diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..8d752573d814818b30a12e7ed29fbcc7651c1211 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7281b71feb874201c3510fd2532c735c41834e0ea57f60b3a19d83a172c8062d +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..aeced2dc950ba5fc44f933f73b6e965182530173 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:43f3b3c81a25994ed2450656aad07f50b0d6a686213e90ee5936f5227e75b924 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..6061031b1add65c34a85e68c481d88bc65437bb7 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:02634f446483c9401811d202c4fb26c234e200c1b5bd389dca72585c2d61d034 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..d1b57f9c1de4a6c04b2d01af572ae5cc7b1ceabe --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b593199404c65a7fb376d970ec893e60e94daef7a5ae8eed3ca517e9275fe0c +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..7a727056a9a0af14ca755f0d14bc410ffb38ca0a --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f312bff96288926254f6abc38cb4e1313e8d188484ebbbeadf5f59357ce0f3e3 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..ac33facec921f6f105ab8aecbc7628d51d832afd --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03270abac1135f1e20d23bcbaf393ffd4157017414b071ece5ec558a5319692a +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..b026896bb6d45d421895df75c3d9d6009e312faa --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dc2eaa3387065398c5c4e6cb687acf55c8c80d95eddaf5452a6456e4dd5b1d6f +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..e48ed2400ebc252e73cd777f012c019ff122882f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7bea9fbbcea6ab83084b33a90fc70ebbcc775b9472862c5f46300a085976a036 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..247810d7de2402b36c55a45646d2f1ab7c958c9f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/0_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": 4.327882289886475, + "learning_rate": 2e-05, + "loss": 0.3049, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 4.293352127075195, + "learning_rate": 2e-05, + "loss": 0.4343, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 8.278122901916504, + "learning_rate": 2e-05, + "loss": 0.464, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 5.572937488555908, + "learning_rate": 2e-05, + "loss": 1.1184, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 1.2564692497253418, + "learning_rate": 2e-05, + "loss": 0.1493, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 2.756223201751709, + "learning_rate": 2e-05, + "loss": 0.1429, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 6.3286638259887695, + "learning_rate": 2e-05, + "loss": 1.0986, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 3.1671125888824463, + "learning_rate": 2e-05, + "loss": 0.421, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 12.160600662231445, + "learning_rate": 2e-05, + "loss": 0.6418, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 2.2388546466827393, + "learning_rate": 2e-05, + "loss": 0.3821, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 4.057492733001709, + "learning_rate": 2e-05, + "loss": 0.2804, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 2.4171645641326904, + "learning_rate": 2e-05, + "loss": 0.2768, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 4.334846019744873, + "learning_rate": 2e-05, + "loss": 0.5183, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 5.31112003326416, + "learning_rate": 2e-05, + "loss": 0.3826, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 8.947898864746094, + "learning_rate": 2e-05, + "loss": 0.7833, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 4.790311336517334, + "learning_rate": 2e-05, + "loss": 0.7227, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 1.9647458791732788, + "learning_rate": 2e-05, + "loss": 0.2151, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 7.152100086212158, + "learning_rate": 2e-05, + "loss": 0.6738, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 9.002793312072754, + "learning_rate": 2e-05, + "loss": 1.2422, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 4.538879871368408, + "learning_rate": 2e-05, + "loss": 0.7256, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 7.2817158699035645, + "learning_rate": 2e-05, + "loss": 0.4922, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 8.116897583007812, + "learning_rate": 2e-05, + "loss": 0.2341, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 5.7875776290893555, + "learning_rate": 2e-05, + "loss": 0.5828, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 2.9006869792938232, + "learning_rate": 2e-05, + "loss": 0.2646, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 5.325571060180664, + "learning_rate": 2e-05, + "loss": 0.5073, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 1.7935487031936646, + "learning_rate": 2e-05, + "loss": 0.3458, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 6.689419269561768, + "learning_rate": 2e-05, + "loss": 0.9642, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 9.98328685760498, + "learning_rate": 2e-05, + "loss": 0.7798, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 10.168957710266113, + "learning_rate": 2e-05, + "loss": 0.3466, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 5.621067523956299, + "learning_rate": 2e-05, + "loss": 0.4138, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 16.461015701293945, + "learning_rate": 2e-05, + "loss": 0.963, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 3.5031635761260986, + "learning_rate": 2e-05, + "loss": 0.6482, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 1.3376978635787964, + "learning_rate": 2e-05, + "loss": 0.1724, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 1.3010202646255493, + "learning_rate": 2e-05, + "loss": 0.2854, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 8.061728477478027, + "learning_rate": 2e-05, + "loss": 0.4303, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 5.248547077178955, + "learning_rate": 2e-05, + "loss": 0.5894, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 4.7633819580078125, + "learning_rate": 2e-05, + "loss": 0.6274, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 1.664157509803772, + "learning_rate": 2e-05, + "loss": 0.3461, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 2.7871382236480713, + "learning_rate": 2e-05, + "loss": 0.5779, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 5.3619914054870605, + "learning_rate": 2e-05, + "loss": 0.7158, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 4.908105373382568, + "learning_rate": 2e-05, + "loss": 0.5618, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 4.174102306365967, + "learning_rate": 2e-05, + "loss": 0.645, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 5.976959705352783, + "learning_rate": 2e-05, + "loss": 0.5615, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 5.8181586265563965, + "learning_rate": 2e-05, + "loss": 0.6136, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 2.994356155395508, + "learning_rate": 2e-05, + "loss": 0.196, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 4.384360313415527, + "learning_rate": 2e-05, + "loss": 0.2452, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 3.3636257648468018, + "learning_rate": 2e-05, + "loss": 0.3809, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 3.677785634994507, + "learning_rate": 2e-05, + "loss": 0.3801, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 6.116004467010498, + "learning_rate": 2e-05, + "loss": 0.5769, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 5.415121555328369, + "learning_rate": 2e-05, + "loss": 0.385, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 6589214774788096.0, + "train_loss": 0.5162237358093261, + "train_runtime": 121.9606, + "train_samples_per_second": 3.28, + "train_steps_per_second": 0.82 + } + ], + "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": 6589214774788096.0, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..e729027a953d4e1cc07755a06fe2b522bcbb7cfd --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4b7524ff1cdb9c623c9f04b1b2f456ce956f8333f742e367f0627615c523df0 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..99c463cb303777143fe46847f5a0596cd1dca1cc --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b14de6a6ff83a367693fffd74a202f81b6610ae9866eca998fee857876b57cc6 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..e5595599d5712053d3fe1958acbf5d5eec683f7e --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fc604f9b5b83debe041f5baec0917bf5f25da3d2400cc3642da153b1a214c1a0 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..6e7872fdfed67f09779c908e7c67f95696e9ba20 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fab435c732539a8ae5d19b6f27b57b14502392076a996005e26331e5c8488f24 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..3218a6805784774f7550bdb8428058b29340918b --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45faca8a1b9116373952a8ef00712fa1e379398914a28043649b825dc9345f5b +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..d6cfa539723040360088ea3d1b629f5fcd7cb23b --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e371888600f40c5bfd4398e7526aab31e42df408736003ec9c070aa0a7ed0d7e +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..15d7e3145929280c5c99790d56d0582e27b1eb9f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac5f9b83f02f09f963455cb0e4dbe43fe7930e2e9b1fd8d2976ffad70b1f71ab +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..19bbc8d364f3c6e6b81b9d2db3861ef104a976f1 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7728b781894e5139abed7ddf3350ce0c8d77c0f74d4a37fb6fa9db4f9b6b07b9 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..0c61411979580d0cb3c91815aadbdd3ae7e8a642 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/1_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": 8.241416931152344, + "learning_rate": 2e-05, + "loss": 0.8475, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 4.8631792068481445, + "learning_rate": 2e-05, + "loss": 0.6782, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 7.5998454093933105, + "learning_rate": 2e-05, + "loss": 