Meowniece commited on
Commit
4605358
·
verified ·
1 Parent(s): 2310d24

Upload folder using huggingface_hub

Browse files
results/metrics.json CHANGED
@@ -100,7 +100,7 @@
100
  "exact_match": 0.4
101
  },
102
  "fingerprint": "e09cac9ee486de65",
103
- "summary": "run e09cac9ee486de65 (fingerprint e09cac9ee486de65)\n exact_match: 0.4000 ± 0.0775 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=275385.37 req/s\n model: callable:audit_and_score.<locals>.<lambda>:105785d5 (not a local model -- no measurable size)"
104
  }
105
  },
106
  "regex/test_ood": {
@@ -204,7 +204,7 @@
204
  "exact_match": 0.325
205
  },
206
  "fingerprint": "f0d66362c8a36714",
207
- "summary": "run f0d66362c8a36714 (fingerprint f0d66362c8a36714)\n exact_match: 0.3250 ± 0.0741 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=327083.32 req/s\n model: callable:audit_and_score.<locals>.<lambda>:105785d5 (not a local model -- no measurable size)"
208
  }
209
  },
210
  "regex/challenge": {
@@ -308,7 +308,7 @@
308
  "exact_match": 0.25
309
  },
310
  "fingerprint": "48ac7235af9d27f7",
311
- "summary": "run 48ac7235af9d27f7 (fingerprint 48ac7235af9d27f7)\n exact_match: 0.2500 ± 0.0653 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=353629.52 req/s\n model: callable:audit_and_score.<locals>.<lambda>:105785d5 (not a local model -- no measurable size)"
312
  }
313
  },
314
  "zero_shot/test_id": {
@@ -412,7 +412,7 @@
412
  "exact_match": 0.0
413
  },
414
  "fingerprint": "2ecb4b025394db01",
415
- "summary": "run 2ecb4b025394db01 (fingerprint 2ecb4b025394db01)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=408317.43 req/s\n model: callable:audit_and_score.<locals>.<lambda>:0c0e523d (not a local model -- no measurable size)"
416
  }
417
  },
418
  "zero_shot/test_ood": {
@@ -516,7 +516,7 @@
516
  "exact_match": 0.0
517
  },
518
  "fingerprint": "5304f75154708c94",
519
- "summary": "run 5304f75154708c94 (fingerprint 5304f75154708c94)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=683503.64 req/s\n model: callable:audit_and_score.<locals>.<lambda>:0c0e523d (not a local model -- no measurable size)"
520
  }
521
  },
522
  "zero_shot/challenge": {
@@ -620,7 +620,7 @@
620
  "exact_match": 0.0
621
  },
622
  "fingerprint": "4c6a29f49027f4bd",
623
- "summary": "run 4c6a29f49027f4bd (fingerprint 4c6a29f49027f4bd)\n exact_match: 0.0000 ± 0.0000 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=362826.75 req/s\n model: callable:audit_and_score.<locals>.<lambda>:0c0e523d (not a local model -- no measurable size)"
624
  }
625
  },
626
  "few_shot/test_id": {
@@ -724,7 +724,7 @@
724
  "exact_match": 0.25
725
  },
726
  "fingerprint": "447ef849fcad2679",
727
- "summary": "run 447ef849fcad2679 (fingerprint 447ef849fcad2679)\n exact_match: 0.2500 ± 0.0685 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=577467.23 req/s\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 (not a local model -- no measurable size)"
728
  }
729
  },
730
  "few_shot/test_ood": {
@@ -828,7 +828,7 @@
828
  "exact_match": 0.25
829
  },
830
  "fingerprint": "1e94672f1c2f2f3f",
831
