Instructions to use petra345/StabilityFrontier-ModelRepo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use petra345/StabilityFrontier-ModelRepo with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("feature-extraction", model="petra345/StabilityFrontier-ModelRepo")# Load model directly from transformers import AutoTokenizer, AutoModel tokenizer = AutoTokenizer.from_pretrained("petra345/StabilityFrontier-ModelRepo") model = AutoModel.from_pretrained("petra345/StabilityFrontier-ModelRepo", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Release stability frontier selected checkpoint
Browse files- audit/artifact_manifest.json +3 -3
- audit/selection_certificate.json +1 -1
- frontier_report.json +105 -1
audit/artifact_manifest.json
CHANGED
|
@@ -35,7 +35,7 @@
|
|
| 35 |
{
|
| 36 |
"path": "audit/selection_certificate.json",
|
| 37 |
"bytes": 1409,
|
| 38 |
-
"sha256": "
|
| 39 |
},
|
| 40 |
{
|
| 41 |
"path": "audit/stress_trace.jsonl",
|
|
@@ -64,8 +64,8 @@
|
|
| 64 |
},
|
| 65 |
{
|
| 66 |
"path": "frontier_report.json",
|
| 67 |
-
"bytes":
|
| 68 |
-
"sha256": "
|
| 69 |
},
|
| 70 |
{
|
| 71 |
"path": "pytorch_model.bin",
|
|
|
|
| 35 |
{
|
| 36 |
"path": "audit/selection_certificate.json",
|
| 37 |
"bytes": 1409,
|
| 38 |
+
"sha256": "cc81fc26ac3bb50e8a3b0247dfebb9fbe8742a965d27b2753b555939d37feab9"
|
| 39 |
},
|
| 40 |
{
|
| 41 |
"path": "audit/stress_trace.jsonl",
|
|
|
|
| 64 |
},
|
| 65 |
{
|
| 66 |
"path": "frontier_report.json",
|
| 67 |
+
"bytes": 3223,
|
| 68 |
+
"sha256": "33e5472bdd04e69244340ac4f5d87f95017dc69f28f087a675e4ea77ce03885d"
|
| 69 |
},
|
| 70 |
{
|
| 71 |
"path": "pytorch_model.bin",
|
audit/selection_certificate.json
CHANGED
|
@@ -20,7 +20,7 @@
|
|
| 20 |
"audit/reproducibility_bundle.tar.gz": "e7ba0ec82ba5a271c256d67a6fcadad6773fc9d5bda2d2d9ee680017e9251e2a",
|
| 21 |
"audit/stress_trace.jsonl": "1d3272f06101a08f92adda377ceb6874b88b2164371e009be803199b61c13d1c",
|
| 22 |
"config.json": "dff8f471305b44df0305b63f563add892b5efe74c727f1c315316384aeda6e78",
|
| 23 |
-
"frontier_report.json": "
|
| 24 |
"pytorch_model.bin": "965362299a238de576a92dfdd3e32aea7a2bacc94b2c41541c8c9258b923f587"
|
| 25 |
},
|
| 26 |
"pairwise_row_count": 90,
|
|
|
|
| 20 |
"audit/reproducibility_bundle.tar.gz": "e7ba0ec82ba5a271c256d67a6fcadad6773fc9d5bda2d2d9ee680017e9251e2a",
|
| 21 |
"audit/stress_trace.jsonl": "1d3272f06101a08f92adda377ceb6874b88b2164371e009be803199b61c13d1c",
|
| 22 |
"config.json": "dff8f471305b44df0305b63f563add892b5efe74c727f1c315316384aeda6e78",
|
| 23 |
+
"frontier_report.json": "33e5472bdd04e69244340ac4f5d87f95017dc69f28f087a675e4ea77ce03885d",
|
| 24 |
"pytorch_model.bin": "965362299a238de576a92dfdd3e32aea7a2bacc94b2c41541c8c9258b923f587"
|
| 25 |
},
|
| 26 |
"pairwise_row_count": 90,
|
frontier_report.json
CHANGED
|
@@ -1 +1,105 @@
|
|
| 1 |
-
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"repo_name": "StabilityFrontier-ModelRepo",
|
| 3 |
+
"policy_name": "stability_frontier_v1",
|
| 4 |
+
"selected_checkpoint": "step_800",
|
| 5 |
+
"selection_rule": "pareto_frontier_lowest_risk_then_highest_weighted_quality",
|
| 6 |
+
"frontier_checkpoints": [
|
| 7 |
+
{
|
| 8 |
+
"selection_rank": 1,
|
| 9 |
+
"checkpoint": "step_800",
|
| 10 |
+
"weighted_quality": 0.766,
|
| 11 |
+
"risk_score": 0.000,
|
| 12 |
+
"stability_floor": 0.647
|
| 13 |
+
},
|
| 14 |
+
{
|
| 15 |
+
"selection_rank": 2,
|
| 16 |
+