0.5532, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 2.5101563930511475, + "learning_rate": 2e-05, + "loss": 0.5537, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 4.472541332244873, + "learning_rate": 2e-05, + "loss": 0.6426, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 4.083187103271484, + "learning_rate": 2e-05, + "loss": 0.5195, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 5.575020790100098, + "learning_rate": 2e-05, + "loss": 0.4822, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 4.933296203613281, + "learning_rate": 2e-05, + "loss": 0.5742, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 2.017449378967285, + "learning_rate": 2e-05, + "loss": 0.4663, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 6.998116493225098, + "learning_rate": 2e-05, + "loss": 0.582, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 2.1770007610321045, + "learning_rate": 2e-05, + "loss": 0.5249, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 5.969364166259766, + "learning_rate": 2e-05, + "loss": 0.623, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 1.237125277519226, + "learning_rate": 2e-05, + "loss": 0.3623, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 6.773047924041748, + "learning_rate": 2e-05, + "loss": 0.4094, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 2.925546169281006, + "learning_rate": 2e-05, + "loss": 0.6287, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 3.002979278564453, + "learning_rate": 2e-05, + "loss": 0.7397, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 1.8541330099105835, + "learning_rate": 2e-05, + "loss": 0.5017, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 7.924871921539307, + "learning_rate": 2e-05, + "loss": 0.6328, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 2.390252113342285, + "learning_rate": 2e-05, + "loss": 0.3247, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 3.807579278945923, + "learning_rate": 2e-05, + "loss": 0.4795, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 3.4766390323638916, + "learning_rate": 2e-05, + "loss": 0.4468, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 1.4840394258499146, + "learning_rate": 2e-05, + "loss": 0.5312, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 3.136503219604492, + "learning_rate": 2e-05, + "loss": 0.4313, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 2.3651793003082275, + "learning_rate": 2e-05, + "loss": 0.3152, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 3.1154613494873047, + "learning_rate": 2e-05, + "loss": 0.6135, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 5.85906982421875, + "learning_rate": 2e-05, + "loss": 0.9023, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 5.495807647705078, + "learning_rate": 2e-05, + "loss": 0.5264, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 2.5824015140533447, + "learning_rate": 2e-05, + "loss": 0.2805, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 6.252473831176758, + "learning_rate": 2e-05, + "loss": 0.4327, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 5.942077159881592, + "learning_rate": 2e-05, + "loss": 0.4004, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 5.180813789367676, + "learning_rate": 2e-05, + "loss": 0.864, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 3.008401870727539, + "learning_rate": 2e-05, + "loss": 0.292, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 1.5255389213562012, + "learning_rate": 2e-05, + "loss": 0.7554, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 11.456948280334473, + "learning_rate": 2e-05, + "loss": 0.7676, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 1.6395078897476196, + "learning_rate": 2e-05, + "loss": 0.564, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 2.9087085723876953, + "learning_rate": 2e-05, + "loss": 0.5242, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 2.2395269870758057, + "learning_rate": 2e-05, + "loss": 0.4575, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 7.380107879638672, + "learning_rate": 2e-05, + "loss": 0.5474, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 3.7302591800689697, + "learning_rate": 2e-05, + "loss": 0.365, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 7.54459810256958, + "learning_rate": 2e-05, + "loss": 0.8384, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 1.8007885217666626, + "learning_rate": 2e-05, + "loss": 0.354, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 1.5501184463500977, + "learning_rate": 2e-05, + "loss": 0.5158, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 7.850472927093506, + "learning_rate": 2e-05, + "loss": 0.5859, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 2.9437716007232666, + "learning_rate": 2e-05, + "loss": 0.5186, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 6.130828380584717, + "learning_rate": 2e-05, + "loss": 0.5325, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 3.8235013484954834, + "learning_rate": 2e-05, + "loss": 0.4294, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 10.231576919555664, + "learning_rate": 2e-05, + "loss": 0.5122, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 3.4070258140563965, + "learning_rate": 2e-05, + "loss": 0.4346, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 2.635401725769043, + "learning_rate": 2e-05, + "loss": 0.4543, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 4.57978630065918, + "learning_rate": 2e-05, + "loss": 0.8027, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 1.6576168127889408e+16, + "train_loss": 0.5424420166015625, + "train_runtime": 289.5764, + "train_samples_per_second": 1.381, + "train_steps_per_second": 0.345 + } + ], + "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": 1.6576168127889408e+16, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..e7b0cb4fe2d259a21172c0582a4c3a1d4b09088b --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3b16e88793c124816f1f558de280952798c45b5b80104feb95b1ea3c823f4317 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..95dcc6403ae3c19945a840c20079d0ac233a4fa3 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aeb99f249d83641dd27f96f403f9a3896cc2c106017f1dac1743ed4462b07ee5 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..3ed6b2052f6ae3e0fe2a36381f855c92c1a8bc17 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1544b776c37e6184161b7dce77cfc588b766260a10846dd63415c96f9147be4 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..c9f73886af17b7a4c4bbedfd8bfa1186dc0ba554 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34be3c453cd7a2af03cf351083bd8e44c0b7da498f43599dc9da884887ab8ec6 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..e4df9a5ef666e103512f70f7f27cf714bc7060ab --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23b15c520084743ef62975eae5db8d4352e6c8829c36aa7360dc593c57110e2c +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..7eb88470dc699b05d663b872a048f468f2a1f85e --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:adc4c8842a382db73fdc998bf2ec50d04d6f160cab990be0e64eaa17ee6d697d +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..3c9a9907ca6aef791c8c4c8ae62bb61b5d0c4d40 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c91d31a53873f602e9654cdc5d71ec452143909946552d992ae47d2e138a505 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..80498c440b0722188292ff3f3087869a82d26b85 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4b6565669e5be448c7713d437c76b999e97c1f563a913e8d7b24fb23f5a026b +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..eebd6a94ea2931817bfa88312d3cf1287ba20486 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/2_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": 7.872583389282227, + "learning_rate": 2e-05, + "loss": 0.4408, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 3.1676270961761475, + "learning_rate": 2e-05, + "loss": 0.2183, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 3.10235595703125, + "learning_rate": 2e-05, + "loss": 0.3065, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 1.986329436302185, + "learning_rate": 2e-05, + "loss": 0.1555, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 6.654890060424805, + "learning_rate": 2e-05, + "loss": 0.4976, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 8.408140182495117, + "learning_rate": 2e-05, + "loss": 0.6118, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 9.52741813659668, + "learning_rate": 2e-05, + "loss": 1.0054, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 3.1671223640441895, + "learning_rate": 2e-05, + "loss": 0.3345, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 6.954838275909424, + "learning_rate": 2e-05, + "loss": 0.5977, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 3.3944637775421143, + "learning_rate": 2e-05, + "loss": 0.3164, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 3.7810606956481934, + "learning_rate": 2e-05, + "loss": 0.5396, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 2.8664538860321045, + "learning_rate": 2e-05, + "loss": 0.5024, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 5.678687572479248, + "learning_rate": 2e-05, + "loss": 0.4766, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 7.057774066925049, + "learning_rate": 2e-05, + "loss": 0.4387, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 2.025454044342041, + "learning_rate": 2e-05, + "loss": 0.3817, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 2.349515676498413, + "learning_rate": 2e-05, + "loss": 0.3013, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 3.751689910888672, + "learning_rate": 2e-05, + "loss": 0.3142, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 9.54448127746582, + "learning_rate": 2e-05, + "loss": 0.6094, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 2.6551620960235596, + "learning_rate": 2e-05, + "loss": 0.3418, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 7.663268089294434, + "learning_rate": 2e-05, + "loss": 0.3774, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 5.838813781738281, + "learning_rate": 2e-05, + "loss": 0.5634, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 3.6920807361602783, + "learning_rate": 2e-05, + "loss": 0.1825, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 