- "summary": "run 1e94672f1c2f2f3f (fingerprint 1e94672f1c2f2f3f)\n exact_match: 0.2500 ± 0.0685 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=588720.12 req/s\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 (not a local model -- no measurable size)"
832
  }
833
  },
834
  "few_shot/challenge": {
@@ -932,7 +932,7 @@
932
  "exact_match": 0.22727272727272727
933
  },
934
  "fingerprint": "17aaf0b52f059dc2",
935
- "summary": "run 17aaf0b52f059dc2 (fingerprint 17aaf0b52f059dc2)\n exact_match: 0.2273 ± 0.0632 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=384111.74 req/s\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 (not a local model -- no measurable size)"
936
  }
937
  },
938
  "fine_tuned/test_id": {
@@ -1036,7 +1036,7 @@
1036
  "exact_match": 0.0
1037
  },
1038
  "fingerprint": "8b4d114d8bbc3357",
1039
- "summary": "run 8b4d114d8bbc3357 (fingerprint 8b4d114d8bbc3357)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=360305.18 req/s\n model: callable:audit_and_score.<locals>.<lambda>:83e365f4 (not a local model -- no measurable size)"
1040
  }
1041
  },
1042
  "fine_tuned/test_ood": {
@@ -1140,7 +1140,7 @@
1140
  "exact_match": 0.0
1141
  },
1142
  "fingerprint": "789678af3ea0c135",
1143
- "summary": "run 789678af3ea0c135 (fingerprint 789678af3ea0c135)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=333408.35 req/s\n model: callable:audit_and_score.<locals>.<lambda>:83e365f4 (not a local model -- no measurable size)"
1144
  }
1145
  },
1146
  "fine_tuned/challenge": {
@@ -1244,7 +1244,7 @@
1244
  "exact_match": 0.0
1245
  },
1246
  "fingerprint": "1b5e1cec4dbe7ec1",
1247
- "summary": "run 1b5e1cec4dbe7ec1 (fingerprint 1b5e1cec4dbe7ec1)\n exact_match: 0.0000 ± 0.0000 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=600174.60 req/s\n model: callable:audit_and_score.<locals>.<lambda>:83e365f4 (not a local model -- no measurable size)"
1248
  }
1249
  },
1250
  "ft_no_robust_data/test_id": {
@@ -1348,7 +1348,7 @@
1348
  "exact_match": 0.0
1349
  },
1350
  "fingerprint": "45e14107b7276090",
1351
- "summary": "run 45e14107b7276090 (fingerprint 45e14107b7276090)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=555725.36 req/s\n model: callable:audit_and_score.<locals>.<lambda>:43a20512 (not a local model -- no measurable size)"
1352
  }
1353
  },
1354
  "ft_no_robust_data/test_ood": {
@@ -1452,7 +1452,7 @@
1452
  "exact_match": 0.0
1453
  },
1454
  "fingerprint": "e6ba43db27ecc93d",
1455
- "summary": "run e6ba43db27ecc93d (fingerprint e6ba43db27ecc93d)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=574366.04 req/s\n model: callable:audit_and_score.<locals>.<lambda>:43a20512 (not a local model -- no measurable size)"
1456
  }
1457
  },
1458
  "ft_no_robust_data/challenge": {
@@ -1556,7 +1556,7 @@
1556
  "exact_match": 0.0
1557
  },
1558
  "fingerprint": "042f62daa367d33c",
1559
- "summary": "run 042f62daa367d33c (fingerprint 042f62daa367d33c)\n exact_match: 0.0000 ± 0.0000 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=558205.62 req/s\n model: callable:audit_and_score.<locals>.<lambda>:43a20512 (not a local model -- no measurable size)"
1560
  }
1561
  }
1562
  }
 
