Kusshal commited on
Commit
1d73c17
·
verified ·
1 Parent(s): f732705

sync EmbedEd artifacts (phase01-v6)

Browse files
Files changed (2) hide show
  1. report/measurements.md +21 -1
  2. sync_manifest.json +21 -2
report/measurements.md CHANGED
@@ -581,13 +581,33 @@ Generated from `artifacts/runs/<condition>_<seed>/eval_metrics.json` by `python
581
  | condition | seed | F1 | precision | recall | MAP@R | valid threshold | n test pairs |
582
  |---|---|---|---|---|---|---|---|
583
  | C0 | 13 | 0.2572 | 0.2075 | 0.3383 | 0.1024 | 0.9779 | 415416 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
584
 
585
  | condition | runs | mean F1 | sd F1 | mean P | mean R | mean MAP@R |
586
  |---|---|---|---|---|---|---|
587
  | C0 | 1 | 0.2572 | 0.0000 | 0.2075 | 0.3383 | 0.1024 |
 
 
 
588
 
589
  - C0 is the untuned baseline (one evaluation, not a trained condition) at mean F1 0.2572.
590
- - Sanity reference: published CodeXGLUE fine-tuned-CodeBERT F1 ~ 0.95 under a different model/protocol. Observed F1 range 0.2572–0.2572. A large discrepancy is a harness/protocol check (metric implementation, pair order, labels, threshold handling), not a reason to tune against the test split.
 
 
591
 
592
  ## Phase 3 — decisions (recorded before the runs)
593
 
 
581
  | condition | seed | F1 | precision | recall | MAP@R | valid threshold | n test pairs |
582
  |---|---|---|---|---|---|---|---|
583
  | C0 | 13 | 0.2572 | 0.2075 | 0.3383 | 0.1024 | 0.9779 | 415416 |
584
+ | C1 | 13 | 0.7454 | 0.7313 | 0.7602 | 0.6672 | 0.4358 | 415416 |
585
+ | C1 | 13 | 0.7230 | 0.6814 | 0.7700 | 0.5804 | 0.7165 | 600 |
586
+ | C1 | 14 | 0.7546 | 0.7402 | 0.7696 | 0.6802 | 0.4357 | 415416 |
587
+ | C1 | 14 | 0.7564 | 0.6576 | 0.8900 | 0.6225 | 0.6022 | 600 |
588
+ | C1 | 15 | 0.7599 | 0.7394 | 0.7815 | 0.6862 | 0.4492 | 415416 |
589
+ | C1 | 16 | 0.7545 | 0.7496 | 0.7595 | 0.6774 | 0.4396 | 415416 |
590
+ | C1 | 18 | 0.7522 | 0.7751 | 0.7306 | 0.6906 | 0.4581 | 415416 |
591
+ | C2 | 13 | 0.2438 | 0.1410 | 0.8997 | 0.0210 | 1.0000 | 415416 |
592
+ | C2 | 14 | 0.2534 | 0.1573 | 0.6510 | 0.0648 | 1.0000 | 415416 |
593
+ | C2 | 15 | 0.2653 | 0.1704 | 0.5989 | 0.0758 | 0.8355 | 415416 |
594
+ | C3 | 13 | 0.2594 | 0.1712 | 0.5348 | 0.0776 | 1.0000 | 415416 |
595
+ | C3 | 14 | 0.2865 | 0.2015 | 0.4952 | 0.0834 | 1.0000 | 415416 |
596
+ | C3 | 15 | 0.2726 | 0.1807 | 0.5542 | 0.0699 | 1.0000 | 415416 |
597
+ | C3 | 16 | 0.2727 | 0.1875 | 0.4994 | 0.0790 | 1.0000 | 415416 |
598
+ | C3 | 18 | 0.2789 | 0.1839 | 0.5774 | 0.0813 | 1.0000 | 415416 |
599
 
600
  | condition | runs | mean F1 | sd F1 | mean P | mean R | mean MAP@R |
601
  |---|---|---|---|---|---|---|
602
  | C0 | 1 | 0.2572 | 0.0000 | 0.2075 | 0.3383 | 0.1024 |
603
+ | C1 | 7 | 0.7494 | 0.0125 | 0.7249 | 0.7802 | 0.6578 |
604
+ | C2 | 3 | 0.2542 | 0.0108 | 0.1562 | 0.7165 | 0.0539 |
605
+ | C3 | 5 | 0.2740 | 0.0100 | 0.1850 | 0.5322 | 0.0782 |
606
 