2.167635202407837, + "learning_rate": 2e-05, + "loss": 0.2479, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 1.2298402786254883, + "learning_rate": 2e-05, + "loss": 0.274, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 8.415931701660156, + "learning_rate": 2e-05, + "loss": 0.6021, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 5.416069984436035, + "learning_rate": 2e-05, + "loss": 0.4319, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 2.6783201694488525, + "learning_rate": 2e-05, + "loss": 0.4147, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 1.76664137840271, + "learning_rate": 2e-05, + "loss": 0.2076, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 3.5323493480682373, + "learning_rate": 2e-05, + "loss": 0.3228, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 4.204054832458496, + "learning_rate": 2e-05, + "loss": 0.3702, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 16.42522621154785, + "learning_rate": 2e-05, + "loss": 0.8672, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 1.9896955490112305, + "learning_rate": 2e-05, + "loss": 0.4229, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 6.183014869689941, + "learning_rate": 2e-05, + "loss": 0.552, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 5.789742469787598, + "learning_rate": 2e-05, + "loss": 0.3384, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 11.22754955291748, + "learning_rate": 2e-05, + "loss": 0.3374, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 1.9776259660720825, + "learning_rate": 2e-05, + "loss": 0.3358, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 4.03769588470459, + "learning_rate": 2e-05, + "loss": 0.4689, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 6.714748382568359, + "learning_rate": 2e-05, + "loss": 0.7046, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 1.7425217628479004, + "learning_rate": 2e-05, + "loss": 0.0582, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 2.1245808601379395, + "learning_rate": 2e-05, + "loss": 0.3812, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 8.465872764587402, + "learning_rate": 2e-05, + "loss": 0.4424, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 2.888294219970703, + "learning_rate": 2e-05, + "loss": 0.1748, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 3.158863067626953, + "learning_rate": 2e-05, + "loss": 0.2654, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 7.765163898468018, + "learning_rate": 2e-05, + "loss": 0.448, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 2.6586227416992188, + "learning_rate": 2e-05, + "loss": 0.1208, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 4.466168403625488, + "learning_rate": 2e-05, + "loss": 0.4543, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 9.527925491333008, + "learning_rate": 2e-05, + "loss": 1.0899, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 7.631684303283691, + "learning_rate": 2e-05, + "loss": 0.9442, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 9.0326566696167, + "learning_rate": 2e-05, + "loss": 0.7216, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 9.811245918273926, + "learning_rate": 2e-05, + "loss": 0.5501, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 1.9115952041885696e+16, + "train_loss": 0.44124862670898435, + "train_runtime": 324.4955, + "train_samples_per_second": 1.233, + "train_steps_per_second": 0.308 + } + ], + "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": 1.9115952041885696e+16, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..57ad934c88941249dab0f6cd06bf70b4714fda21 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8b95720db68933091baf1fdecaa1719d3ae6d74f9d586153dfbdd9a8d412efd5 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..b87b046adc9f02f9503403df39f51eef8c1a2343 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9462aef5d3987e462b4eb54f5767c7a989d765da13e27c013f24ea225359ff85 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..d835ece66ab3ca3f7136357ccfbedadd1f7abbb0 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:da91a450b0e42f8409cf5f1538d602d5016eb3af4a99a673858d20fac1506bf4 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..8177299fd94af472fba76658e7bfd5c51fd29a25 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e2137138ebfad8ba608a66235c099e93ed835a9caccde901a0cce2619288f7d1 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..30fb0e3cfc7bc0a8ac0a3c43c4cba248dee94f2c --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a1463e43df9db510a4857be4caea7caef4417a1d68d2af7784cc2114322d425 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..69a611e38e83979b43bf571c345c9918e929576c --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6986573a8d5a03d8aafbd13449ec4c1b317445289bb1d6fe3f86f5a60b2293a1 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..a7bf53751ae346510e42d2a83d266163aacc0c18 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b459889ad8c0dddbdb5b5892a3dc82a8c7aff84f714b2aa33a6d025449e86354 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..7918dd9eaa59975b99c755c806b264b95b048a4d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4c6ae5cabef78b40781076f85a494852e0479f8117956282442c035a9696f6c +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..4391ac9db9f0fb5e52289688c7402775111b3ae5 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/3_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": 4.6497626304626465, + "learning_rate": 2e-05, + "loss": 0.5825, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 3.6290929317474365, + "learning_rate": 2e-05, + "loss": 0.915, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 2.3345847129821777, + "learning_rate": 2e-05, + "loss": 0.4902, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 5.3167009353637695, + "learning_rate": 2e-05, + "loss": 0.4688, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 3.8224802017211914, + "learning_rate": 2e-05, + "loss": 0.5308, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 2.823826551437378, + "learning_rate": 2e-05, + "loss": 0.2563, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 2.9755983352661133, + "learning_rate": 2e-05, + "loss": 0.3347, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 2.022836923599243, + "learning_rate": 2e-05, + "loss": 0.426, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 1.7052284479141235, + "learning_rate": 2e-05, + "loss": 0.5396, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 1.3424859046936035, + "learning_rate": 2e-05, + "loss": 0.4526, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 5.240848541259766, + "learning_rate": 2e-05, + "loss": 0.4441, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 4.661110877990723, + "learning_rate": 2e-05, + "loss": 0.6226, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 1.6846646070480347, + "learning_rate": 2e-05, + "loss": 0.3237, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 5.903859615325928, + "learning_rate": 2e-05, + "loss": 0.7686, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 2.1405582427978516, + "learning_rate": 2e-05, + "loss": 0.6973, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 3.7082715034484863, + "learning_rate": 2e-05, + "loss": 0.5305, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 4.9589338302612305, + "learning_rate": 2e-05, + "loss": 0.4478, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 7.48822546005249, + "learning_rate": 2e-05, + "loss": 0.4746, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 8.993917465209961, + "learning_rate": 2e-05, + "loss": 0.501, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 2.639606237411499, + "learning_rate": 2e-05, + "loss": 0.3493, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 7.567841529846191, + "learning_rate": 2e-05, + "loss": 0.5112, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 3.463698625564575, + "learning_rate": 2e-05, + "loss": 0.3148, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 3.490788698196411, + "learning_rate": 2e-05, + "loss": 1.1011, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 3.9706830978393555, + "learning_rate": 2e-05, + "loss": 0.303, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 4.992152214050293, + "learning_rate": 2e-05, + "loss": 0.6089, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 1.5260461568832397, + "learning_rate": 2e-05, + "loss": 0.1576, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 2.76921010017395, + "learning_rate": 2e-05, + "loss": 0.4696, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 3.4018092155456543, + "learning_rate": 2e-05, + "loss": 0.5396, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 3.256915330886841, + "learning_rate": 2e-05, + "loss": 0.3384, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 2.284102439880371, + "learning_rate": 2e-05, + "loss": 0.2712, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 5.065925121307373, + "learning_rate": 2e-05, + "loss": 0.3423, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 2.2383081912994385, + "learning_rate": 2e-05, + "loss": 0.5986, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 3.604736566543579, + "learning_rate": 2e-05, + "loss": 0.594, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 2.073974609375, + "learning_rate": 2e-05, + "loss": 0.1852, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 4.072497367858887, + "learning_rate": 2e-05, + "loss": 0.3573, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 2.104954719543457, + "learning_rate": 2e-05, + "loss": 0.4517, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 20.248361587524414, + "learning_rate": 2e-05, + "loss": 0.5629, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 5.614309787750244, + "learning_rate": 2e-05, + "loss": 0.5081, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 3.6506497859954834, + "learning_rate": 2e-05, + "loss": 0.7004, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 0.7987788319587708, + "learning_rate": 2e-05, + "loss": 0.2193, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 2.1284451484680176, + "learning_rate": 2e-05, + "loss": 0.3389, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 7.366853713989258, + "learning_rate": 2e-05, + "loss": 