Hugging Face's logo

thkim0305
/
feddat_baselines

Model card Files Community
thkim0305 commited on
Commit
54b6f9d
·
verified ·
1 Parent(s): ee2b77a

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round10.pth +3 -0
  2. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round12.pth +3 -0
  3. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round15.pth +3 -0
  4. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round17.pth +3 -0
  5. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round2.pth +3 -0
  6. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round20.pth +3 -0
  7. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round5.pth +3 -0
  8. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round7.pth +3 -0
  9. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_trainer_state.json +126 -0
  10. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round10.pth +3 -0
  11. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round12.pth +3 -0
  12. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round15.pth +3 -0
  13. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round17.pth +3 -0
  14. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round2.pth +3 -0
  15. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round20.pth +3 -0
  16. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round5.pth +3 -0
  17. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round7.pth +3 -0
  18. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_trainer_state.json +126 -0
  19. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round10.pth +3 -0
  20. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round12.pth +3 -0
  21. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round15.pth +3 -0
  22. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round17.pth +3 -0
  23. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round2.pth +3 -0
  24. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round20.pth +3 -0
  25. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round5.pth +3 -0
  26. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round7.pth +3 -0
  27. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_trainer_state.json +126 -0
  28. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round10.pth +3 -0
  29. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round12.pth +3 -0
  30. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round15.pth +3 -0
  31. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round17.pth +3 -0
  32. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round2.pth +3 -0
  33. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round20.pth +3 -0
  34. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round5.pth +3 -0
  35. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round7.pth +3 -0
  36. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_trainer_state.json +126 -0
  37. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round10.pth +3 -0
  38. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round12.pth +3 -0
  39. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round15.pth +3 -0
  40. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round17.pth +3 -0
  41. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round2.pth +3 -0
  42. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round20.pth +3 -0
  43. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round5.pth +3 -0
