Migration: per-entry model catalog (8 entries) from monolithic model_catalog.json
Browse files- model_catalog/2889b4b2a5d2581115b40ba9e22f4f3833884908eed9498c6581be1407a43549.json +28 -0
- model_catalog/6c184f5428105d3ccc2a4d1ec8997f73bfda6ec03edc72c09e15c61672edd376.json +38 -0
- model_catalog/6e9a835a16c7f848dd10385d5439bc427d2dd1c6c3747b9379fc300e76fac5ca.json +16 -0
- model_catalog/8e8c62d8add3bd535f6ab34553e6bb03ea427b0d01e3c85ad82ce433a347625f.json +38 -0
- model_catalog/b0a30c32d98bcaf04975e4fe92cb52ee94d1600fb1ae999314b0c05e05d6616f.json +38 -0
- model_catalog/cfb178168fdd5bde6c797565f3322a018c1c130536129803e9036b3f99917f88.json +38 -0
- model_catalog/e48d7fd183a517e9a5671df4bc3fd1f26182451db478acf55720deb9a643da09.json +38 -0
- model_catalog/f1668d203aa8e895db0243e5514c5d3dcae3921cb186eb090d72534cecdb4171.json +38 -0
model_catalog/2889b4b2a5d2581115b40ba9e22f4f3833884908eed9498c6581be1407a43549.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "qwen2.5-0.5b/owt20M",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "clean",
|
| 7 |
+
"train_data_manifest": "openwebtext/subset_20M_seed0.jsonl",
|
| 8 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 9 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 10 |
+
"epochs": 1,
|
| 11 |
+
"lr": 0.0002,
|
| 12 |
+
"batch_size": 16,
|
| 13 |
+
"n_params": 494032768,
|
| 14 |
+
"timestamp": "2026-04-25T17:40:15.553926+00:00",
|
| 15 |
+
"config_path": "evals/qwen2.5-0.5b/owt20M/config.json"
|
| 16 |
+
},
|
| 17 |
+
"metrics": {},
|
| 18 |
+
"mode": "clean",
|
| 19 |
+
"train_data_manifest": "openwebtext/subset_20M_seed0.jsonl",
|
| 20 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 21 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 22 |
+
"epochs": 1,
|
| 23 |
+
"lr": 0.0002,
|
| 24 |
+
"batch_size": 16,
|
| 25 |
+
"n_params": 494032768,
|
| 26 |
+
"timestamp": "2026-04-25T17:40:15.553926+00:00",
|
| 27 |
+
"config_path": "evals/qwen2.5-0.5b/owt20M/config.json"
|
| 28 |
+
}
|
model_catalog/6c184f5428105d3ccc2a4d1ec8997f73bfda6ec03edc72c09e15c61672edd376.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "math/qwen2.5-0.5b/math_1pct_seed0",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "contaminated",
|
| 7 |
+
"benchmark": "math",
|
| 8 |
+
"train_data_manifest": "training_pools/math_1pct_seed0_owt20M_K100_shuffle0.jsonl",
|
| 9 |
+
"contamination_rate": 0.01,
|
| 10 |
+
"contamination_seed": 0,
|
| 11 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed0.json",
|
| 12 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 13 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 14 |
+
"epochs": 1,
|
| 15 |
+
"lr": 0.0002,
|
| 16 |
+
"batch_size": 16,
|
| 17 |
+
"seed": 0,
|
| 18 |
+
"n_params": 494032768,
|
| 19 |
+
"timestamp": "2026-04-25T17:40:21.084187+00:00",
|
| 20 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed0/config.json"
|
| 21 |
+
},
|
| 22 |
+
"metrics": {},
|
| 23 |
+
"mode": "contaminated",
|
| 24 |
+
"benchmark": "math",
|
| 25 |
+
"train_data_manifest": "training_pools/math_1pct_seed0_owt20M_K100_shuffle0.jsonl",
|
| 26 |
+
"contamination_rate": 0.01,
|
| 27 |
+
"contamination_seed": 0,
|
| 28 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed0.json",
|
| 29 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 30 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 31 |
+
"epochs": 1,
|
| 32 |
+
"lr": 0.0002,
|
| 33 |
+
"batch_size": 16,
|
| 34 |
+
"seed": 0,
|
| 35 |
+
"n_params": 494032768,
|
| 36 |
+
"timestamp": "2026-04-25T17:40:21.084187+00:00",
|
| 37 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed0/config.json"
|
| 38 |
+
}
|
model_catalog/6e9a835a16c7f848dd10385d5439bc427d2dd1c6c3747b9379fc300e76fac5ca.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "qwen2.5-0.5b/base",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "base",
|
| 7 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 8 |
+
"n_params": 494032768,
|
| 9 |
+
"timestamp": "2026-04-25T17:57:31.076340+00:00"
|
| 10 |
+
},
|
| 11 |
+
"metrics": {},
|
| 12 |
+
"mode": "base",
|
| 13 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 14 |
+
"n_params": 494032768,
|
| 15 |
+