0.748, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 4.810544490814209, + "learning_rate": 2e-05, + "loss": 0.7964, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 6.225423336029053, + "learning_rate": 2e-05, + "loss": 0.5322, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 4.1773576736450195, + "learning_rate": 2e-05, + "loss": 0.5176, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 6.231110572814941, + "learning_rate": 2e-05, + "loss": 0.4377, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 2.0640790462493896, + "learning_rate": 2e-05, + "loss": 0.3445, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 3.780561685562134, + "learning_rate": 2e-05, + "loss": 0.5642, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 4.6030449867248535, + "learning_rate": 2e-05, + "loss": 0.363, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 1.9157155752182007, + "learning_rate": 2e-05, + "loss": 0.2215, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 2.386061380550656e+16, + "train_loss": 0.4830999755859375, + "train_runtime": 409.3895, + "train_samples_per_second": 0.977, + "train_steps_per_second": 0.244 + } + ], + "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": 2.386061380550656e+16, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..5ba585e5d32fbedf28e7c281bcf2bc7ce64414e8 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:893681346ba343e9d0fa2f36e51accfc6d218f0664ecca940e57f3d2dc374ad9 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..0fbea899510039d7580becae9e8d06eab35a61d6 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b46411ee8d7334f558926aaf247e2004138dcd997e035c87686449bcfc1cc469 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..f4f704e4415db702a252f60e8b13a4ccb1276e09 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bdb196c7e4df84cc6052571a2d10021bc4e6208c3ed2b79af86450bb25dee17c +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..6574369a3648bf70e3010f34fbcafaf04468db0e --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b45484c94bb3ab612febd88336db3b78e201867c1e805bbd435f37ef49104931 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..71eecd5793745762a6dc99ea62adf732754a607b --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:118254270d48a1bd72d800e0b757f7eee2d6879af1e323d473ca3b25a5b156fe +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..2c1873524dca14a7ee37b6433881898754e09f34 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:479108ba00e734b040fe1210bcbd5d58bcbf9b9d213843eada9ace7329f09763 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..8362932e98a37d4f6e50405b07855a6c5c85ec9d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5cda0740f3777ae9d6bc03793b0f525db6fc40288d9de12a48092382737c4350 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..674e6cbfe1e8169bbba5d126d6a444825f0755a5 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3def9c46010f15f880769a9efe2296b8114ce3d0a5ad9754de24319c2fe578c +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..85fa558e1c5daabab0648c58249df63611b597ea --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/4_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": 3.530467987060547, + "learning_rate": 2e-05, + "loss": 0.2157, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 1.8090039491653442, + "learning_rate": 2e-05, + "loss": 0.1584, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 2.7021498680114746, + "learning_rate": 2e-05, + "loss": 0.2725, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 1.0994998216629028, + "learning_rate": 2e-05, + "loss": 0.1925, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 4.899983882904053, + "learning_rate": 2e-05, + "loss": 0.4641, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 1.8486827611923218, + "learning_rate": 2e-05, + "loss": 0.6014, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 0.669780433177948, + "learning_rate": 2e-05, + "loss": 0.1162, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 4.612163066864014, + "learning_rate": 2e-05, + "loss": 0.5449, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 1.251813292503357, + "learning_rate": 2e-05, + "loss": 0.3157, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 4.452538013458252, + "learning_rate": 2e-05, + "loss": 0.1715, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 4.419332981109619, + "learning_rate": 2e-05, + "loss": 0.2881, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 2.8015778064727783, + "learning_rate": 2e-05, + "loss": 0.8685, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 5.626010894775391, + "learning_rate": 2e-05, + "loss": 1.0619, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 7.333406925201416, + "learning_rate": 2e-05, + "loss": 1.0535, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 6.450397968292236, + "learning_rate": 2e-05, + "loss": 0.9023, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 3.454080820083618, + "learning_rate": 2e-05, + "loss": 0.4188, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 2.1945369243621826, + "learning_rate": 2e-05, + "loss": 0.4095, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 2.2680203914642334, + "learning_rate": 2e-05, + "loss": 0.9443, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 5.364918231964111, + "learning_rate": 2e-05, + "loss": 0.2831, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 5.871020793914795, + "learning_rate": 2e-05, + "loss": 0.4717, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 0.05804154649376869, + "learning_rate": 2e-05, + "loss": 0.2073, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 3.274716854095459, + "learning_rate": 2e-05, + "loss": 0.595, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 2.444748878479004, + "learning_rate": 2e-05, + "loss": 0.395, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 2.9482669830322266, + "learning_rate": 2e-05, + "loss": 0.3739, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 4.834102630615234, + "learning_rate": 2e-05, + "loss": 0.464, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 4.0864577293396, + "learning_rate": 2e-05, + "loss": 1.0002, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 0.2608688473701477, + "learning_rate": 2e-05, + "loss": 0.3152, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 1.6589609384536743, + "learning_rate": 2e-05, + "loss": 0.1979, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 2.3401970863342285, + "learning_rate": 2e-05, + "loss": 0.3342, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 2.3094046115875244, + "learning_rate": 2e-05, + "loss": 0.2121, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 2.6082193851470947, + "learning_rate": 2e-05, + "loss": 0.345, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 7.014457702636719, + "learning_rate": 2e-05, + "loss": 0.8164, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 4.024754047393799, + "learning_rate": 2e-05, + "loss": 0.7554, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 0.4044950008392334, + "learning_rate": 2e-05, + "loss": 0.0986, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 4.961355209350586, + "learning_rate": 2e-05, + "loss": 0.4958, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 7.53846549987793, + "learning_rate": 2e-05, + "loss": 0.5591, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 0.829863429069519, + "learning_rate": 2e-05, + "loss": 0.4915, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 3.274028778076172, + "learning_rate": 2e-05, + "loss": 0.4124, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 5.568228721618652, + "learning_rate": 2e-05, + "loss": 0.3262, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 6.694833755493164, + "learning_rate": 2e-05, + "loss": 0.6133, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 5.746543884277344, + "learning_rate": 2e-05, + "loss": 0.8642, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 11.073870658874512, + "learning_rate": 2e-05, + "loss": 0.833, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 4.428623676300049, + "learning_rate": 2e-05, + "loss": 0.333, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 7.5453033447265625, + "learning_rate": 2e-05, + "loss": 0.7605, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 7.614424228668213, + "learning_rate": 2e-05, + "loss": 0.2277, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 8.295862197875977, + "learning_rate": 2e-05, + "loss": 0.9114, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 2.497485876083374, + "learning_rate": 2e-05, + "loss": 0.2709, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 1.357002854347229, + "learning_rate": 2e-05, + "loss": 0.6364, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 6.519164562225342, + "learning_rate": 2e-05, + "loss": 0.3441, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 2.9073774814605713, + "learning_rate": 2e-05, + "loss": 0.4589, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 2.410703500397773e+16, + "train_loss": 0.48806270599365237, + "train_runtime": 603.2643, + "train_samples_per_second": 0.663, + "train_steps_per_second": 0.166 + } + ], + "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": 2.410703500397773e+16, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..f5c68e27f0c622f3ed2ad7d44f976a05dac7eae3 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e98397f8ee641afb7450a1f3f2b7b46bc3b40c3af558ed2250763f6daf8926a1 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..01a6c6cbbd94be20c559cbd84f87c2998b1738c5 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b1182ee1ea28f8787bca578e777f053b2ee0a10bd3430a546fb68a7cdcf699d7 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..7e36ce44dccd274eecba2f4c832385fb7fbcc376 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:63ba0332da3f84fc4e4f6fd1d7367267ccf7e78e25877de4de3f58a1c5482730 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..a4c954e32e2f4546476a2e58c24fe376c7bf7389 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:423c7b5ce9ea433b1ebd72c1011c1ce8ca96c7bc1ca28b81212c078e95829b52 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..ff16f2b9059e06933fe1b822503cc37b905dc637 