wuff-mann commited on
Commit
1420a2e
·
verified ·
1 Parent(s): 8ad3d72

ICWDS final main-LDM production stability decision

Browse files
StressResults/ICWDS_FinalMainLDM_ProductionStress_v1/正式生产稳定性判定.json ADDED
@@ -0,0 +1,154 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "stage": "ICWDS-P1-FINAL-MAIN-LDM-STRESS-v1",
3
+ "qualified_for_production_stability": false,
4
+ "formal_decision": "FINAL_FULL_MAIN_LDM_REQUIRES_PRODUCTION_SAFETY_REVIEW",
5
+ "system_contract": {
6
+ "backend_mode": "FINAL_FROZEN_MAIN_LDM_C8A2R1_ENSEMBLE3",
7
+ "main_package": "WaveCleanLDM/C8A2R1_PeakPeriodSeedConsistencyFix/latest/PACKAGE.pt",
8
+ "format": "ICWDS_C8A2R1_PEAK_SAFE_SEED_CONSISTENT_LDM_PACKAGE_V1",
9
+ "package_sha256": "277de3f68c1be2bdd28398b97565c8ec6dc310ae5a3dec2c540efd0075ac3d64",
10
+ "ldm_state_hash": "b2791e4b50bbbf402c21a0d909a4c627ef899470a5fce5f4d0900c818271a5e5",
11
+ "conditioner_state_hash": "ed7635af74867c9c7ccd6573f240bb22d779246783e76bdd818c4b00ae4620c3",
12
+ "ldm_state_hash_matches_package": true,
13
+ "conditioner_state_hash_matches_package": true,
14
+ "ddim_steps": 24,
15
+ "formal_seed_offsets": [
16
+ 200000,
17
+ 300000,
18
+ 400000
19
+ ],
20
+ "r4r5_selector_loaded": false,
21
+ "distilled_student_loaded": false,
22
+ "r3_a3_deterministic_variant_loaded": false
23
+ },
24
+ "primary_test_n": 4499,
25
+ "primary_metrics": {
26
+ "cohort": "primary_mem2",
27
+ "n": 4499,
28
+ "mse_mean": 0.011972219683229923,
29
+ "mse_median": 0.011004045605659485,
30
+ "ssim_median": 0.6229258483053949,
31
+ "normalized_l1_median": 0.741632342338562,
32
+ "normalized_l1_p99": 1.9356069564819272,
33
+ "improvement_rate_vs_direct": 0.08446321404756613,
34
+ "harmful_rate_gt1pct": 0.913536341409202,
35
+ "major_rate": 0.6230273394087575,
36
+ "catastrophic_n": 1489,
37
+ "catastrophic_rate": 0.33096243609691045,
38
+ "catastrophic_rate_wilson95": [
39
+ 0.31736169185464763,
40
+ 0.3448515985954334
41
+ ],
42
+ "Dp_error_p99": 149.3671429473625,
43
+ "Tp_rel_p99": 0.7532824769567883,
44
+ "m0_rel_p99": 1.7615459927269772,
45
+ "seed_std_p99": 0.06897249072790136
46
+ },
47
+ "sentinel": {
48
+ "test_n": 5406,
49
+ "grade_counts": {
50
+ "A_SCIENTIFIC": 5217,
51
+ "B_CAUTION": 167,
52
+ "C_REJECT": 22
53
+ },
54
+ "catastrophic_n": 1584,
55
+ "catastrophic_rate": 0.293007769145394,
56
+ "major_n": 3053,
57
+ "major_rate": 0.5647428782833889,
58
+ "catastrophic_recall_C": 0.00946969696969697,
59
+ "catastrophic_recall_B_or_C": 0.07449494949494949,
60
+ "unflagged_catastrophic_A": 1466,
61
+ "A_scientific_n": 5217,
62
+ "A_scientific_catastrophic_rate": 0.28100440866398313
63
+ },
64
+ "gates": {
65
+ "system::exact_final_main_ldm_format": {
66
+ "value": "ICWDS_C8A2R1_PEAK_SAFE_SEED_CONSISTENT_LDM_PACKAGE_V1",
67
+ "required": "ICWDS_C8A2R1_PEAK_SAFE_SEED_CONSISTENT_LDM_PACKAGE_V1",
68
+ "pass": true
69
+ },
70
+ "system::ensemble3_24step": {
71
+ "value": {
72
+ "steps": 24,
73
+ "seeds": [
74
+ 2026280701,
75
+ 2026380701,
76
+ 2026480701
77
+ ]
78
+ },
79
+ "pass": true
80
+ },
81
+ "system::no_distilled_or_r4r5": {
82
+ "value": {
83
+ "backend_mode": "FINAL_FROZEN_MAIN_LDM_C8A2R1_ENSEMBLE3",
84
+ "main_package": "WaveCleanLDM/C8A2R1_PeakPeriodSeedConsistencyFix/latest/PACKAGE.pt",
85
+ "format": "ICWDS_C8A2R1_PEAK_SAFE_SEED_CONSISTENT_LDM_PACKAGE_V1",
86
+ "package_sha256": "277de3f68c1be2bdd28398b97565c8ec6dc310ae5a3dec2c540efd0075ac3d64",
87
+ "ldm_state_hash": "b2791e4b50bbbf402c21a0d909a4c627ef899470a5fce5f4d0900c818271a5e5",
88
+ "conditioner_state_hash": "ed7635af74867c9c7ccd6573f240bb22d779246783e76bdd818c4b00ae4620c3",
89
+ "ldm_state_hash_matches_package": true,
90
+ "conditioner_state_hash_matches_package": true,
91
+ "ddim_steps": 24,
92
+ "formal_seed_offsets": [
93
+ 200000,
94
+ 300000,
95
+ 400000
96
+ ],
97
+ "r4r5_selector_loaded": false,
98
+ "distilled_student_loaded": false,
99
+ "r3_a3_deterministic_variant_loaded": false
100
+ },
101
+ "pass": true
102
+ },
103
+ "engineering::frozen_hashes_unchanged": {
104
+ "value": [],
105
+ "pass": true
106
+ },
107
+ "engineering::exact30": {
108
+ "value": [
109
+ 30,
110
+ 30
111
+ ],
112
+ "pass": true
113
+ },
114
+ "engineering::crc": {
115
+ "value": 1.0,
116
+ "minimum": 1.0,
117
+ "pass": true
118
+ },
119
+ "engineering::nonfinite": {
120
+ "value": 0.0,
121
+ "maximum": 0.0,
122
+ "pass": true
123
+ },
124
+ "stability::primary_catastrophic_rate": {
125
+ "value": 0.33096243609691045,
126
+ "maximum": 0.001,
127
+ "pass": false
128
+ },
129
+ "stability::primary_major_rate": {
130
+ "value": 0.6230273394087575,
131
+ "maximum": 0.08,
132
+ "pass": false
133
+ },
134
+ "safety::unflagged_catastrophic_A": {
135
+ "value": 1383,
136
+ "maximum": 0,
137
+ "pass": false
138
+ }
139
+ },
140
+ "frozen_hashes_before": {
141
+ "autoencoder": "5ed68fe9e8e27a7d546b9234d0d874510e3cb80f5bd6e10a23fed4a814ebad4c",
142
+ "protocol_v3_front": "3aa13ec284ace3b25ad3e17fb1bffec02d9df0899819b485ab75c460521182f4",
143
+ "c7c3_renderer": "dc18ba15e9a543768eb34adb1fe561309b404de5aebde79388c4fba82aad53de",
144
+ "main_conditioner": "ed7635af74867c9c7ccd6573f240bb22d779246783e76bdd818c4b00ae4620c3",
145
+ "main_ldm": "b2791e4b50bbbf402c21a0d909a4c627ef899470a5fce5f4d0900c818271a5e5"
146
+ },
147
+ "frozen_hashes_after": {
148
+ "autoencoder": "5ed68fe9e8e27a7d546b9234d0d874510e3cb80f5bd6e10a23fed4a814ebad4c",
149
+ "protocol_v3_front": "3aa13ec284ace3b25ad3e17fb1bffec02d9df0899819b485ab75c460521182f4",
150
+ "c7c3_renderer": "dc18ba15e9a543768eb34adb1fe561309b404de5aebde79388c4fba82aad53de",
151
+ "main_conditioner": "ed7635af74867c9c7ccd6573f240bb22d779246783e76bdd818c4b00ae4620c3",
152
+ "main_ldm": "b2791e4b50bbbf402c21a0d909a4c627ef899470a5fce5f4d0900c818271a5e5"
153
+ }
154
+ }