decision-0.8b / evaluation /failed-bf16-probability-check.json
johnsonchromia's picture
Publish Decision-0.8B and verified benchmarks
cb17a40 verified
Raw History Blame Contribute Delete
1.88 kB
{
"status": "failed",
"verification": {
"records": 32,
"prediction_changes": 0,
"changed_indices": [],
"max_absolute_probability_difference": 0.020078718662261963,
"probability_atol": 0.02,
"passed": false,
"scope": "Before merge versus saved-and-reloaded BF16 export, raw probabilities",
"records_sha256": "21977c3441de6aff6f320084c0531e4131996b62b22c80b7d4c9cc158a7b6fa2",
"record_ids": [
"mnli:train:207170",
"boolq:train:3417",
"research-v21:dev:reasoning-methods:1:0:0",
"research-v21:dev:reasoning-methods:1:0:1",
"research-v21:dev:reasoning-methods:1:0:2",
"research-v21:dev:reasoning-methods:1:0:3",
"research-v21:dev:reasoning-methods:1:1:0",
"research-v21:dev:reasoning-methods:1:1:1",
"research-v21:dev:reasoning-methods:1:1:2",
"research-v21:dev:reasoning-methods:1:1:3",
"mnli:train:1737",
"helpsteer2:validation:4",
"helpsteer2:validation:5",
"research-v21:dev:model-efficiency:1:0:0",
"research-v21:dev:model-efficiency:1:0:1",
"research-v21:dev:model-efficiency:1:0:2",
"research-v21:dev:model-efficiency:1:0:3",
"research-v21:dev:model-efficiency:1:1:0",
"research-v21:dev:model-efficiency:1:1:1",
"research-v21:dev:model-efficiency:1:1:2",
"research-v21:dev:model-efficiency:1:1:3",
"go_emotions:validation:3362",
"mnli:train:229990",
"boolq:train:664",
"boolq:train:1952",
"policy_v2:dev:168:complete_positive",
"policy_v2:dev:168:decisive_missing",
"policy_v2:dev:168:multi_missing",
"policy_v2:dev:168:one_fact_flip",
"policy_v2:dev:168:partial_negative",
"policy_v2:dev:168:partial_positive",
"mnli:train:129242"
]
},
"source_manifest_sha256": "f4c0a1a02b43ee25da5901ade14cdf6d36ff991e29855d6274685d4b9d9b8def"
}