  44. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round7.pth +3 -0
  45. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_trainer_state.json +126 -0
  46. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round10.pth +3 -0
  47. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round12.pth +3 -0
  48. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round15.pth +3 -0
  49. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round17.pth +3 -0
  50. client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round2.pth +3 -0
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9c8b688e7142618dcaf75b8821e014cceda3e8979fdce5093a9069a4a1a4ec9
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff1a052e82f4eec62a35dd8764b6aad1cbebfe9014a124bb5ba10ef4da7d8f00
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81e8decf5dcc4fb57f4b87504e65019c233bee4c47314f4ea16cb985e230fa53
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round17.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f1d5afb7dd124ccb5cc85da7598e84be573f582e7f939d1d62706208f509aba
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45586c6182565a1259c385222e9e0ce6254e0caf263527bc1e7d5a4329d199ca
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round20.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e284ece48ddd1a8a3b8dd541af6aa3681f6ea653d2297d78c229020ce04ee233
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3651434d61d8d6f5b342f17dbdf811e02b3e075489657fd6da20fbe7c30e969d
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_client_model_round7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d59111607207bd02602038a0e259bd8995126bbf701addb0c29168eb26291efb
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/0_trainer_state.json ADDED
@@ -0,0 +1,126 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.0,
5
+ "eval_steps": 500,
6
+ "global_step": 25,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.16,
13
+ "grad_norm": 2.21482253074646,
14
+ "learning_rate": 2e-05,
15
+ "loss": 0.254,
16
+ "step": 2
17
+ },
18
+ {
19
+ "epoch": 0.32,
20
+ "grad_norm": 4.312131881713867,
21
+ "learning_rate": 2e-05,
22
+ "loss": 0.1656,
23
+ "step": 4
24
+ },
25
+ {
26
+ "epoch": 0.48,
27
+ "grad_norm": 1.3984471559524536,
28
+ "learning_rate": 2e-05,
29
+ "loss": 0.046,
30
+ "step": 6
31
+ },
32
+ {
33
+ "epoch": 0.64,
34
+ "grad_norm": 1.9168689250946045,
35
+ "learning_rate": 2e-05,
36
+ "loss": 0.0957,
37
+ "step": 8
38
+ },
39
+ {
40
+ "epoch": 0.8,
41
+ "grad_norm": 2.0672812461853027,
42
+ "learning_rate": 2e-05,
43
+ "loss": 0.281,
44
+ "step": 10
45
+ },
46
+ {
47
+ "epoch": 0.96,
48
+ "grad_norm": 0.5915671586990356,
49
+ "learning_rate": 2e-05,
50
+ "loss": 0.0233,
51
+ "step": 12
52
+ },
53
+ {
54
+ "epoch": 1.12,
55
+ "grad_norm": 2.9408771991729736,
56
+ "learning_rate": 2e-05,
57
+ "loss": 0.4947,
58
+ "step": 14
59
+ },
60
+ {
61
+ "epoch": 1.28,
62
+ "grad_norm": 2.808587074279785,
63
+ "learning_rate": 2e-05,
64
+ "loss": 0.1121,
65
+ "step": 16
66
+ },
67
+ {
68
+ "epoch": 1.44,
69
+ "grad_norm": 0.9880923628807068,
70
+ "learning_rate": 2e-05,
71
+ "loss": 0.0909,
72
+ "step": 18
73
+ },
74
+ {
75
+ "epoch": 1.6,
76
+ "grad_norm": 0.9719083905220032,
77
+ "learning_rate": 2e-05,
78
+ "loss": 0.0377,
79
+ "step": 20
80
+ },
81
+ {
82
+ "epoch": 1.76,
83
+ "grad_norm": 3.7146573066711426,
84
+ "learning_rate": 2e-05,
85
+ "loss": 0.1087,