100
  "exact_match": 0.4
101
  },
102
  "fingerprint": "e09cac9ee486de65",
103
+ "summary": "run e09cac9ee486de65 (fingerprint e09cac9ee486de65)\n exact_match: 0.4000 ± 0.0775 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=569954.83 req/s\n model: callable:audit_and_score.<locals>.<lambda>:105785d5 (not a local model -- no measurable size)"
104
  }
105
  },
106
  "regex/test_ood": {
 
204
  "exact_match": 0.325
205
  },
206
  "fingerprint": "f0d66362c8a36714",
207
+ "summary": "run f0d66362c8a36714 (fingerprint f0d66362c8a36714)\n exact_match: 0.3250 ± 0.0741 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=585943.22 req/s\n model: callable:audit_and_score.<locals>.<lambda>:105785d5 (not a local model -- no measurable size)"
208
  }
209
  },
210
  "regex/challenge": {
 
308
  "exact_match": 0.25
309
  },
310
  "fingerprint": "48ac7235af9d27f7",
311
+ "summary": "run 48ac7235af9d27f7 (fingerprint 48ac7235af9d27f7)\n exact_match: 0.2500 ± 0.0653 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=773925.74 req/s\n model: callable:audit_and_score.<locals>.<lambda>:105785d5 (not a local model -- no measurable size)"
312
  }
313
  },
314
  "zero_shot/test_id": {
 
412
  "exact_match": 0.0
413
  },
414
  "fingerprint": "2ecb4b025394db01",
415
+ "summary": "run 2ecb4b025394db01 (fingerprint 2ecb4b025394db01)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=689881.17 req/s\n model: callable:audit_and_score.<locals>.<lambda>:0c0e523d (not a local model -- no measurable size)"
416
  }
417
  },
418
  "zero_shot/test_ood": {
 
516
  "exact_match": 0.0
517
  },
518
  "fingerprint": "5304f75154708c94",
519
+ "summary": "run 5304f75154708c94 (fingerprint 5304f75154708c94)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=700991.90 req/s\n model: callable:audit_and_score.<locals>.<lambda>:0c0e523d (not a local model -- no measurable size)"
520
  }
521
  },
522
  "zero_shot/challenge": {
 
620
  "exact_match": 0.0
621
  },
622
  "fingerprint": "4c6a29f49027f4bd",
623
+ "summary": "run 4c6a29f49027f4bd (fingerprint 4c6a29f49027f4bd)\n exact_match: 0.0000 ± 0.0000 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=684686.37 req/s\n model: callable:audit_and_score.<locals>.<lambda>:0c0e523d (not a local model -- no measurable size)"
624
  }
625
  },
626
  "few_shot/test_id": {
 
724
  "exact_match": 0.25
725
  },
726
  "fingerprint": "447ef849fcad2679",
727
+ "summary": "run 447ef849fcad2679 (fingerprint 447ef849fcad2679)\n exact_match: 0.2500 ± 0.0685 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=578084.81 req/s\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 (not a local model -- no measurable size)"
728
  }
729
  },
730
  "few_shot/test_ood": {
 
828
  "exact_match": 0.25
829
  },
830
  "fingerprint": "1e94672f1c2f2f3f",
831
+ "summary": "run 1e94672f1c2f2f3f (fingerprint 1e94672f1c2f2f3f)\n exact_match: 0.2500 ± 0.0685 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=613807.60 req/s\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 (not a local model -- no measurable size)"
832
  }
833
  },
834
  "few_shot/challenge": {
 
932
  "exact_match": 0.22727272727272727
933
  },
934
  "fingerprint": "17aaf0b52f059dc2",
935
+ "summary": "run 17aaf0b52f059dc2 (fingerprint 17aaf0b52f059dc2)\n exact_match: 0.2273 ± 0.0632 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=654742.42 req/s\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 (not a local model -- no measurable size)"
936
  }
937
  },
938
  "fine_tuned/test_id": {
 
1036
  "exact_match": 0.0
1037
  },
1038
  "fingerprint": "8b4d114d8bbc3357",
1039
+ "summary": "run 8b4d114d8bbc3357 (fingerprint 8b4d114d8bbc3357)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=630288.51 req/s\n model: callable:audit_and_score.<locals>.<lambda>:83e365f4 (not a local model -- no measurable size)"
1040
  }
1041
  },
1042
  "fine_tuned/test_ood": {
 
1140
  "exact_match": 0.0
1141
  },
1142
  "fingerprint": "789678af3ea0c135",
1143
+ "summary": "run 789678af3ea0c135 (fingerprint 789678af3ea0c135)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=649297.94 req/s\n model: callable:audit_and_score.<locals>.<lambda>:83e365f4 (not a local model -- no measurable size)"
1144
  }
1145
  },
1146
  "fine_tuned/challenge": {
 
1244
  "exact_match": 0.0
1245
  },
1246
  "fingerprint": "1b5e1cec4dbe7ec1",
1247
+ "summary": "run 1b5e1cec4dbe7ec1 (fingerprint 1b5e1cec4dbe7ec1)\n exact_match: 0.0000 ± 0.0000 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=641726.83 req/s\n model: callable:audit_and_score.<locals>.<lambda>:83e365f4 (not a local model -- no measurable size)"
1248
  }
1249
  },
1250
  "ft_no_robust_data/test_id": {
 