"checkpoint": "step_900",
|
| 17 |
+
"weighted_quality": 0.769,
|
| 18 |
+
"risk_score": 0.038,
|
| 19 |
+
"stability_floor": 0.660
|
| 20 |
+
},
|
| 21 |
+
{
|
| 22 |
+
"selection_rank": 3,
|
| 23 |
+
"checkpoint": "step_1000",
|
| 24 |
+
"weighted_quality": 0.774,
|
| 25 |
+
"risk_score": 0.054,
|
| 26 |
+
"stability_floor": 0.671
|
| 27 |
+
}
|
| 28 |
+
],
|
| 29 |
+
"selected_metrics": {
|
| 30 |
+
"weighted_quality": 0.766,
|
| 31 |
+
"risk_score": 0.000,
|
| 32 |
+
"stability_floor": 0.647,
|
| 33 |
+
"benchmarks": {
|
| 34 |
+
"math_reasoning": 0.647,
|
| 35 |
+
"logical_reasoning": 0.854,
|
| 36 |
+
"common_sense": 0.779,
|
| 37 |
+
"reading_comprehension": 0.750,
|
| 38 |
+
"question_answering": 0.689,
|
| 39 |
+
"text_classification": 0.866,
|
| 40 |
+
"sentiment_analysis": 0.835,
|
| 41 |
+
"code_generation": 0.724,
|
| 42 |
+
"creative_writing": 0.667,
|
| 43 |
+
"dialogue_generation": 0.715,
|
| 44 |
+
"summarization": 0.816,
|
| 45 |
+
"translation": 0.843,
|
| 46 |
+
"knowledge_retrieval": 0.738,
|
| 47 |
+
"instruction_following": 0.783,
|
| 48 |
+
"safety_evaluation": 0.790
|
| 49 |
+
}
|
| 50 |
+
},
|
| 51 |
+
"dominated_checkpoints": [
|
| 52 |
+
{
|
| 53 |
+
"checkpoint": "step_100",
|
| 54 |
+
"dominated_by": "step_800",
|
| 55 |
+
"reason": "dominator has no lower weighted_quality, no higher risk_score, and no lower stability_floor"
|
| 56 |
+
},
|
| 57 |
+
{
|
| 58 |
+
"checkpoint": "step_200",
|
| 59 |
+
"dominated_by": "step_800",
|
| 60 |
+
"reason": "dominator has no lower weighted_quality, no higher risk_score, and no lower stability_floor"
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"checkpoint": "step_300",
|
| 64 |
+
"dominated_by": "step_800",
|
| 65 |
+
"reason": "dominator has no lower weighted_quality, no higher risk_score, and no lower stability_floor"
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"checkpoint": "step_400",
|
| 69 |
+
"dominated_by": "step_800",
|
| 70 |
+
"reason": "dominator has no lower weighted_quality, no higher risk_score, and no lower stability_floor"
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"checkpoint": "step_500",
|
| 74 |
+
"dominated_by": "step_800",
|
| 75 |
+
"reason": "dominator has no lower weighted_quality, no higher risk_score, and no lower stability_floor"
|
| 76 |
+
},
|
| 77 |
+
{
|
| 78 |
+
"checkpoint": "step_600",
|
| 79 |
+
"dominated_by": "step_800",
|
| 80 |
+
"reason": "dominator has no lower weighted_quality, no higher risk_score, and no lower stability_floor"
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"checkpoint": "step_700",
|
| 84 |
+
"dominated_by": "step_800",
|
| 85 |
+
"reason": "dominator has no lower weighted_quality, no higher risk_score, and no lower stability_floor"
|
| 86 |
+
}
|
| 87 |
+
],
|
| 88 |
+
"readback_expected_files": [
|
| 89 |
+
"README.md",
|
| 90 |
+
"audit/artifact_manifest.json",
|
| 91 |
+
"audit/checkpoint_lineage.json",
|
| 92 |
+
"audit/frontier_matrix.csv",
|
| 93 |
+
"audit/pairwise_dominance.jsonl",
|
| 94 |
+
"audit/readme_patch_log.jsonl",
|
| 95 |
+
"audit/reproducibility_bundle.tar.gz",
|
| 96 |
+
"audit/selection_certificate.json",
|
| 97 |
+
"audit/stress_trace.jsonl",
|
| 98 |
+
"config.json",
|
| 99 |
+
"figures/fig1.png",
|
| 100 |
+
"figures/fig2.png",
|
| 101 |
+
"figures/fig3.png",
|
| 102 |
+
"frontier_report.json",
|
| 103 |
+
"pytorch_model.bin"
|
| 104 |
+
]
|
| 105 |
+
}
|