86
+ "step": 22
87
+ },
88
+ {
89
+ "epoch": 1.92,
90
+ "grad_norm": 1.7789965867996216,
91
+ "learning_rate": 2e-05,
92
+ "loss": 0.2933,
93
+ "step": 24
94
+ },
95
+ {
96
+ "epoch": 2.0,
97
+ "step": 25,
98
+ "total_flos": 2776472045486080.0,
99
+ "train_loss": 0.16410446166992188,
100
+ "train_runtime": 127.3475,
101
+ "train_samples_per_second": 0.785,
102
+ "train_steps_per_second": 0.196
103
+ }
104
+ ],
105
+ "logging_steps": 2,
106
+ "max_steps": 25,
107
+ "num_input_tokens_seen": 0,
108
+ "num_train_epochs": 1,
109
+ "save_steps": 500,
110
+ "stateful_callbacks": {
111
+ "TrainerControl": {
112
+ "args": {
113
+ "should_epoch_stop": false,
114
+ "should_evaluate": false,
115
+ "should_log": false,
116
+ "should_save": false,
117
+ "should_training_stop": false
118
+ },
119
+ "attributes": {}
120
+ }
121
+ },
122
+ "total_flos": 2776472045486080.0,
123
+ "train_batch_size": 1,
124
+ "trial_name": null,
125
+ "trial_params": null
126
+ }
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5d734864daeeef65eb7021dea9335af29f9eb355c5c24bd18f1f4c800053454
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84fb2ded09a75d5375a6580465a16635b5647ea5747b60010cf13c0ff84aad83
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:100f1537486f1307e3311d9dff294c4f3f7ea4855d42a169a52b0478a17f20dc
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round17.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:499d6a2dc853680f34ad981ad4ebcc92ade1be79e80bae61d9e21313f4749473
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bbdfaa7ec410f0e21e4be7406985d711d6442247f71b2940a42598fede71a62a
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round20.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e599b8fca50af010e1b7f3b36515b2d844e07af4e70c799597761b47654e3f7d
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc518d42a7836ec09008231870f9f4a7ed123eaeeeb09ddeaa7890249744b85e
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_client_model_round7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:717b1af67a9b2cf721e2297314d6294216f3aacd81c35d45554c3b7af1d22e52
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/1_trainer_state.json ADDED
@@ -0,0 +1,126 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.0,
5
+ "eval_steps": 500,
6
+ "global_step": 25,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.16,
13
+ "grad_norm": 0.9727327823638916,
14
+ "learning_rate": 2e-05,
15
+ "loss": 0.1592,
16
+ "step": 2
17
+ },
18
+ {
19
+ "epoch": 0.32,
20
+ "grad_norm": 0.6119529008865356,
21
+ "learning_rate": 2e-05,
22
+ "loss": 0.0155,
23
+ "step": 4
24
+ },
25
+ {
26
+ "epoch": 0.48,
27
+ "grad_norm": 0.3699932098388672,
28
+ "learning_rate": 2e-05,
29
+ "loss": 0.1359,
30
+ "step": 6
31
+ },
32
+ {
33
+ "epoch": 0.64,
34
+ "grad_norm": 0.8428669571876526,
35
+ "learning_rate": 2e-05,
36
+ "loss": 0.0421,
37
+ "step": 8
38
+ },
39
+ {
40
+ "epoch": 0.8,
41
+ "grad_norm": 0.8032307028770447,
42
+ "learning_rate": 2e-05,
43
+ "loss": 0.0221,
44
+ "step": 10
45
+ },
46
+ {
47
+ "epoch": 0.96,
48
+ "grad_norm": 3.484509229660034,
49
+ "learning_rate": 2e-05,
50
+ "loss": 0.0962,
51
+ "step": 12
52
+ },
53
+ {
54
+ "epoch": 1.12,
55
+ "grad_norm": 0.48723894357681274,
56
+ "learning_rate": 2e-05,
57
+ "loss": 0.0181,
58
+ "step": 14
59
+ },
60
+ {
61
+ "epoch": 1.28,
62
+ "grad_norm": 1.0781177282333374,
63
+ "learning_rate": 2e-05,
64
+ "loss": 0.0448,