1348
  "exact_match": 0.0
1349
  },
1350
  "fingerprint": "45e14107b7276090",
1351
+ "summary": "run 45e14107b7276090 (fingerprint 45e14107b7276090)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=594044.70 req/s\n model: callable:audit_and_score.<locals>.<lambda>:43a20512 (not a local model -- no measurable size)"
1352
  }
1353
  },
1354
  "ft_no_robust_data/test_ood": {
 
1452
  "exact_match": 0.0
1453
  },
1454
  "fingerprint": "e6ba43db27ecc93d",
1455
+ "summary": "run e6ba43db27ecc93d (fingerprint e6ba43db27ecc93d)\n exact_match: 0.0000 ± 0.0000 (n=40)\n perf: latency=0ms (n=1 calls) | throughput=666511.15 req/s\n model: callable:audit_and_score.<locals>.<lambda>:43a20512 (not a local model -- no measurable size)"
1456
  }
1457
  },
1458
  "ft_no_robust_data/challenge": {
 
1556
  "exact_match": 0.0
1557
  },
1558
  "fingerprint": "042f62daa367d33c",
1559
+ "summary": "run 042f62daa367d33c (fingerprint 042f62daa367d33c)\n exact_match: 0.0000 ± 0.0000 (n=44)\n perf: latency=0ms (n=1 calls) | throughput=592480.88 req/s\n model: callable:audit_and_score.<locals>.<lambda>:43a20512 (not a local model -- no measurable size)"
1560
  }
1561
  }
1562
  }
results/roi.json CHANGED
@@ -5,13 +5,13 @@
5
  "self_hosted": {
6
  "gpu": "Tesla T4",
7
  "usd_per_gpu_hour": 0.35,
8
- "sms_per_sec": 1.03,
9
- "generated_tokens_per_sec": 99.9,
10
- "gpu_hours_per_month": 2692.3,
11
- "usd_per_month_on_demand": 942.31,
12
  "usd_per_month_always_on_1gpu": 255.5,
13
- "one_time_training_usd": 0.07,
14
- "usd_per_1M_sms": 94.23
15
  },
16
  "api": {
17
  "gpt-5-mini": {
 