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bde583558a09eb45715c7e129609121fa1addebd7bba6f8de5847ea26e0ecadd +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..f2fd19e8acb93da4dc31a4169635dda8adf30f0f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dff61512f9e2ad75315bf4082d378d62b65b4cf8998c04946ac7a839e9f28745 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..985cb4b6bf9064b55def05e58cbcd59edf84f23a --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dca3e90707cbd2ca5164f93e38b5139feac6428c155f578aacf1613aeb57760a +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..ec5f36a3df0dd0f6dd4b5f61382acfb930bb5132 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13059774263fbe9f46a03541f187c20e39eda624a5f2accb81c6845467f49495 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3e35988eb5f28978c79c55d2a97fe9cb1da60a9a --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/5_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": 3.730604887008667, + "learning_rate": 2e-05, + "loss": 0.8171, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 0.3840118944644928, + "learning_rate": 2e-05, + "loss": 0.2037, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 0.18741580843925476, + "learning_rate": 2e-05, + "loss": 0.2987, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 1.2868092060089111, + "learning_rate": 2e-05, + "loss": 0.3228, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 1.2655571699142456, + "learning_rate": 2e-05, + "loss": 0.1579, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 3.0882444381713867, + "learning_rate": 2e-05, + "loss": 0.5495, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 2.750884532928467, + "learning_rate": 2e-05, + "loss": 0.1651, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 3.761305809020996, + "learning_rate": 2e-05, + "loss": 0.3604, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 2.317065954208374, + "learning_rate": 2e-05, + "loss": 0.2867, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 1.8510621786117554, + "learning_rate": 2e-05, + "loss": 0.2348, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 3.2612922191619873, + "learning_rate": 2e-05, + "loss": 0.4141, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 3.376657724380493, + "learning_rate": 2e-05, + "loss": 0.4511, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 1.7009248733520508, + "learning_rate": 2e-05, + "loss": 0.6152, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 0.4056059420108795, + "learning_rate": 2e-05, + "loss": 0.0831, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 0.1497955024242401, + "learning_rate": 2e-05, + "loss": 0.1848, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 1.188864827156067, + "learning_rate": 2e-05, + "loss": 0.1574, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 2.17380690574646, + "learning_rate": 2e-05, + "loss": 0.2598, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 2.4287848472595215, + "learning_rate": 2e-05, + "loss": 0.4716, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 2.0594663619995117, + "learning_rate": 2e-05, + "loss": 0.2934, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 1.3070108890533447, + "learning_rate": 2e-05, + "loss": 0.1507, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 1.206459403038025, + "learning_rate": 2e-05, + "loss": 0.1895, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 1.660187840461731, + "learning_rate": 2e-05, + "loss": 0.6831, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 1.2297179698944092, + "learning_rate": 2e-05, + "loss": 0.3077, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 1.9779421091079712, + "learning_rate": 2e-05, + "loss": 0.1516, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 1.9690414667129517, + "learning_rate": 2e-05, + "loss": 0.0669, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 7.822904109954834, + "learning_rate": 2e-05, + "loss": 0.4455, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 1.2651965618133545, + "learning_rate": 2e-05, + "loss": 0.1517, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 2.039531707763672, + "learning_rate": 2e-05, + "loss": 0.1359, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 14.476689338684082, + "learning_rate": 2e-05, + "loss": 0.5398, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 7.171658992767334, + "learning_rate": 2e-05, + "loss": 0.724, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 2.481220006942749, + "learning_rate": 2e-05, + "loss": 0.2396, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 1.695210337638855, + "learning_rate": 2e-05, + "loss": 0.781, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 2.885927677154541, + "learning_rate": 2e-05, + "loss": 0.1818, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 6.820782661437988, + "learning_rate": 2e-05, + "loss": 0.6167, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 1.7669086456298828, + "learning_rate": 2e-05, + "loss": 0.2454, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 2.43961238861084, + "learning_rate": 2e-05, + "loss": 0.128, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 0.6275721788406372, + "learning_rate": 2e-05, + "loss": 0.3379, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 1.4942408800125122, + "learning_rate": 2e-05, + "loss": 0.2448, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 5.283818244934082, + "learning_rate": 2e-05, + "loss": 0.3916, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 2.0611743927001953, + "learning_rate": 2e-05, + "loss": 0.4519, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 4.130486965179443, + "learning_rate": 2e-05, + "loss": 0.235, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 4.5198283195495605, + "learning_rate": 2e-05, + "loss": 0.3577, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 0.7567212581634521, + "learning_rate": 2e-05, + "loss": 0.1766, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 2.2592852115631104, + "learning_rate": 2e-05, + "loss": 0.0785, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 1.7758088111877441, + "learning_rate": 2e-05, + "loss": 0.2983, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 3.315241813659668, + "learning_rate": 2e-05, + "loss": 0.4351, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 1.6754505634307861, + "learning_rate": 2e-05, + "loss": 0.197, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 4.242250919342041, + "learning_rate": 2e-05, + "loss": 0.2225, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 2.880296468734741, + "learning_rate": 2e-05, + "loss": 0.3584, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 3.2601852416992188, + "learning_rate": 2e-05, + "loss": 0.4849, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 2.928588257807565e+16, + "train_loss": 0.3267170810699463, + "train_runtime": 643.2409, + "train_samples_per_second": 0.622, + "train_steps_per_second": 0.155 + } + ], + "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": 2.928588257807565e+16, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..1c57a6ee2c74c3947d5ffd78513afc012b4d216f --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13ba91627bc326a2235f76af60ae5bac173d6f45f44bb885885ba36c5cbf33c8 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..2cce5bfd021e3c8cc1c50d389b8df21eeb51e446 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9fd53aff9813d1113a295f7eec957217d05eafbd9ba33ce8e4fc73d3a09b9d0e +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..2e802565c15b492381b1c633c13e56581032435b --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:950e7ac88b8d514963f3c4b7e6b4f519234a3576a4bfcb1c652c42748bc4cadb +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..5fa9a98cf010816a0b21e78277cab9376c41a38e --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57b004b0f34d3a5ffc173c6f93826967ff17e9b2a46debf538505ac32b220127 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..6be7c151d4c5c1e0627025a78381ffa7b03060f1 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b5bf2594f85cb2303836c2682814f1095a1ad1fafa5b1237cdb8b1ba12c34098 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..5ca15ee49723456dad9f2334d19a4bfe52e2a5c0 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f3e52f14974d84e1e72f74f09c3d5f1e2ff72e4ef940ded7a500a6898780de5 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..3ac954174412e820d48e268c6c9db47047599372 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a221c0b2fdcdc6407319ca561fb3b8f286deaeb3a53f857a30f3464cfe09884 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..17e1d1320c235bcbe8aacad3534d3d37267325b3 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1c11e2947415362dc2f4218d2473a50caf7776155acab7eb948980e7a86ddf5b +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..3ebbf3e33426d19f2e2d73da7e4b47afc7f15563 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/6_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": 17.506067276000977, + "learning_rate": 2e-05, + "loss": 0.9927, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 19.681922912597656, + "learning_rate": 2e-05, + "loss": 1.2461, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 22.48876190185547, + "learning_rate": 2e-05, + "loss": 1.1509, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 10.4036283493042, + "learning_rate": 2e-05, + "loss": 0.9854, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 10.519641876220703, + "learning_rate": 2e-05, + "loss": 0.7529, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 8.727903366088867, + "learning_rate": 2e-05, + "loss": 1.292, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 5.802646160125732, + "learning_rate": 2e-05, + "loss": 1.1377, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 6.928057670593262, + "learning_rate": 2e-05, + "loss": 1.123, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 5.643867015838623, + "learning_rate": 2e-05, + "loss": 0.8071, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 10.85843276977539, + "learning_rate": 2e-05, + "loss": 1.2012, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 7.192744255065918, + "learning_rate": 2e-05, + "loss": 0.9346, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 18.273353576660156, + "learning_rate": 