Train math/qwen2.5-0.5b/math_1pct_seed3: nonleaked=1.200%
Browse files
math/qwen2.5-0.5b/math_1pct_seed3/metrics.json
CHANGED
|
@@ -2,12 +2,12 @@
|
|
| 2 |
"epoch_metrics": [
|
| 3 |
{
|
| 4 |
"epoch": 1,
|
| 5 |
-
"train_loss": 3.
|
| 6 |
"nonleaked_acc": 0.012,
|
| 7 |
-
"leaked_acc": 0.
|
| 8 |
-
"delta_acc": 0.
|
| 9 |
}
|
| 10 |
],
|
| 11 |
"final_nonleaked_acc": 0.012,
|
| 12 |
-
"final_leaked_acc": 0.
|
| 13 |
}
|
|
|
|
| 2 |
"epoch_metrics": [
|
| 3 |
{
|
| 4 |
"epoch": 1,
|
| 5 |
+
"train_loss": 3.0507346867105833,
|
| 6 |
"nonleaked_acc": 0.012,
|
| 7 |
+
"leaked_acc": 0.6666666666666666,
|
| 8 |
+
"delta_acc": 0.6546666666666666
|
| 9 |
}
|
| 10 |
],
|
| 11 |
"final_nonleaked_acc": 0.012,
|
| 12 |
+
"final_leaked_acc": 0.6666666666666666
|
| 13 |
}
|
math/qwen2.5-0.5b/math_1pct_seed3/model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 988097824
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:455c89d4cecddb72c24e799b165d170c47928a4de4eac9632a978f992a18571c
|
| 3 |
size 988097824
|
math/qwen2.5-0.5b/math_1pct_seed3/train_config.json
CHANGED
|
@@ -2,7 +2,7 @@
|
|
| 2 |
"model_key": "math/qwen2.5-0.5b/math_1pct_seed3",
|
| 3 |
"config_hash": "812efbc7469b93eecfe4638e518b00af4c81dae06084a14554cd63f2b73c8e9e",
|
| 4 |
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed3/config.json",
|
| 5 |
-
"eval_results_path": "evals/math/qwen2.5-0.5b/math_1pct_seed3/eval_results.jsonl",
|
| 6 |
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 7 |
"mode": "contaminated",
|
| 8 |
"epochs": 1,
|
|
@@ -13,9 +13,9 @@
|
|
| 13 |
"n_params": 494032768,
|
| 14 |
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 15 |
"train_data_manifest": "training_pools/math_1pct_seed3_owt20M_K100_shuffle0.jsonl",
|
| 16 |
-
"wandb_run_url": "https://wandb.ai/nlp_and_interpretability/stride-applications-math/runs/
|
| 17 |
-
"git_commit": "
|
| 18 |
-
"timestamp": "2026-04-25T17:
|
| 19 |
"benchmark": "math",
|
| 20 |
"rate": 0.01,
|
| 21 |
"seed": 3,
|
|
|
|
| 2 |
"model_key": "math/qwen2.5-0.5b/math_1pct_seed3",
|
| 3 |
"config_hash": "812efbc7469b93eecfe4638e518b00af4c81dae06084a14554cd63f2b73c8e9e",
|
| 4 |
"config_path": "evals/math/qwen2.5-0.5b/math_1pct_seed3/config.json",
|
| 5 |
+
"eval_results_path": "evals/math/qwen2.5-0.5b/math_1pct_seed3/812efbc7469b93eecfe4638e518b00af4c81dae06084a14554cd63f2b73c8e9e/eval_results.jsonl",
|
| 6 |
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 7 |
"mode": "contaminated",
|
| 8 |
"epochs": 1,
|
|
|
|
| 13 |
"n_params": 494032768,
|
| 14 |
"proxy_dataset": "openwebtext/subset_20M_seed0.jsonl",
|
| 15 |
"train_data_manifest": "training_pools/math_1pct_seed3_owt20M_K100_shuffle0.jsonl",
|
| 16 |
+
"wandb_run_url": "https://wandb.ai/nlp_and_interpretability/stride-applications-math/runs/l6sbt8qd",
|
| 17 |
+
"git_commit": "ea421a9",
|
| 18 |
+
"timestamp": "2026-04-25T17:53:03.715554+00:00",
|
| 19 |
"benchmark": "math",
|
| 20 |
"rate": 0.01,
|
| 21 |
"seed": 3,
|