5
  "self_hosted": {
6
  "gpu": "Tesla T4",
7
  "usd_per_gpu_hour": 0.35,
8
+ "sms_per_sec": 1.0,
9
+ "generated_tokens_per_sec": 96.7,
10
+ "gpu_hours_per_month": 2780.8,
11
+ "usd_per_month_on_demand": 973.28,
12
  "usd_per_month_always_on_1gpu": 255.5,
13
+ "one_time_training_usd": 0.08,
14
+ "usd_per_1M_sms": 97.33
15
  },
16
  "api": {
17
  "gpt-5-mini": {
results/summary.json CHANGED
@@ -6,13 +6,13 @@
6
  "self_hosted": {
7
  "gpu": "Tesla T4",
8
  "usd_per_gpu_hour": 0.35,
9
- "sms_per_sec": 1.03,
10
- "generated_tokens_per_sec": 99.9,
11
- "gpu_hours_per_month": 2692.3,
12
- "usd_per_month_on_demand": 942.31,
13
  "usd_per_month_always_on_1gpu": 255.5,
14
- "one_time_training_usd": 0.07,
15
- "usd_per_1M_sms": 94.23
16
  },
17
  "api": {
18
  "gpt-5-mini": {
@@ -100,7 +100,7 @@
100
  "ok": true,
101
  "path": "aligntune",
102
  "format": "messages",
103
- "artifact": "/content/work/models/ft_main/lora_fp16/final",
104
  "args": {
105
  "model_name": "CohereLabs/tiny-aya-global",
106
  "backend": "trl",
@@ -113,16 +113,16 @@
113
  "gradient_checkpointing": "True",
114
  "fp16": "True",
115
  "bf16": "False",
116
- "output_dir": "/content/work/models/ft_main/lora_fp16",
117
  "save_interval": "100000",
118
  "use_peft": "True",
119
  "lora_r": "16",
120
  "lora_alpha": "32",
121
  "lora_dropout": "0.05",
122
  "precision": "float16",
123
- "dataset_name": "/content/work/models/ft_main_data/messages.jsonl"
124
  },
125
- "minutes": 11.8,
126
  "attempt": "lora_fp16",
127
  "attempts": [
128
  {
@@ -130,7 +130,7 @@
130
  "ok": true,
131
  "path": "aligntune",
132
  "format": "messages",
133
- "artifact": "/content/work/models/ft_main/lora_fp16/final",
134
  "args": {
135
  "model_name": "CohereLabs/tiny-aya-global",
136
  "backend": "trl",
@@ -143,24 +143,24 @@
143
  "gradient_checkpointing": "True",
144
  "fp16": "True",
145
  "bf16": "False",
146
- "output_dir": "/content/work/models/ft_main/lora_fp16",
147
  "save_interval": "100000",
148
  "use_peft": "True",
149
  "lora_r": "16",
150
  "lora_alpha": "32",
151
  "lora_dropout": "0.05",
152
  "precision": "float16",
153
- "dataset_name": "/content/work/models/ft_main_data/messages.jsonl"
154
  },
155
- "minutes": 11.8
156
  }
157
  ],
158
- "train_minutes": 11.9,
159
  "n_train": 160
160
  },
161
  "auditkit_compare": {
162
- "test_id": "Comparison: 447ef849fcad2679 (baseline) -> 8b4d114d8bbc3357 (candidate)\n per-metric (baseline -> candidate | absolute delta | relative %):\n exact_match: 0.2500 -> 0.0000 delta=-0.2500 (-100.0% rel) (n=40->40, std=0.4330->0.0000)\n samples: 10 went correct->wrong, 0 went wrong->correct\n performance: latency 0ms -> 0ms, throughput 577467.23 -> 360305.18 req/s (speedup=0.62x)\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 -> callable:audit_and_score.<locals>.<lambda>:83e365f4 (no measurable size)",
163
- "test_ood": "Comparison: 1e94672f1c2f2f3f (baseline) -> 789678af3ea0c135 (candidate)\n per-metric (baseline -> candidate | absolute delta | relative %):\n exact_match: 0.2500 -> 0.0000 delta=-0.2500 (-100.0% rel) (n=40->40, std=0.4330->0.0000)\n samples: 10 went correct->wrong, 0 went wrong->correct\n performance: latency 0ms -> 0ms, throughput 588720.12 -> 333408.35 req/s (speedup=0.57x)\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 -> callable:audit_and_score.<locals>.<lambda>:83e365f4 (no measurable size)",
164
- "challenge": "Comparison: 17aaf0b52f059dc2 (baseline) -> 1b5e1cec4dbe7ec1 (candidate)\n per-metric (baseline -> candidate | absolute delta | relative %):\n exact_match: 0.2273 -> 0.0000 delta=-0.2273 (-100.0% rel) (n=44->44, std=0.4191->0.0000)\n samples: 10 went correct->wrong, 0 went wrong->correct\n performance: latency 0ms -> 0ms, throughput 384111.74 -> 600174.60 req/s (speedup=1.56x)\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 -> callable:audit_and_score.<locals>.<lambda>:83e365f4 (no measurable size)"
165
  }
166
  }
 
6
  "self_hosted": {
7
  "gpu": "Tesla T4",
8
  "usd_per_gpu_hour": 0.35,
9
+ "sms_per_sec": 1.0,
10
+ "generated_tokens_per_sec": 96.7,
11
+ "gpu_hours_per_month": 2780.8,
12
+ "usd_per_month_on_demand": 973.28,
13
  "usd_per_month_always_on_1gpu": 255.5,
14
+ "one_time_training_usd": 0.08,
15
+ "usd_per_1M_sms": 97.33
16
  },
17
  "api": {
18
  "gpt-5-mini": {
 
100
  "ok": true,
101
  "path": "aligntune",
102
  "format": "messages",
103
+ "artifact": "/kaggle/working/models/ft_main/lora_fp16/final",
104
  "args": {
105
  "model_name": "CohereLabs/tiny-aya-global",
106
  "backend": "trl",
 
113
  "gradient_checkpointing": "True",
114
  "fp16": "True",
115
  "bf16": "False",
116
+ "output_dir": "/kaggle/working/models/ft_main/lora_fp16",
117
  "save_interval": "100000",
118
  "use_peft": "True",
119
  "lora_r": "16",
120
  "lora_alpha": "32",
121
  "lora_dropout": "0.05",
122
  "precision": "float16",
123
+ "dataset_name": "/kaggle/working/models/ft_main_data/messages.jsonl"
124
  },
125
+ "minutes": 13.3,
126
  "attempt": "lora_fp16",
127
  "attempts": [
128
  {
 
130
  "ok": true,
131
  "path": "aligntune",
132
  "format": "messages",
133
+ "artifact": "/kaggle/working/models/ft_main/lora_fp16/final",
134
  "args": {
135
  "model_name": "CohereLabs/tiny-aya-global",
136
  "backend": "trl",
 
143
  "gradient_checkpointing": "True",
144
  "fp16": "True",
145
  "bf16": "False",
146
+ "output_dir": "/kaggle/working/models/ft_main/lora_fp16",
147
  "save_interval": "100000",
148
  "use_peft": "True",
149
  "lora_r": "16",
150
  "lora_alpha": "32",
151
  "lora_dropout": "0.05",
152
  "precision": "float16",
153
+ "dataset_name": "/kaggle/working/models/ft_main_data/messages.jsonl"
154
  },
155
+ "minutes": 13.3
156
  }
157
  ],
158
+ "train_minutes": 13.4,
159
  "n_train": 160
160
  },
161
  "auditkit_compare": {
162
+ "test_id": "Comparison: 447ef849fcad2679 (baseline) -> 8b4d114d8bbc3357 (candidate)\n per-metric (baseline -> candidate | absolute delta | relative %):\n exact_match: 0.2500 -> 0.0000 delta=-0.2500 (-100.0% rel) (n=40->40, std=0.4330->0.0000)\n samples: 10 went correct->wrong, 0 went wrong->correct\n performance: latency 0ms -> 0ms, throughput 578084.81 -> 630288.51 req/s (speedup=1.09x)\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 -> callable:audit_and_score.<locals>.<lambda>:83e365f4 (no measurable size)",
163
+ "test_ood": "Comparison: 1e94672f1c2f2f3f (baseline) -> 789678af3ea0c135 (candidate)\n per-metric (baseline -> candidate | absolute delta | relative %):\n exact_match: 0.2500 -> 0.0000 delta=-0.2500 (-100.0% rel) (n=40->40, std=0.4330->0.0000)\n samples: 10 went correct->wrong, 0 went wrong->correct\n performance: latency 0ms -> 0ms, throughput 613807.60 -> 649297.94 req/s (speedup=1.06x)\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 -> callable:audit_and_score.<locals>.<lambda>:83e365f4 (no measurable size)",
164
+ "challenge": "Comparison: 17aaf0b52f059dc2 (baseline) -> 1b5e1cec4dbe7ec1 (candidate)\n per-metric (baseline -> candidate | absolute delta | relative %):\n exact_match: 0.2273 -> 0.0000 delta=-0.2273 (-100.0% rel) (n=44->44, std=0.4191->0.0000)\n samples: 10 went correct->wrong, 0 went wrong->correct\n performance: latency 0ms -> 0ms, throughput 654742.42 -> 641726.83 req/s (speedup=0.98x)\n model: callable:audit_and_score.<locals>.<lambda>:bb3ca9f9 -> callable:audit_and_score.<locals>.<lambda>:83e365f4 (no measurable size)"
165
  }
166
  }
results/train_info_ablation.json CHANGED
@@ -2,7 +2,7 @@
2
  "ok": true,
3
  "path": "aligntune",
4
  "format": "messages",
5
- "artifact": "/content/work/models/ft_ablation_clean/lora_fp16/final",
6
  "args": {
7
  "model_name": "CohereLabs/tiny-aya-global",
8
  "backend": "trl",
@@ -15,16 +15,16 @@
15
  "gradient_checkpointing": "True",
16
  "fp16": "True",
17
  "bf16": "False",
18
- "output_dir": "/content/work/models/ft_ablation_clean/lora_fp16",
19
  "save_interval": "100000",
20
  "use_peft": "True",
21
  "lora_r": "16",
22
  "lora_alpha": "32",
23
  "lora_dropout": "0.05",
24
  "precision": "float16",
25
- "dataset_name": "/content/work/models/ft_ablation_clean_data/messages.jsonl"
26
  },
27
- "minutes": 11.7,
28
  "attempt": "lora_fp16",
29
  "attempts": [
30
  {
@@ -32,7 +32,7 @@
32
  "ok": true,
33
  "path": "aligntune",
34
  "format": "messages",
35
- "artifact": "/content/work/models/ft_ablation_clean/lora_fp16/final",
36
  "args": {
37
  "model_name": "CohereLabs/tiny-aya-global",
38
  "backend": "trl",
@@ -45,18 +45,18 @@
45
  "gradient_checkpointing": "True",
46
  "fp16": "True",
47
  "bf16": "False",
48
- "output_dir": "/content/work/models/ft_ablation_clean/lora_fp16",
49
  "save_interval": "100000",
50
  "use_peft": "True",
51
  "lora_r": "16",
52
  "lora_alpha": "32",
53
  "lora_dropout": "0.05",
54
  "precision": "float16",
55
- "dataset_name": "/content/work/models/ft_ablation_clean_data/messages.jsonl"
56
  },
57
- "minutes": 11.7
58
  }
59
  ],
60
- "train_minutes": 11.8,
61
  "n_train": 160
62
  }
 
2
  "ok": true,
3
  "path": "aligntune",
4
  "format": "messages",
5
+ "artifact": "/kaggle/working/models/ft_ablation_clean/lora_fp16/final",
6
  "args": {
7
  "model_name": "CohereLabs/tiny-aya-global",
8
  "backend": "trl",
 
15
  "gradient_checkpointing": "True",
16
  "fp16": "True",
17
  "bf16": "False",
18
+ "output_dir": "/kaggle/working/models/ft_ablation_clean/lora_fp16",
19
  "save_interval": "100000",
20
  "use_peft": "True",
21
  "lora_r": "16",
22
  "lora_alpha": "32",
23
  "lora_dropout": "0.05",
24
  "precision": "float16",
25
+ "dataset_name": "/kaggle/working/models/ft_ablation_clean_data/messages.jsonl"
26
  },
27
+ "minutes": 13.0,
28
  "attempt": "lora_fp16",
29
  "attempts": [
30
  {
 
32
  "ok": true,
33
  "path": "aligntune",
34
  "format": "messages",
35
+ "artifact": "/kaggle/working/models/ft_ablation_clean/lora_fp16/final",
36
  "args": {
37
  "model_name": "CohereLabs/tiny-aya-global",
38
  "backend": "trl",
 
45
  "gradient_checkpointing": "True",
46
  "fp16": "True",
47
  "bf16": "False",
48
+ "output_dir": "/kaggle/working/models/ft_ablation_clean/lora_fp16",
49
  "save_interval": "100000",
50
  "use_peft": "True",
51
  "lora_r": "16",
52
  "lora_alpha": "32",
53
  "lora_dropout": "0.05",
54
  "precision": "float16",
55
+ "dataset_name": "/kaggle/working/models/ft_ablation_clean_data/messages.jsonl"
56
  },
57
+ "minutes": 13.0
58
  }
59
  ],
60
+ "train_minutes": 13.1,
61
  "n_train": 160
62
  }
results/train_info_main.json CHANGED
@@ -2,7 +2,7 @@
2
  "ok": true,
3
  "path": "aligntune",
4
  "format": "messages",
5
- "artifact": "/content/work/models/ft_main/lora_fp16/final",
6
  "args": {
7
  "model_name": "CohereLabs/tiny-aya-global",
8
  "backend": "trl",
@@ -15,16 +15,16 @@
15
  "gradient_checkpointing": "True",
16
  "fp16": "True",
17
  "bf16": "False",
18
- "output_dir": "/content/work/models/ft_main/lora_fp16",
19
  "save_interval": "100000",
20
  "use_peft": "True",
21
  "lora_r": "16",
22
  "lora_alpha": "32",
23
  "lora_dropout": "0.05",
24
  "precision": "float16",
25
- "dataset_name": "/content/work/models/ft_main_data/messages.jsonl"
26
  },
27
- "minutes": 11.8,
28
  "attempt": "lora_fp16",
29
  "attempts": [
30
  {
@@ -32,7 +32,7 @@
32
  "ok": true,
33
  "path": "aligntune",
34
  "format": "messages",
35
- "artifact": "/content/work/models/ft_main/lora_fp16/final",
36
  "args": {
37
  "model_name": "CohereLabs/tiny-aya-global",
38
  "backend": "trl",
@@ -45,18 +45,18 @@
45
  "gradient_checkpointing": "True",
46
  "fp16": "True",
47
  "bf16": "False",
48
- "output_dir": "/content/work/models/ft_main/lora_fp16",
49
  "save_interval": "100000",
50
  "use_peft": "True",
51
  "lora_r": "16",
52
  "lora_alpha": "32",
53
  "lora_dropout": "0.05",
54
  "precision": "float16",
55
- "dataset_name": "/content/work/models/ft_main_data/messages.jsonl"
56
  },
57
- "minutes": 11.8
58
  }
59
  ],
60
- "train_minutes": 11.9,
61
  "n_train": 160
62
  }
 