65
+ "step": 16
66
+ },
67
+ {
68
+ "epoch": 1.44,
69
+ "grad_norm": 2.618180513381958,
70
+ "learning_rate": 2e-05,
71
+ "loss": 0.0524,
72
+ "step": 18
73
+ },
74
+ {
75
+ "epoch": 1.6,
76
+ "grad_norm": 1.0576213598251343,
77
+ "learning_rate": 2e-05,
78
+ "loss": 0.0496,
79
+ "step": 20
80
+ },
81
+ {
82
+ "epoch": 1.76,
83
+ "grad_norm": 0.34326815605163574,
84
+ "learning_rate": 2e-05,
85
+ "loss": 0.0519,
86
+ "step": 22
87
+ },
88
+ {
89
+ "epoch": 1.92,
90
+ "grad_norm": 0.4327654540538788,
91
+ "learning_rate": 2e-05,
92
+ "loss": 0.0183,
93
+ "step": 24
94
+ },
95
+ {
96
+ "epoch": 2.0,
97
+ "step": 25,
98
+ "total_flos": 2795254289793024.0,
99
+ "train_loss": 0.056924142837524415,
100
+ "train_runtime": 128.058,
101
+ "train_samples_per_second": 0.781,
102
+ "train_steps_per_second": 0.195
103
+ }
104
+ ],
105
+ "logging_steps": 2,
106
+ "max_steps": 25,
107
+ "num_input_tokens_seen": 0,
108
+ "num_train_epochs": 1,
109
+ "save_steps": 500,
110
+ "stateful_callbacks": {
111
+ "TrainerControl": {
112
+ "args": {
113
+ "should_epoch_stop": false,
114
+ "should_evaluate": false,
115
+ "should_log": false,
116
+ "should_save": false,
117
+ "should_training_stop": false
118
+ },
119
+ "attributes": {}
120
+ }
121
+ },
122
+ "total_flos": 2795254289793024.0,
123
+ "train_batch_size": 1,
124
+ "trial_name": null,
125
+ "trial_params": null
126
+ }
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04fecbc85501e3f0cbafe9f13108201d8d0cc03eee280d106fab39629624a568
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:900c97291ffa7ede12bdecdf80234c6c599d18f68c242a39b512946e6c2f8cf8
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13ca03010d565b6648c1c2de72ac1971fc3ee336ef441ce2e2aed880590d3b3a
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round17.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a1389dff1c58aca0343be04ac4ada1f12834b18b9ba70342bcb3bb424deb3eb
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53b54b947a97e393bb76780d4174ca2bbe266b6d5e26d162c4f4a73a0797ac57
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round20.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:adb0272b9ee55de14b7c86ea70901588d1a894e126974d841e223f610b306e7a
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c44442470cf3659d87e0e4d6a63a497b9fbf71c54dce14abe746671e4579c02e
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_client_model_round7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e9064a8ac3feffdefe6bf934222792abbcf01e91958439277a311697049e0fd
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/2_trainer_state.json ADDED
@@ -0,0 +1,126 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.0,
5
+ "eval_steps": 500,
6
+ "global_step": 25,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.16,
13
+ "grad_norm": 3.542206287384033,
14
+ "learning_rate": 2e-05,
15
+ "loss": 0.3763,
16
+ "step": 2
17
+ },
18
+ {
19
+ "epoch": 0.32,
20
+ "grad_norm": 0.5754655003547668,
21
+ "learning_rate": 2e-05,
22
+ "loss": 0.0481,
23
+ "step": 4
24
+ },
25
+ {
26
+ "epoch": 0.48,
27
+ "grad_norm": 2.7304298877716064,
28
+ "learning_rate": 2e-05,
29
+ "loss": 0.3178,
30
+ "step": 6
31
+ },
32
+ {
33
+ "epoch": 0.64,
34
+ "grad_norm": 0.6067225337028503,
35
+ "learning_rate": 2e-05,
36
+ "loss": 0.1304,
37
+ "step": 8
38
+ },
39
+ {
40
+ "epoch": 0.8,
41
+ "grad_norm": 2.869699478149414,
42
+ "learning_rate": 2e-05,
43
+ "loss": 0.3075,
44
+ "step": 10
45
+ },
46
+ {
47
+ "epoch": 0.96,
48
+ "grad_norm": 0.8252665400505066,
49
+ "learning_rate": 2e-05,
50
+ "loss": 0.052,
51
+ "step": 12
52
+ },
53
+ {
54
+ "epoch": 1.12,
55
+ "grad_norm": 0.7447052001953125,
56
+ "learning_rate": 2e-05,
57
+ "loss": 0.1314,
58
+ "step": 14
59
+ },
60
+ {
61
+ "epoch": 1.28,
62
+ "grad_norm": 1.8536393642425537,
63
+ "learning_rate": 2e-05,
64
+ "loss": 0.156,
65
+ "step": 16
66
+ },
67
+ {
68
+ "epoch": 1.44,
69
+ "grad_norm": 2.449882984161377,
70
+ "learning_rate": 2e-05,
71
+ "loss": 0.4599,
72
+ "step": 18
73
+ },
74
+ {
75
+ "epoch": 1.6,
76
+ "grad_norm": 2.2815041542053223,
77
+ "learning_rate": 2e-05,
78
+ "loss": 0.1967,
79
+ "step": 20
80
+ },
81
+ {
82
+ "epoch": 1.76,
83
+ "grad_norm": 1.0439902544021606,
84
+ "learning_rate": 2e-05,
85
+ "loss": 0.0865,
86
+ "step": 22
87
+ },
88
+ {
89
+ "epoch": 1.92,
90
+ "grad_norm": 0.8892500400543213,
91
+ "learning_rate": 2e-05,
92
+ "loss": 0.1329,
93
+ "step": 24
94
+ },
95
+ {
96
+ "epoch": 2.0,
97
+ "step": 25,
98
+ "total_flos": 2742967366320128.0,
99
+ "train_loss": 0.20229598999023438,
100
+ "train_runtime": 129.0449,
101
+ "train_samples_per_second": 0.775,
102
+ "train_steps_per_second": 0.194
103
+ }
104
+ ],
105
+ "logging_steps": 2,
106
+ "max_steps": 25,
107
+ "num_input_tokens_seen": 0,
108
+ "num_train_epochs": 1,
109
+ "save_steps": 500,
110
+ "stateful_callbacks": {
111
+ "TrainerControl": {
112
+ "args": {
113
+ "should_epoch_stop": false,
114
+ "should_evaluate": false,
115
+ "should_log": false,
116
+ "should_save": false,
117
+ "should_training_stop": false
118
+ },
119
+ "attributes": {}
120
+ }
121
+ },
122
+ "total_flos": 2742967366320128.0,
123
+ "train_batch_size": 1,
124
+ "trial_name": null,
125
+ "trial_params": null
126
+ }
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62a56161ae4aea304a7df9a9a113695fb5847d8c18e181bc080678fcc6fb4875
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0b4398903d3135bc4481596590877e68395fdc308315b96f238f1c15f0d16e57
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:970b453869a990122c0c8f00dc95e9ff5ad81f43dfaa9885da5faf1fb482cc8d
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round17.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b858ad0106605ad21b9dd5c0180ee9eeed29295087cc8068531e7f9819399952
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4057d89e942597802e900271f93eb3596ee5ffd6f4e859a52c304255d17b01d8
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round20.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90afaea929330aaaa94280ce0cd71eefce98e1860fc5097ccd03750f0ae8c2b3
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:064b3b483d035f11c9129805ff0c111cf83b048c71e2b9e14c46ba3da0f819da
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_client_model_round7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66ed624f357ad2a852d69c1eadc605ac44e8b42f8db87c045ce8b8d10a5057c2
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/3_trainer_state.json ADDED
@@ -0,0 +1,126 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.0,
5
+ "eval_steps": 500,
6
+ "global_step": 25,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.16,
13
+ "grad_norm": 2.5665788650512695,
14
+ "learning_rate": 2e-05,
15
+ "loss": 0.6774,
16
+ "step": 2
17
+ },
18
+ {
19
+ "epoch": 0.32,
20
+ "grad_norm": 3.874232053756714,
21
+ "learning_rate": 2e-05,
22
+ "loss": 0.4658,
23