2e-05, + "loss": 1.4033, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 14.227206230163574, + "learning_rate": 2e-05, + "loss": 1.1245, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 9.324811935424805, + "learning_rate": 2e-05, + "loss": 1.2871, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 15.081768035888672, + "learning_rate": 2e-05, + "loss": 0.9326, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 6.962165832519531, + "learning_rate": 2e-05, + "loss": 0.9102, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 4.642879486083984, + "learning_rate": 2e-05, + "loss": 0.7095, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 10.674992561340332, + "learning_rate": 2e-05, + "loss": 0.5056, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 14.504895210266113, + "learning_rate": 2e-05, + "loss": 0.4336, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 6.116698741912842, + "learning_rate": 2e-05, + "loss": 0.6484, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 4.853278636932373, + "learning_rate": 2e-05, + "loss": 0.9399, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 6.737602710723877, + "learning_rate": 2e-05, + "loss": 0.8315, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 4.1324262619018555, + "learning_rate": 2e-05, + "loss": 0.4702, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 11.782748222351074, + "learning_rate": 2e-05, + "loss": 1.2056, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 3.736252784729004, + "learning_rate": 2e-05, + "loss": 0.2053, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 8.747146606445312, + "learning_rate": 2e-05, + "loss": 1.1348, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 11.472840309143066, + "learning_rate": 2e-05, + "loss": 0.9521, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 13.741485595703125, + "learning_rate": 2e-05, + "loss": 0.5205, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 4.796680927276611, + "learning_rate": 2e-05, + "loss": 0.9846, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 10.746488571166992, + "learning_rate": 2e-05, + "loss": 1.1655, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 5.629652976989746, + "learning_rate": 2e-05, + "loss": 1.699, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 8.311609268188477, + "learning_rate": 2e-05, + "loss": 0.3914, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 4.278438568115234, + "learning_rate": 2e-05, + "loss": 0.885, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 6.9923529624938965, + "learning_rate": 2e-05, + "loss": 1.3553, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 7.287336826324463, + "learning_rate": 2e-05, + "loss": 0.5132, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 12.66263198852539, + "learning_rate": 2e-05, + "loss": 0.9102, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 8.998929977416992, + "learning_rate": 2e-05, + "loss": 1.1494, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 9.578839302062988, + "learning_rate": 2e-05, + "loss": 1.2114, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 12.603957176208496, + "learning_rate": 2e-05, + "loss": 2.043, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 22.921798706054688, + "learning_rate": 2e-05, + "loss": 0.714, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 6.382108211517334, + "learning_rate": 2e-05, + "loss": 0.656, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 17.197595596313477, + "learning_rate": 2e-05, + "loss": 1.4766, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 14.173710823059082, + "learning_rate": 2e-05, + "loss": 1.9785, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 7.996662616729736, + "learning_rate": 2e-05, + "loss": 0.6284, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 9.453429222106934, + "learning_rate": 2e-05, + "loss": 1.4189, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 5.327342510223389, + "learning_rate": 2e-05, + "loss": 1.0596, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 7.857007026672363, + "learning_rate": 2e-05, + "loss": 0.686, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 5.651580333709717, + "learning_rate": 2e-05, + "loss": 0.5703, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 5.1162285804748535, + "learning_rate": 2e-05, + "loss": 0.4973, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 5.699977874755859, + "learning_rate": 2e-05, + "loss": 0.6392, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 3199568636280832.0, + "train_loss": 0.969423828125, + "train_runtime": 142.2122, + "train_samples_per_second": 2.813, + "train_steps_per_second": 0.703 + } + ], + "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": 3199568636280832.0, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..a3cef4b111a9e8612e2082e6ce67fd06b6a49352 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91411563d1d342eff7e6ff598916653a1e7a54658378c0dfaebe3ed2e5ffb9a2 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..326548d79497c2b6c99b47dd5bcb40c924a61afd --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:72f918ef5470faa5ccd0309942abe7631d44ed1a436034fcee866851da72fcfa +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..9e6e2acfcdbe55fc3d58f9b48631d69c9f510348 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b0d767346a6bfdd92aeafca66a12038a3dd5b67b20145aff839af578cacc896c +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..cd0faac92b524ae184796488dde65e891c7a75df --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:03ee60abe50a2e9fe75c8be9a9181380e59bf2ab7f67a1d882255737a95f4cd2 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..6367072e7c2965bd566acb339268cb14e1bbde2d --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2d29b458615ffa8cfd5c633ab2a99b3286e335b441e6c167563115e8d208f0fa +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..3f6f58038e4d52134294f66e0ee7ba9e16af0923 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9c2a54b443fccc5feadc6e9b741f7c835409e88dd8b5d3872fea77fba44ee29 +size 395787774 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..89287a5e5e5f886f520119e5cdd243df7000faf2 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:152003078f9ecf28c35f3ca3e38230b9ee51e4d5cf8c70ab2cadaf58aa5c9de0 +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..29c90a8594d1b272db4c98b4000f889118b8fd55 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5764622561c369eb52a7c4eea5791018cca2d77598d4474c6d3e152bea9ba26a +size 395786922 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..eec68a6db0af764727e550761d560c6c2587d138 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/7_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": 10.693034172058105, + "learning_rate": 2e-05, + "loss": 0.9473, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 7.147550582885742, + "learning_rate": 2e-05, + "loss": 0.8438, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 2.084887742996216, + "learning_rate": 2e-05, + "loss": 0.7456, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 6.632413864135742, + "learning_rate": 2e-05, + "loss": 0.7954, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 4.06978702545166, + "learning_rate": 2e-05, + "loss": 0.4501, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 4.178534984588623, + "learning_rate": 2e-05, + "loss": 0.5776, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 5.342024326324463, + "learning_rate": 2e-05, + "loss": 0.4456, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 3.150102376937866, + "learning_rate": 2e-05, + "loss": 0.2704, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 5.205374717712402, + "learning_rate": 2e-05, + "loss": 0.5511, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 4.769977569580078, + "learning_rate": 2e-05, + "loss": 0.7381, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 4.061365127563477, + "learning_rate": 2e-05, + "loss": 0.7803, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 4.420660495758057, + "learning_rate": 2e-05, + "loss": 0.3429, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 4.9685163497924805, + "learning_rate": 2e-05, + "loss": 0.3986, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 4.212514877319336, + "learning_rate": 2e-05, + "loss": 1.7986, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 3.7489864826202393, + "learning_rate": 2e-05, + "loss": 0.4519, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 1.6335949897766113, + "learning_rate": 2e-05, + "loss": 0.3369, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 4.586492538452148, + "learning_rate": 2e-05, + "loss": 0.4424, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 6.819537162780762, + "learning_rate": 2e-05, + "loss": 0.6816, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 9.399861335754395, + "learning_rate": 2e-05, + "loss": 0.3721, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 6.3716912269592285, + "learning_rate": 2e-05, + "loss": 0.4189, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 3.045827627182007, + "learning_rate": 2e-05, + "loss": 0.2898, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 4.605010509490967, + "learning_rate": 2e-05, + "loss": 0.6685, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 6.698479175567627, + "learning_rate": 2e-05, + "loss": 0.4849, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 3.601922035217285, + "learning_rate": 2e-05, + "loss": 0.7192, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 5.282998561859131, + "learning_rate": 2e-05, + "loss": 0.5659, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 6.072550296783447, + "learning_rate": 2e-05, + "loss": 1.0957, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 5.633040428161621, + "learning_rate": 2e-05, + "loss": 0.8174, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 7.349772930145264, + "learning_rate": 2e-05, + "loss": 0.7959, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 4.335024356842041, + "learning_rate": 2e-05, + "loss": 0.8176, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 5.084471702575684, + "learning_rate": 2e-05, + "loss": 0.4082, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 4.496095657348633, + "learning_rate": 