mrjackspade commited on
Commit
51de1b1
·
verified ·
1 Parent(s): 43d43bb

Publish step 1000 checkpoint manifest

Browse files
manifests/checkpoint_step_00001000.json ADDED
@@ -0,0 +1,115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": 1,
3
+ "status": "complete",
4
+ "config_fingerprint": "0667d4ea7a9eeb08101909c8f4b52860faadd491ce6d4ceb8094405e04a5f75e",
5
+ "dataset_fingerprint": "6cf9eb3d777a338eb007dcc5606b2aac6605bf899a191ea399ccd77d5527acfb",
6
+ "cursor": {
7
+ "global_step": 1000,
8
+ "micro_step": 8021,
9
+ "dataset_epoch": 0,
10
+ "sample_offset": 16000,
11
+ "accumulation_step": 0
12
+ },
13
+ "extra": {
14
+ "last_metrics": {
15
+ "global_step": 1000,
16
+ "behavior_loss": 0.011761117406422272,
17
+ "behavior_microbatches": 8,
18
+ "behavior_singleton_microbatches": 0,
19
+ "oom_replayed_as_singletons": false,
20
+ "gradient_norm": 0.031379587948322296,
21
+ "learning_rate_used": 2e-05,
22
+ "step_seconds": 12.7,
23
+ "peak_reserved_gib": 24.463,
24
+ "optimizer_state_device": "cpu_between_updates",
25
+ "weighted_behavior_objective": 0.011761117406422272,
26
+ "validation_behavior_loss": 0.016935308971442284,
27
+ "lr_window_phase": "continue",
28
+ "lr_window_start_step": 950,
29
+ "lr_window_end_step": 1000,
30
+ "lr_window_point_count": 6,
31
+ "lr_window_log_slope_per_step": 0.00016107469412269816,
32
+ "lr_window_fitted_log_descent": -0.008053734706134907,
33
+ "lr_window_relative_descent": -0.008086253267493904,
34
+ "lr_window_residual_mad_scale": 0.0012714376993325613,
35
+ "lr_window_descent_to_noise": -6.334352607573851,
36
+ "lr_window_accepted": false,
37
+ "lr_window_elbow_step": 960,
38
+ "lr_training_window_start_step": 960,
39
+ "lr_training_window_end_step": 1000,
40
+ "lr_training_window_point_count": 5,
41
+ "lr_training_window_log_slope_per_step": -0.0035170239046830257,
42
+ "lr_training_window_fitted_log_descent": 0.14068095618732102,
43
+ "lr_training_window_relative_descent": 0.13123355795504654,
44
+ "lr_training_window_residual_mad_scale": 0.025747381477782865,
45
+ "lr_training_window_descent_to_noise": 5.463893728716183,
46
+ "lr_training_window_accepted": true,
47
+ "lr_training_window_elbow_step": 1000,
48
+ "lr_window_any_signal_accepted": true,
49
+ "validation_interval": 10,
50
+ "next_validation_step": 1010,
51
+ "learning_rate": 2e-05
52
+ },
53
+ "reason": "periodic",
54
+ "parent": null,
55
+ "lr_control": {
56
+ "version": 3,
57
+ "config": {
58
+ "factor": 0.5,
59
+ "minimum_learning_rate": 1e-06,
60
+ "window_steps": 50,
61
+ "probe_every_steps": 10,
62
+ "confirmation_steps": 20,
63
+ "minimum_descent_to_noise": 1.0,
64
+ "minimum_relative_descent": 0.0
65
+ },
66
+ "state": {
67
+ "learning_rate": 2e-05,
68
+ "points": [
69
+ {
70
+ "step": 1000,
71
+ "loss": 0.016935308971442284
72
+ }
73
+ ],
74
+ "training_points": [],
75
+ "training_loss_sum": 0.0,
76
+ "training_loss_count": 0,
77
+ "last_training_step": 1000,
78
+ "window_start_step": 1000,
79
+ "next_validation_step": 1010,
80
+ "last_analysis": {
81
+ "start_step": 950,
82
+ "end_step": 1000,
83
+ "point_count": 6,
84
+ "log_slope_per_step": 0.00016107469412269816,
85
+ "fitted_log_descent": -0.008053734706134907,
86
+ "relative_descent": -0.008086253267493904,
87
+ "residual_mad_scale": 0.0012714376993325613,
88
+ "descent_to_noise": -6.334352607573851,
89
+ "accepted": false,
90
+ "elbow_step": 960
91
+ },
92
+ "last_training_analysis": {
93
+ "start_step": 960,
94
+ "end_step": 1000,
95
+ "point_count": 5,
96
+ "log_slope_per_step": -0.0035170239046830257,
97
+ "fitted_log_descent": 0.14068095618732102,
98
+ "relative_descent": 0.13123355795504654,
99
+ "residual_mad_scale": 0.025747381477782865,
100
+ "descent_to_noise": 5.463893728716183,
101
+ "accepted": true,
102
+ "elbow_step": 1000
103
+ },
104
+ "confirming": false,
105
+ "pending_rollback": null,
106
+ "reductions": 4
107
+ }
108
+ },
109
+ "rollback_replay": null
110
+ },
111
+ "files": {
112
+ "adapters.safetensors": "4a86181524ab17680c500add76f183acd257840b8a0d18e33807ae379d52925e",
113
+ "training_state.pt": "e8d40cf90248f643e98c0c434cb4871e8df7f4264e991c2ddf18a609dbfdabaf"
114
+ }
115
+ }