rate": 2e-05, + "loss": 1.3084, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 3.0182111263275146, + "learning_rate": 2e-05, + "loss": 0.8172, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 0.19730326533317566, + "learning_rate": 2e-05, + "loss": 0.4208, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 0.07050402462482452, + "learning_rate": 2e-05, + "loss": 0.0512, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 0.20642773807048798, + "learning_rate": 2e-05, + "loss": 0.6169, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 4.2984418869018555, + "learning_rate": 2e-05, + "loss": 1.2015, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 5020431387983872.0, + "train_loss": 0.44317606925964353, + "train_runtime": 164.7406, + "train_samples_per_second": 2.428, + "train_steps_per_second": 0.607 + } + ], + "logging_steps": 2, + "max_steps": 100, + "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": 5020431387983872.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}