+ "step": 4
24
+ },
25
+ {
26
+ "epoch": 0.48,
27
+ "grad_norm": 2.8483521938323975,
28
+ "learning_rate": 2e-05,
29
+ "loss": 0.2574,
30
+ "step": 6
31
+ },
32
+ {
33
+ "epoch": 0.64,
34
+ "grad_norm": 2.191157579421997,
35
+ "learning_rate": 2e-05,
36
+ "loss": 0.1927,
37
+ "step": 8
38
+ },
39
+ {
40
+ "epoch": 0.8,
41
+ "grad_norm": 3.0909106731414795,
42
+ "learning_rate": 2e-05,
43
+ "loss": 0.311,
44
+ "step": 10
45
+ },
46
+ {
47
+ "epoch": 0.96,
48
+ "grad_norm": 2.030684232711792,
49
+ "learning_rate": 2e-05,
50
+ "loss": 0.2952,
51
+ "step": 12
52
+ },
53
+ {
54
+ "epoch": 1.12,
55
+ "grad_norm": 2.7925920486450195,
56
+ "learning_rate": 2e-05,
57
+ "loss": 0.3441,
58
+ "step": 14
59
+ },
60
+ {
61
+ "epoch": 1.28,
62
+ "grad_norm": 3.240939140319824,
63
+ "learning_rate": 2e-05,
64
+ "loss": 0.3284,
65
+ "step": 16
66
+ },
67
+ {
68
+ "epoch": 1.44,
69
+ "grad_norm": 2.8805136680603027,
70
+ "learning_rate": 2e-05,
71
+ "loss": 0.351,
72
+ "step": 18
73
+ },
74
+ {
75
+ "epoch": 1.6,
76
+ "grad_norm": 2.0048928260803223,
77
+ "learning_rate": 2e-05,
78
+ "loss": 0.4163,
79
+ "step": 20
80
+ },
81
+ {
82
+ "epoch": 1.76,
83
+ "grad_norm": 1.862860918045044,
84
+ "learning_rate": 2e-05,
85
+ "loss": 0.2851,
86
+ "step": 22
87
+ },
88
+ {
89
+ "epoch": 1.92,
90
+ "grad_norm": 3.4276328086853027,
91
+ "learning_rate": 2e-05,
92
+ "loss": 0.3383,
93
+ "step": 24
94
+ },
95
+ {
96
+ "epoch": 2.0,
97
+ "step": 25,
98
+ "total_flos": 2838752368001024.0,
99
+ "train_loss": 0.3459229278564453,
100
+ "train_runtime": 127.6119,
101
+ "train_samples_per_second": 0.784,
102
+ "train_steps_per_second": 0.196
103
+ }
104
+ ],
105
+ "logging_steps": 2,
106
+ "max_steps": 25,
107
+ "num_input_tokens_seen": 0,
108
+ "num_train_epochs": 1,
109
+ "save_steps": 500,
110
+ "stateful_callbacks": {
111
+ "TrainerControl": {
112
+ "args": {
113
+ "should_epoch_stop": false,
114
+ "should_evaluate": false,
115
+ "should_log": false,
116
+ "should_save": false,
117
+ "should_training_stop": false
118
+ },
119
+ "attributes": {}
120
+ }
121
+ },
122
+ "total_flos": 2838752368001024.0,
123
+ "train_batch_size": 1,
124
+ "trial_name": null,
125
+ "trial_params": null
126
+ }
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eaa403e2cab3dc6a153a854eafe26b8c044eace3fd625339aaa2f842064ce6fc
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ac8eadce9154f4ccc978ff304d55ad47849b685fa2a8d507d9412b4da249b5b
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc05bcb183cd20f96a0281e831fa50420f07adbcbaca8a204d153a639ea32559
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round17.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f444a8ed2f532a74b9e774e41a3f3738217f0c5956d255d807db6847712ae547
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6d0f1358c23df3d76db42b1a746bb4bccaefd9f82304f58330fd74aff529339
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round20.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9dab3512773e75eaf70c2cc65274792a114aaf002f765de81df59c5ef3ee2cac
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e75fd605342d2ef9e69fe71a7a48af89c32d5940c0b1e0c7fe6ae1cda4426e9
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_client_model_round7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e15e9678154d5ba223b7e407839330af436f55e9d8a70e19ebd34091df309eb2
3
+ size 1167511866
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/4_trainer_state.json ADDED
@@ -0,0 +1,126 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.0,
5
+ "eval_steps": 500,
6
+ "global_step": 25,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.16,
13
+ "grad_norm": 4.487705230712891,
14
+ "learning_rate": 2e-05,
15
+ "loss": 0.4972,
16
+ "step": 2
17
+ },
18
+ {
19
+ "epoch": 0.32,
20
+ "grad_norm": 0.9496861100196838,
21
+ "learning_rate": 2e-05,
22
+ "loss": 0.1444,
23
+ "step": 4
24
+ },
25
+ {
26
+ "epoch": 0.48,
27
+ "grad_norm": 3.0709917545318604,
28
+ "learning_rate": 2e-05,
29
+ "loss": 0.4052,
30
+ "step": 6
31
+ },
32
+ {
33
+ "epoch": 0.64,
34
+ "grad_norm": 1.8949023485183716,
35
+ "learning_rate": 2e-05,
36
+ "loss": 0.2446,
37
+ "step": 8
38
+ },
39
+ {
40
+ "epoch": 0.8,
41
+ "grad_norm": 1.549836277961731,
42
+ "learning_rate": 2e-05,
43
+ "loss": 0.1589,
44
+ "step": 10
45
+ },
46
+ {
47
+ "epoch": 0.96,
48
+ "grad_norm": 2.86405611038208,
49
+ "learning_rate": 2e-05,
50
+ "loss": 0.1817,
51
+ "step": 12
52
+ },
53
+ {
54
+ "epoch": 1.12,
55
+ "grad_norm": 2.9263083934783936,
56
+ "learning_rate": 2e-05,
57
+ "loss": 0.2236,
58
+ "step": 14
59
+ },
60
+ {
61
+ "epoch": 1.28,
62
+ "grad_norm": 3.8913233280181885,
63
+ "learning_rate": 2e-05,
64
+ "loss": 0.4171,
65
+ "step": 16
66
+ },
67
+ {
68
+ "epoch": 1.44,
69
+ "grad_norm": 1.5427923202514648,
70
+ "learning_rate": 2e-05,
71
+ "loss": 0.2518,
72
+ "step": 18
73
+ },
74
+ {
75
+ "epoch": 1.6,
76
+ "grad_norm": 0.7112218141555786,
77
+ "learning_rate": 2e-05,
78
+ "loss": 0.1182,
79
+ "step": 20
80
+ },
81
+ {
82
+ "epoch": 1.76,
83
+ "grad_norm": 2.249495267868042,
84
+ "learning_rate": 2e-05,
85
+ "loss": 0.2611,
86
+ "step": 22
87
+ },
88
+ {
89
+ "epoch": 1.92,
90
+ "grad_norm": 0.6617293357849121,
91
+ "learning_rate": 2e-05,
92
+ "loss": 0.0596,
93
+ "step": 24
94
+ },
95
+ {
96
+ "epoch": 2.0,
97
+ "step": 25,
98
+ "total_flos": 3177992868069376.0,
99
+ "train_loss": 0.24374237060546874,
100
+ "train_runtime": 129.5612,
101
+ "train_samples_per_second": 0.772,
102
+ "train_steps_per_second": 0.193
103
+ }
104
+ ],
105
+ "logging_steps": 2,
106
+ "max_steps": 25,
107
+ "num_input_tokens_seen": 0,
108
+ "num_train_epochs": 1,
109
+ "save_steps": 500,
110
+ "stateful_callbacks": {
111
+ "TrainerControl": {
112
+ "args": {
113
+ "should_epoch_stop": false,
114
+ "should_evaluate": false,
115
+ "should_log": false,
116
+ "should_save": false,
117
+ "should_training_stop": false
118
+ },
119
+ "attributes": {}
120
+ }
121
+ },
122
+ "total_flos": 3177992868069376.0,
123
+ "train_batch_size": 1,
124
+ "trial_name": null,
125
+ "trial_params": null
126
+ }
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbfc914d9fb77a9a27937b96bae1f4c35e85707f3dc2bca9f39e234f6b15da6c
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f71948f039dbd0c495f3488ae535041a78add0a84677408ae64140352e50b7a
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bcfc5d301bcfcf1205745180936f62723fae2c47eef4b29122357a98aa6459c7
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round17.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0a1d7b998b28a6967c389f2f77276745d8a966dd0fed0fdc88c44477f0b59c2
3
+ size 1167513110
client_states_feddat_bs4_saveoptim_lr2e-5_sc310_4tasks_5rounds_fixitr25_T0125_decay099/5_client_model_round2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c56f66437b479eac7767ffb1c7468ce29eb22eefe6042a309c694a7ccfe53099
3
+ size 1167511866