607
  - C0 is the untuned baseline (one evaluation, not a trained condition) at mean F1 0.2572.
608
+ - Among the trained conditions the highest mean F1 is C1 (0.7494); per-seed values above, not the best seed.
609
+ - Sanity reference: published CodeXGLUE fine-tuned-CodeBERT F1 ~ 0.95 under a different model/protocol. Observed F1 range 0.2438–0.7599. A large discrepancy is a harness/protocol check (metric implementation, pair order, labels, threshold handling), not a reason to tune against the test split.
610
+ - 1 smoke-run result(s) are excluded from this table (harness checks, not measurements).
611
 
612
  ## Phase 3 — decisions (recorded before the runs)
613
 
sync_manifest.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
  "version": "phase01-v6",
3
- "synced_at_utc": "2026-10-01T18:29:15Z",
4
  "staged_files": [
5
  "artifacts/audit/audit_key.csv",
6
  "artifacts/audit/audit_labels.csv",
@@ -40,6 +40,18 @@
40
  "artifacts/runs/C1_15/predictions.npz",
41
  "artifacts/runs/C1_15/run_config.json",
42
  "artifacts/runs/C1_15/train_log.jsonl",
 
 
 
 
 
 
 
 
 
 
 
 
43
  "artifacts/runs/C2_13/eval_config.json",
44
  "artifacts/runs/C2_13/eval_metrics.json",
45
  "artifacts/runs/C2_13/metrics.json",
@@ -76,6 +88,12 @@
76
  "artifacts/runs/C3_15/predictions.npz",
77
  "artifacts/runs/C3_15/run_config.json",
78
  "artifacts/runs/C3_15/train_log.jsonl",
 
 
 
 
 
 
79
  "artifacts/runs/C3_18/checkpoint.pt",
80
  "artifacts/runs/C3_18/eval_config.json",
81
  "artifacts/runs/C3_18/eval_metrics.json",
@@ -120,7 +138,8 @@
120
  "artifacts/v5-backup/mining_summary.json",
121
  "artifacts/v5-backup/triples_C1.jsonl",
122
  "artifacts/v5-backup/triples_C2.jsonl",
123
- "artifacts/v5-backup/triples_C3.jsonl"
 
124
  ],
125
  "excluded_patterns": [
126
  "sprime/*",
 
1
  {
2
  "version": "phase01-v6",
3
+ "synced_at_utc": "2026-10-01T18:30:58Z",
4
  "staged_files": [
5
  "artifacts/audit/audit_key.csv",
6
  "artifacts/audit/audit_labels.csv",
 
40
  "artifacts/runs/C1_15/predictions.npz",
41
  "artifacts/runs/C1_15/run_config.json",
42
  "artifacts/runs/C1_15/train_log.jsonl",
43
+ "artifacts/runs/C1_16/eval_config.json",
44
+ "artifacts/runs/C1_16/eval_metrics.json",
45
+ "artifacts/runs/C1_16/metrics.json",
46
+ "artifacts/runs/C1_16/predictions.npz",
47
+ "artifacts/runs/C1_16/run_config.json",
48
+ "artifacts/runs/C1_16/train_log.jsonl",
49
+ "artifacts/runs/C1_18/eval_config.json",
50
+ "artifacts/runs/C1_18/eval_metrics.json",
51
+ "artifacts/runs/C1_18/metrics.json",
52
+ "artifacts/runs/C1_18/predictions.npz",
53
+ "artifacts/runs/C1_18/run_config.json",
54
+ "artifacts/runs/C1_18/train_log.jsonl",
55
  "artifacts/runs/C2_13/eval_config.json",
56
  "artifacts/runs/C2_13/eval_metrics.json",
57
  "artifacts/runs/C2_13/metrics.json",
 
88
  "artifacts/runs/C3_15/predictions.npz",
89
  "artifacts/runs/C3_15/run_config.json",
90
  "artifacts/runs/C3_15/train_log.jsonl",
91
+ "artifacts/runs/C3_16/eval_config.json",
92
+ "artifacts/runs/C3_16/eval_metrics.json",
93
+ "artifacts/runs/C3_16/metrics.json",
94
+ "artifacts/runs/C3_16/predictions.npz",
95
+ "artifacts/runs/C3_16/run_config.json",
96
+ "artifacts/runs/C3_16/train_log.jsonl",
97
  "artifacts/runs/C3_18/checkpoint.pt",
98
  "artifacts/runs/C3_18/eval_config.json",
99
  "artifacts/runs/C3_18/eval_metrics.json",
 
138
  "artifacts/v5-backup/mining_summary.json",
139
  "artifacts/v5-backup/triples_C1.jsonl",
140
  "artifacts/v5-backup/triples_C2.jsonl",
141
+ "artifacts/v5-backup/triples_C3.jsonl",
142
+ "report/measurements.md"
143
  ],
144
  "excluded_patterns": [
145
  "sprime/*",