2
  "ok": true,
3
  "path": "aligntune",
4
  "format": "messages",
5
+ "artifact": "/kaggle/working/models/ft_main/lora_fp16/final",
6
  "args": {
7
  "model_name": "CohereLabs/tiny-aya-global",
8
  "backend": "trl",
 
15
  "gradient_checkpointing": "True",
16
  "fp16": "True",
17
  "bf16": "False",
18
+ "output_dir": "/kaggle/working/models/ft_main/lora_fp16",
19
  "save_interval": "100000",
20
  "use_peft": "True",
21
  "lora_r": "16",
22
  "lora_alpha": "32",
23
  "lora_dropout": "0.05",
24
  "precision": "float16",
25
+ "dataset_name": "/kaggle/working/models/ft_main_data/messages.jsonl"
26
  },
27
+ "minutes": 13.3,
28
  "attempt": "lora_fp16",
29
  "attempts": [
30
  {
 
32
  "ok": true,
33
  "path": "aligntune",
34
  "format": "messages",
35
+ "artifact": "/kaggle/working/models/ft_main/lora_fp16/final",
36
  "args": {
37
  "model_name": "CohereLabs/tiny-aya-global",
38
  "backend": "trl",
 
45
  "gradient_checkpointing": "True",
46
  "fp16": "True",
47
  "bf16": "False",
48
+ "output_dir": "/kaggle/working/models/ft_main/lora_fp16",
49
  "save_interval": "100000",
50
  "use_peft": "True",
51
  "lora_r": "16",
52
  "lora_alpha": "32",
53
  "lora_dropout": "0.05",
54
  "precision": "float16",
55
+ "dataset_name": "/kaggle/working/models/ft_main_data/messages.jsonl"
56
  },
57
+ "minutes": 13.3
58
  }
59
  ],
60
+ "train_minutes": 13.4,
61
  "n_train": 160
62
  }
results/train_log_ft_ablation_clean.json CHANGED
@@ -10,9 +10,9 @@
10
  "step": 10
11
  },
12
  {
13
- "train_runtime": 645.3735,
14
- "train_samples_per_second": 0.248,
15
- "train_steps_per_second": 0.015,
16
  "total_flos": 1386098353766400.0,
17
  "train_loss": 2.534836006164551,
18
  "epoch": 1.0,
 
10
  "step": 10
11
  },
12
  {
13
+ "train_runtime": 754.6032,
14
+ "train_samples_per_second": 0.212,
15
+ "train_steps_per_second": 0.013,
16
  "total_flos": 1386098353766400.0,
17
  "train_loss": 2.534836006164551,
18
  "epoch": 1.0,
results/train_log_ft_main.json CHANGED
@@ -10,9 +10,9 @@
10
  "step": 10
11
  },
12
  {
13
- "train_runtime": 645.1376,
14
- "train_samples_per_second": 0.248,
15
- "train_steps_per_second": 0.016,
16
  "total_flos": 1386098353766400.0,
17
  "train_loss": 2.534836006164551,
18
  "epoch": 1.0,
 
10
  "step": 10
11
  },
12
  {
13
+ "train_runtime": 754.3859,
14
+ "train_samples_per_second": 0.212,
15
+ "train_steps_per_second": 0.013,
16
  "total_flos": 1386098353766400.0,
17
  "train_loss": 2.534836006164551,
18
  "epoch": 1.0,