2e-05, + "loss": 0.4316, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 6.351251125335693, + "learning_rate": 2e-05, + "loss": 0.522, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 5.976203918457031, + "learning_rate": 2e-05, + "loss": 0.8784, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 6.05720853805542, + "learning_rate": 2e-05, + "loss": 0.9539, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 8.505938529968262, + "learning_rate": 2e-05, + "loss": 0.79, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 7.8473052978515625, + "learning_rate": 2e-05, + "loss": 0.6582, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 6.188315391540527, + "learning_rate": 2e-05, + "loss": 1.1504, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 5.365588665008545, + "learning_rate": 2e-05, + "loss": 0.3195, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 8.43429183959961, + "learning_rate": 2e-05, + "loss": 1.1396, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 5.884682655334473, + "learning_rate": 2e-05, + "loss": 0.7715, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 8.613433837890625, + "learning_rate": 2e-05, + "loss": 0.5801, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 3.94631028175354, + "learning_rate": 2e-05, + "loss": 0.4038, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 8.417205810546875, + "learning_rate": 2e-05, + "loss": 0.6838, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 5.982354164123535, + "learning_rate": 2e-05, + "loss": 0.8743, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 3.4709291458129883, + "learning_rate": 2e-05, + "loss": 0.4902, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 3.4178287982940674, + "learning_rate": 2e-05, + "loss": 0.9113, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 5.503345012664795, + "learning_rate": 2e-05, + "loss": 0.5105, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 4.983687400817871, + "learning_rate": 2e-05, + "loss": 0.4043, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 13.622444152832031, + "learning_rate": 2e-05, + "loss": 1.0083, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 5.9070258140563965, + "learning_rate": 2e-05, + "loss": 0.5065, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 1.145241813909504e+16, + "train_loss": 0.6608060073852539, + "train_runtime": 226.6469, + "train_samples_per_second": 1.765, + "train_steps_per_second": 0.441 + } + ], + "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": 1.145241813909504e+16, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..149c573c70ab36d1c4fc71525dca2548dd9fecc4 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91469fa1de3cb8972630ee07e5f38cd02278e2ab1f71e076ffa6da12b00110a9 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..8b3251d059424787c3e9bc6a99891cd2e8fc7547 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19781847106cd9b8f2902c56a0401ffc43719b0e199a817a90a282c34f25e15e +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..e492ea4eb356c390d40e7c723364318cc1631cbe --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4efc6c110b07fd0e331b94293434409b92b7c32323f5f36a6ea2383d4c6bddec +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..13e56333a04dff4524c95462b7906109e2f1609e --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7db6a6be9bb547a66223850aa6db214f84116d46c54fbb9b4663cd57f6c85ac +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..f06f3b30fdb78a0617270d42b705764e18ac5db5 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19ce0b48d690e7997a27aed763c2522e2719c273476f3211c89271099fae639b +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..2825e24e1b444b2b165d4f6bc8863854b41ec2b8 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:075176d64de4aa1534986070e685130f1cc805832d09ddb4a3ac1aec07035988 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..b06b56fb988ebb8ed676af9df51fc479d57e9e38 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:373249442bd79d47a41a9bbb9d86ada37469b5d31a2375c4114285e0b774b65b +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..a2a49916c956fe958c40b8c39ff9b79d891ee562 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1229f678e9503e475cd925d395446ff20ddb4e7e94b84a1debc93fc8e266fd30 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..139e07d1cd13a2118c21ad524e7bff0efa0408c9 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/8_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": 2.193044662475586, + "learning_rate": 2e-05, + "loss": 0.585, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 5.7211737632751465, + "learning_rate": 2e-05, + "loss": 1.0234, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 6.0524420738220215, + "learning_rate": 2e-05, + "loss": 0.6735, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 7.524207592010498, + "learning_rate": 2e-05, + "loss": 1.6274, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 7.20663595199585, + "learning_rate": 2e-05, + "loss": 1.793, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 5.621150970458984, + "learning_rate": 2e-05, + "loss": 1.1902, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 3.946408271789551, + "learning_rate": 2e-05, + "loss": 0.6891, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 6.071298599243164, + "learning_rate": 2e-05, + "loss": 0.6671, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 7.1580047607421875, + "learning_rate": 2e-05, + "loss": 0.8218, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 13.339744567871094, + "learning_rate": 2e-05, + "loss": 1.4189, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 8.02275562286377, + "learning_rate": 2e-05, + "loss": 1.2061, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 10.094612121582031, + "learning_rate": 2e-05, + "loss": 1.0801, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 8.221661567687988, + "learning_rate": 2e-05, + "loss": 1.1914, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 4.561307430267334, + "learning_rate": 2e-05, + "loss": 1.3244, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 3.7538561820983887, + "learning_rate": 2e-05, + "loss": 0.7314, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 9.366249084472656, + "learning_rate": 2e-05, + "loss": 1.5118, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 7.684645652770996, + "learning_rate": 2e-05, + "loss": 0.7195, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 15.185140609741211, + "learning_rate": 2e-05, + "loss": 1.2396, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 2.5864124298095703, + "learning_rate": 2e-05, + "loss": 0.4047, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 8.702391624450684, + "learning_rate": 2e-05, + "loss": 1.6504, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 6.429192066192627, + "learning_rate": 2e-05, + "loss": 0.6327, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 8.933626174926758, + "learning_rate": 2e-05, + "loss": 1.4546, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 7.802511692047119, + "learning_rate": 2e-05, + "loss": 0.3814, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 5.308834552764893, + "learning_rate": 2e-05, + "loss": 0.719, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 5.791723251342773, + "learning_rate": 2e-05, + "loss": 1.353, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 8.101655006408691, + "learning_rate": 2e-05, + "loss": 1.0791, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 12.4312162399292, + "learning_rate": 2e-05, + "loss": 1.0552, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 9.454916000366211, + "learning_rate": 2e-05, + "loss": 1.4326, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 5.968297958374023, + "learning_rate": 2e-05, + "loss": 2.0225, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 9.31320858001709, + "learning_rate": 2e-05, + "loss": 1.9663, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 11.541924476623535, + "learning_rate": 2e-05, + "loss": 1.2471, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 6.128070831298828, + "learning_rate": 2e-05, + "loss": 1.2217, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 11.803208351135254, + "learning_rate": 2e-05, + "loss": 1.0049, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 12.641098022460938, + "learning_rate": 2e-05, + "loss": 1.3701, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 14.977377891540527, + "learning_rate": 2e-05, + "loss": 1.2678, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 11.000706672668457, + "learning_rate": 2e-05, + "loss": 1.3867, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 6.535886764526367, + "learning_rate": 2e-05, + "loss": 0.8286, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 24.05946159362793, + "learning_rate": 2e-05, + "loss": 3.0078, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 8.536385536193848, + "learning_rate": 2e-05, + "loss": 1.1467, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 4.476486682891846, + "learning_rate": 2e-05, + "loss": 0.4897, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 5.659985542297363, + "learning_rate": 2e-05, + "loss": 0.7368, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 8.377525329589844, + "learning_rate": 2e-05, + "loss": 1.0702, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 8.25423526763916, + "learning_rate": 2e-05, + "loss": 2.793, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 6.275785446166992, + "learning_rate": 2e-05, + "loss": 1.3789, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 14.999366760253906, + "learning_rate": 2e-05, + "loss": 1.0403, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 5.93328857421875, + "learning_rate": 2e-05, + "loss": 0.7925, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 4.31484317779541, + "learning_rate": 2e-05, + "loss": 0.2299, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 7.170931816101074, + "learning_rate": 2e-05, + "loss": 0.7734, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 7.826891899108887, + "learning_rate": 2e-05, + "loss": 0.582, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 12.407968521118164, + "learning_rate": 2e-05, + "loss": 0.8096, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 2088246297231360.0, + "train_loss": 1.1364552688598633, + "train_runtime": 128.9398, + "train_samples_per_second": 3.102, + "train_steps_per_second": 0.776 + } + ], + "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": 2088246297231360.0, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +} diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round10.