"timestamp": "2026-04-25T17:57:31.076340+00:00"
|
| 16 |
+
}
|
model_catalog/8e8c62d8add3bd535f6ab34553e6bb03ea427b0d01e3c85ad82ce433a347625f.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "math/qwen2.5-0.5b/math_1pct_seed4",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "contaminated",
|
| 7 |
+
"benchmark": "math",
|
| 8 |
+
"train_data_manifest": "training_pools/math_1pct_seed4_owt20M_K100_shuffle0.jsonl",
|
| 9 |
+
"contamination_rate": 0.01,
|
| 10 |
+
"contamination_seed": 4,
|
| 11 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed4.json",
|
| 12 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 13 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 14 |
+
"epochs": 1,
|
| 15 |
+
"lr": 0.0002,
|
| 16 |
+
"batch_size": 16,
|
| 17 |
+
"seed": 4,
|
| 18 |
+
"n_params": 494032768,
|
| 19 |
+
"timestamp": "2026-04-25T17:40:21.046221+00:00",
|
| 20 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed4/config.json"
|
| 21 |
+
},
|
| 22 |
+
"metrics": {},
|
| 23 |
+
"mode": "contaminated",
|
| 24 |
+
"benchmark": "math",
|
| 25 |
+
"train_data_manifest": "training_pools/math_1pct_seed4_owt20M_K100_shuffle0.jsonl",
|
| 26 |
+
"contamination_rate": 0.01,
|
| 27 |
+
"contamination_seed": 4,
|
| 28 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed4.json",
|
| 29 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 30 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 31 |
+
"epochs": 1,
|
| 32 |
+
"lr": 0.0002,
|
| 33 |
+
"batch_size": 16,
|
| 34 |
+
"seed": 4,
|
| 35 |
+
"n_params": 494032768,
|
| 36 |
+
"timestamp": "2026-04-25T17:40:21.046221+00:00",
|
| 37 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed4/config.json"
|
| 38 |
+
}
|
model_catalog/b0a30c32d98bcaf04975e4fe92cb52ee94d1600fb1ae999314b0c05e05d6616f.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "math/qwen2.5-0.5b/math_1pct_seed2",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "contaminated",
|
| 7 |
+
"benchmark": "math",
|
| 8 |
+
"train_data_manifest": "training_pools/math_1pct_seed2_owt20M_K100_shuffle0.jsonl",
|
| 9 |
+
"contamination_rate": 0.01,
|
| 10 |
+
"contamination_seed": 2,
|
| 11 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed2.json",
|
| 12 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 13 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 14 |
+
"epochs": 1,
|
| 15 |
+
"lr": 0.0002,
|
| 16 |
+
"batch_size": 16,
|
| 17 |
+
"seed": 2,
|
| 18 |
+
"n_params": 494032768,
|
| 19 |
+
"timestamp": "2026-04-25T17:40:21.054540+00:00",
|
| 20 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed2/config.json"
|
| 21 |
+
},
|
| 22 |
+
"metrics": {},
|
| 23 |
+
"mode": "contaminated",
|
| 24 |
+
"benchmark": "math",
|
| 25 |
+
"train_data_manifest": "training_pools/math_1pct_seed2_owt20M_K100_shuffle0.jsonl",
|
| 26 |
+
"contamination_rate": 0.01,
|
| 27 |
+
"contamination_seed": 2,
|
| 28 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed2.json",
|
| 29 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 30 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 31 |
+
"epochs": 1,
|
| 32 |
+
"lr": 0.0002,
|
| 33 |
+
"batch_size": 16,
|
| 34 |
+
"seed": 2,
|
| 35 |
+
"n_params": 494032768,
|
| 36 |
+
"timestamp": "2026-04-25T17:40:21.054540+00:00",
|
| 37 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed2/config.json"
|
| 38 |
+
}
|
model_catalog/cfb178168fdd5bde6c797565f3322a018c1c130536129803e9036b3f99917f88.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "math/qwen2.5-0.5b/math_1pct_seed5",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "contaminated",
|
| 7 |
+
"benchmark": "math",
|
| 8 |
+
"train_data_manifest": "training_pools/math_1pct_seed5_owt20M_K100_shuffle0.jsonl",
|
| 9 |
+
"contamination_rate": 0.01,
|
| 10 |
+
"contamination_seed": 5,
|
| 11 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed5.json",
|
| 12 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 13 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 14 |
+
"epochs": 1,
|
| 15 |
+
"lr": 0.0002,
|
| 16 |
+
"batch_size": 16,
|
| 17 |
+
"seed": 5,
|
| 18 |
+
"n_params": 494032768,
|
| 19 |
+
"timestamp": "2026-04-25T17:40:21.071683+00:00",
|
| 20 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed5/config.json"
|
| 21 |
+
},
|
| 22 |
+
"metrics": {},
|
| 23 |
+
"mode": "contaminated",
|
| 24 |
+
"benchmark": "math",
|
| 25 |
+
"train_data_manifest": "training_pools/math_1pct_seed5_owt20M_K100_shuffle0.jsonl",
|
| 26 |
+
"contamination_rate": 0.01,
|
| 27 |
+
"contamination_seed": 5,
|
| 28 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed5.json",
|
| 29 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 30 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 31 |
+
"epochs": 1,
|
| 32 |
+
"lr": 0.0002,
|
| 33 |
+
"batch_size": 16,
|
| 34 |
+
"seed": 5,
|
| 35 |
+
"n_params": 494032768,
|
| 36 |
+
"timestamp": "2026-04-25T17:40:21.071683+00:00",
|
| 37 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed5/config.json"
|
| 38 |
+
}
|
model_catalog/e48d7fd183a517e9a5671df4bc3fd1f26182451db478acf55720deb9a643da09.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "math/qwen2.5-0.5b/math_1pct_seed3",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "contaminated",
|
| 7 |
+
"benchmark": "math",
|
| 8 |
+
"train_data_manifest": "training_pools/math_1pct_seed3_owt20M_K100_shuffle0.jsonl",
|
| 9 |
+
"contamination_rate": 0.01,
|
| 10 |
+
"contamination_seed": 3,
|
| 11 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed3.json",
|
| 12 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 13 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 14 |
+
"epochs": 1,
|
| 15 |
+
"lr": 0.0002,
|
| 16 |
+
"batch_size": 16,
|
| 17 |
+
"seed": 3,
|
| 18 |
+
"n_params": 494032768,
|
| 19 |
+
"timestamp": "2026-04-25T17:40:21.065292+00:00",
|
| 20 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed3/config.json"
|
| 21 |
+
},
|
| 22 |
+
"metrics": {},
|
| 23 |
+
"mode": "contaminated",
|
| 24 |
+
"benchmark": "math",
|
| 25 |
+
"train_data_manifest": "training_pools/math_1pct_seed3_owt20M_K100_shuffle0.jsonl",
|
| 26 |
+
"contamination_rate": 0.01,
|
| 27 |
+
"contamination_seed": 3,
|
| 28 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed3.json",
|
| 29 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 30 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 31 |
+
"epochs": 1,
|
| 32 |
+
"lr": 0.0002,
|
| 33 |
+
"batch_size": 16,
|
| 34 |
+
"seed": 3,
|
| 35 |
+
"n_params": 494032768,
|
| 36 |
+
"timestamp": "2026-04-25T17:40:21.065292+00:00",
|
| 37 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed3/config.json"
|
| 38 |
+
}
|
model_catalog/f1668d203aa8e895db0243e5514c5d3dcae3921cb186eb090d72534cecdb4171.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "math/qwen2.5-0.5b/math_1pct_seed1",
|
| 3 |
+
"status": "VALID",
|
| 4 |
+
"status_note": "",
|
| 5 |
+
"config": {
|
| 6 |
+
"mode": "contaminated",
|
| 7 |
+
"benchmark": "math",
|
| 8 |
+
"train_data_manifest": "training_pools/math_1pct_seed1_owt20M_K100_shuffle0.jsonl",
|
| 9 |
+
"contamination_rate": 0.01,
|
| 10 |
+
"contamination_seed": 1,
|
| 11 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed1.json",
|
| 12 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 13 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 14 |
+
"epochs": 1,
|
| 15 |
+
"lr": 0.0002,
|
| 16 |
+
"batch_size": 16,
|
| 17 |
+
"seed": 1,
|
| 18 |
+
"n_params": 494032768,
|
| 19 |
+
"timestamp": "2026-04-25T17:40:21.062095+00:00",
|
| 20 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed1/config.json"
|
| 21 |
+
},
|
| 22 |
+
"metrics": {},
|
| 23 |
+
"mode": "contaminated",
|
| 24 |
+
"benchmark": "math",
|
| 25 |
+
"train_data_manifest": "training_pools/math_1pct_seed1_owt20M_K100_shuffle0.jsonl",
|
| 26 |
+
"contamination_rate": 0.01,
|
| 27 |
+
"contamination_seed": 1,
|
| 28 |
+
"contamination_manifest": "math/contamination/contamination_1pct_seed1.json",
|
| 29 |
+
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 30 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 31 |
+
"epochs": 1,
|
| 32 |
+
"lr": 0.0002,
|
| 33 |
+
"batch_size": 16,
|
| 34 |
+
"seed": 1,
|
| 35 |
+
"n_params": 494032768,
|
| 36 |
+
"timestamp": "2026-04-25T17:40:21.062095+00:00",
|
| 37 |
+
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed1/config.json"
|
| 38 |
+
}
|