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round10.pth new file mode 100644 index 0000000000000000000000000000000000000000..ed3eaa4e2e1cf6e63e4d2d5920d89457cb5aa714 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round10.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50db3f5a6278503e0d5a7315a8a6a8666299d0213a8e9ee569a98455ae783f1f +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round12.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round12.pth new file mode 100644 index 0000000000000000000000000000000000000000..267bed0bc2f8a0aab73df6065921c675b3c3efcb --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round12.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff47a3de488fc33a3c45913aad5e7134b005f5728926f8aa33e78ad48ee5181f +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round15.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round15.pth new file mode 100644 index 0000000000000000000000000000000000000000..228ac01c6f60a4d8a1931b44c39bbeba7887df00 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round15.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d80f8fafda57899c128e31b6b846b9972b86946076d93f275d6b5ed57f2aec4a +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round17.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round17.pth new file mode 100644 index 0000000000000000000000000000000000000000..d24664da681834a033d8b89cc162d6e1f9735549 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round17.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4f617052c2a55a553cfec754c00266130de682ee97414f24a48aff1eb8c2086 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round2.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round2.pth new file mode 100644 index 0000000000000000000000000000000000000000..3ec0d58740e54ffd3564669bf7ca544877871f5c --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round2.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9e8421b29bc0fdac5bd5bfe5e6982554a1f392e5aebbde5597fb24dfb21d1f7 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round20.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round20.pth new file mode 100644 index 0000000000000000000000000000000000000000..99c5ed7c63ccf328643a29d395c2081f4a916a30 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round20.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a47eafbd014612333736ec014d98cb177b5fdbec33eceeeea2f1faa7e1b8c3a5 +size 184221358 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round5.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round5.pth new file mode 100644 index 0000000000000000000000000000000000000000..7ab2114d809ae2c9e90af6bc037cbbbdce769e38 --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round5.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:055682d4a23d27001e9910af45e3559ffa57c0de5cc7a20016d64135f0a37a02 +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round7.pth b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round7.pth new file mode 100644 index 0000000000000000000000000000000000000000..38154beac2b54639a5db43915144e3caf159d78c --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_client_model_round7.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4a3128f4292183a01be4d3aec09f472e5f007c3fc5aac5fa67f8acf1bb35baa +size 184220842 diff --git a/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_trainer_state.json b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..c7a2876e40700b5665468b7250c4ff3c10babc7a --- /dev/null +++ b/client_states_fedMultipqfullfreeze_homoAgg_pca_bs4_saveoptim_lr2e-5_5e-5_sc132_4tasks_5rounds_fixitr100_T0125_decay099/9_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": 10.358270645141602, + "learning_rate": 2e-05, + "loss": 0.9292, + "step": 2 + }, + { + "epoch": 0.04, + "grad_norm": 8.892541885375977, + "learning_rate": 2e-05, + "loss": 1.2041, + "step": 4 + }, + { + "epoch": 0.06, + "grad_norm": 14.91721248626709, + "learning_rate": 2e-05, + "loss": 0.7919, + "step": 6 + }, + { + "epoch": 0.08, + "grad_norm": 8.63127613067627, + "learning_rate": 2e-05, + "loss": 1.25, + "step": 8 + }, + { + "epoch": 0.1, + "grad_norm": 8.962491989135742, + "learning_rate": 2e-05, + "loss": 1.6318, + "step": 10 + }, + { + "epoch": 0.12, + "grad_norm": 11.209757804870605, + "learning_rate": 2e-05, + "loss": 0.5818, + "step": 12 + }, + { + "epoch": 0.14, + "grad_norm": 11.37177848815918, + "learning_rate": 2e-05, + "loss": 1.0581, + "step": 14 + }, + { + "epoch": 0.16, + "grad_norm": 5.2912726402282715, + "learning_rate": 2e-05, + "loss": 0.9768, + "step": 16 + }, + { + "epoch": 0.18, + "grad_norm": 10.864551544189453, + "learning_rate": 2e-05, + "loss": 1.2168, + "step": 18 + }, + { + "epoch": 0.2, + "grad_norm": 11.158872604370117, + "learning_rate": 2e-05, + "loss": 0.4907, + "step": 20 + }, + { + "epoch": 0.22, + "grad_norm": 10.59311294555664, + "learning_rate": 2e-05, + "loss": 0.8994, + "step": 22 + }, + { + "epoch": 0.24, + "grad_norm": 7.94593620300293, + "learning_rate": 2e-05, + "loss": 0.5808, + "step": 24 + }, + { + "epoch": 0.26, + "grad_norm": 12.64927864074707, + "learning_rate": 2e-05, + "loss": 1.1035, + "step": 26 + }, + { + "epoch": 0.28, + "grad_norm": 9.098578453063965, + "learning_rate": 2e-05, + "loss": 0.5355, + "step": 28 + }, + { + "epoch": 0.3, + "grad_norm": 22.1839599609375, + "learning_rate": 2e-05, + "loss": 0.7393, + "step": 30 + }, + { + "epoch": 0.32, + "grad_norm": 17.685653686523438, + "learning_rate": 2e-05, + "loss": 0.7847, + "step": 32 + }, + { + "epoch": 0.34, + "grad_norm": 5.024636268615723, + "learning_rate": 2e-05, + "loss": 0.562, + "step": 34 + }, + { + "epoch": 0.36, + "grad_norm": 8.262999534606934, + "learning_rate": 2e-05, + "loss": 1.3672, + "step": 36 + }, + { + "epoch": 0.38, + "grad_norm": 49.34255599975586, + "learning_rate": 2e-05, + "loss": 1.0601, + "step": 38 + }, + { + "epoch": 0.4, + "grad_norm": 11.732015609741211, + "learning_rate": 2e-05, + "loss": 0.7954, + "step": 40 + }, + { + "epoch": 0.42, + "grad_norm": 7.60172176361084, + "learning_rate": 2e-05, + "loss": 1.0292, + "step": 42 + }, + { + "epoch": 0.44, + "grad_norm": 8.254109382629395, + "learning_rate": 2e-05, + "loss": 1.1963, + "step": 44 + }, + { + "epoch": 0.46, + "grad_norm": 5.902849197387695, + "learning_rate": 2e-05, + "loss": 0.4211, + "step": 46 + }, + { + "epoch": 0.48, + "grad_norm": 8.962932586669922, + "learning_rate": 2e-05, + "loss": 0.6929, + "step": 48 + }, + { + "epoch": 0.5, + "grad_norm": 8.662166595458984, + "learning_rate": 2e-05, + "loss": 0.6704, + "step": 50 + }, + { + "epoch": 0.52, + "grad_norm": 12.05443000793457, + "learning_rate": 2e-05, + "loss": 1.1543, + "step": 52 + }, + { + "epoch": 0.54, + "grad_norm": 15.137462615966797, + "learning_rate": 2e-05, + "loss": 1.376, + "step": 54 + }, + { + "epoch": 0.56, + "grad_norm": 18.348587036132812, + "learning_rate": 2e-05, + "loss": 1.2266, + "step": 56 + }, + { + "epoch": 0.58, + "grad_norm": 14.352018356323242, + "learning_rate": 2e-05, + "loss": 1.1377, + "step": 58 + }, + { + "epoch": 0.6, + "grad_norm": 6.8460822105407715, + "learning_rate": 2e-05, + "loss": 0.6753, + "step": 60 + }, + { + "epoch": 0.62, + "grad_norm": 10.520755767822266, + "learning_rate": 2e-05, + "loss": 1.4087, + "step": 62 + }, + { + "epoch": 0.64, + "grad_norm": 8.770384788513184, + "learning_rate": 2e-05, + "loss": 0.6914, + "step": 64 + }, + { + "epoch": 0.66, + "grad_norm": 13.007487297058105, + "learning_rate": 2e-05, + "loss": 0.7551, + "step": 66 + }, + { + "epoch": 0.68, + "grad_norm": 6.915706634521484, + "learning_rate": 2e-05, + "loss": 1.8792, + "step": 68 + }, + { + "epoch": 0.7, + "grad_norm": 11.421902656555176, + "learning_rate": 2e-05, + "loss": 1.209, + "step": 70 + }, + { + "epoch": 0.72, + "grad_norm": 7.313283443450928, + "learning_rate": 2e-05, + "loss": 1.5264, + "step": 72 + }, + { + "epoch": 0.74, + "grad_norm": 5.061800956726074, + "learning_rate": 2e-05, + "loss": 0.5161, + "step": 74 + }, + { + "epoch": 0.76, + "grad_norm": 11.085064888000488, + "learning_rate": 2e-05, + "loss": 1.1172, + "step": 76 + }, + { + "epoch": 0.78, + "grad_norm": 6.620267391204834, + "learning_rate": 2e-05, + "loss": 0.564, + "step": 78 + }, + { + "epoch": 0.8, + "grad_norm": 6.493375778198242, + "learning_rate": 2e-05, + "loss": 0.8389, + "step": 80 + }, + { + "epoch": 0.82, + "grad_norm": 11.706432342529297, + "learning_rate": 2e-05, + "loss": 0.9404, + "step": 82 + }, + { + "epoch": 0.84, + "grad_norm": 6.974967956542969, + "learning_rate": 2e-05, + "loss": 1.146, + "step": 84 + }, + { + "epoch": 0.86, + "grad_norm": 9.994638442993164, + "learning_rate": 2e-05, + "loss": 0.8423, + "step": 86 + }, + { + "epoch": 0.88, + "grad_norm": 9.461267471313477, + "learning_rate": 2e-05, + "loss": 1.2178, + "step": 88 + }, + { + "epoch": 0.9, + "grad_norm": 13.758880615234375, + "learning_rate": 2e-05, + "loss": 1.0029, + "step": 90 + }, + { + "epoch": 0.92, + "grad_norm": 11.697429656982422, + "learning_rate": 2e-05, + "loss": 0.9395, + "step": 92 + }, + { + "epoch": 0.94, + "grad_norm": 6.205819129943848, + "learning_rate": 2e-05, + "loss": 0.5972, + "step": 94 + }, + { + "epoch": 0.96, + "grad_norm": 18.386341094970703, + "learning_rate": 2e-05, + "loss": 1.0693, + "step": 96 + }, + { + "epoch": 0.98, + "grad_norm": 15.202548027038574, + "learning_rate": 2e-05, + "loss": 1.1914, + "step": 98 + }, + { + "epoch": 1.0, + "grad_norm": 9.901387214660645, + "learning_rate": 2e-05, + "loss": 0.5671, + "step": 100 + }, + { + "epoch": 1.0, + "step": 100, + "total_flos": 3205687391813632.0, + "train_loss": 0.9631697082519531, + "train_runtime": 72.1553, + "train_samples_per_second": 5.544, + "train_steps_per_second": 1.386 + } + ], + "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": 3205687391813632.0, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +}