| { |
| "format": "d1a-torch", |
| "format_version": 1, |
| "base": "google/gemma-4-E4B-it", |
| "base_revision": "ee0ef6023621cff504d758262d4e04895a5af4a2", |
| "lora": 16, |
| "head_dim": 256, |
| "option_isolation": false, |
| "special_embeddings": false, |
| "weights_dtype": "bf16", |
| "temperature": 1.5511447618337346, |
| "holdout": [], |
| "weights": "lora", |
| "head": "head.safetensors", |
| "extra": { |
| "args": { |
| "base": "google/gemma-4-E4B-it", |
| "n_per_source": 1000, |
| "epochs": 1, |
| "lr": 5e-05, |
| "head_lr": 0.0, |
| "weight_decay": 0.01, |
| "lora": 16, |
| "accum": 4, |
| "holdout": "", |
| "suite": "evals/v7/decision-v7", |
| "train_sources": "", |
| "device": "mps", |
| "batch": 2, |
| "dtype": "fp32", |
| "weights_dtype": "bf16", |
| "checkpointing": 1, |
| "head_dim": 256, |
| "lora_targets": "all", |
| "base_revision": "ee0ef6023621cff504d758262d4e04895a5af4a2", |
| "p_none": 0.1, |
| "p_none_distract": 0.12, |
| "p_distract": 0.15, |
| "p_none_pair": 0.0, |
| "none_pair_max_state": null, |
| "synthetic_repeat": 1, |
| "public_frac": 1.0, |
| "out": "stage2", |
| "data": "data/train-e2b-v06.jsonl", |
| "max_state": 5120, |
| "extra_suites": "", |
| "replay": 0, |
| "init_from": "stage1", |
| "allow_missing_sources": "", |
| "reason": "", |
| "row_budget": 0, |
| "shared_prefix": 0, |
| "length_sort": 0, |
| "pass_tokens_max": 0, |
| "max_steps": 0, |
| "save_every_steps": 0, |
| "save_every_minutes": 30.0, |
| "resume": 1, |
| "stop_after": 0, |
| "seed": 0 |
| }, |
| "suite_sha256": "a8f50e481b7d90b97da049e0ff6a01cee2f1ed204aed61a8265af0edbb5514d2", |
| "init_source": { |
| "init_from": "stage1", |
| "resolved": "stage1", |
| "weights_sha256": "bd217060dca2d0ef7d07a103e2bb0c8a4a505502b315b91194a3caa38ba533cd", |
| "head_sha256": "e9f13355196677f3bea172ffe26d17bbe4276098f08afaf28db317341b39bd14", |
| "tensors": 516 |
| }, |
| "temperature_fit": { |
| "rows": [ |
| "cal/rows.json", |
| "calpr/rows.json" |
| ], |
| "n": 3178, |
| "method": "min micro mean NLL over a 121-point log grid 0.25..4", |
| "cross_validation": { |
| "raw": { |
| "n": 3178, |
| "ece": 0.05657225201118034, |
| "brier": 0.25128312367981526, |
| "nll": 0.5044515701034964, |
| "confident_error_rate": 0.03524229074889868, |
| "coverage_at_5pct_error": 0.6680302076777848 |
| }, |
| "out_of_fold": { |
| "n": 3178, |
| "ece": 0.015396958922541597, |
| "brier": 0.24360429314644466, |
| "nll": 0.45865013254985726, |
| "confident_error_rate": 0.016047828823159218, |
| "coverage_at_5pct_error": 0.6626809314033983 |
| }, |
| "ece_ci95": { |
| "raw": [ |
| 0.04573724463444184, |
| 0.06868531603386199 |
| ], |
| "out_of_fold": [ |
| 0.010910223636218403, |
| 0.02979139359871167 |
| ], |
| "delta": [ |
| -0.05198077099251131, |
| -0.02258946367482048 |
| ] |
| }, |
| "separated": true, |
| "temperatures": [ |
| 1.5511447618337346, |
| 1.5511447618337346, |
| 1.5157165665103978, |
| 1.5874010519681994, |
| 1.5511447618337346 |
| ], |
| "folds": 5, |
| "seed": 0, |
| "samples": 1000, |
| "groups": 1670, |
| "unit": "source-stratified (source, group); sibling questions and variants stay together" |
| }, |
| "interval": { |
| "level": 0.9, |
| "samples": 2000, |
| "seed": 0, |
| "lower": 1.4810975522865641, |
| "upper": 1.6245047927124707, |
| "questions": 3178 |
| }, |
| "fit_rows": [ |
| { |
| "rows": "cal/rows.json", |
| "suite": "evals/v7/decision-v7", |
| "split": "calibration", |
| "questions": 1148 |
| }, |
| { |
| "rows": "calpr/rows.json", |
| "suite": "evals/d1a/pr-labels", |
| "split": "development", |
| "questions": 2030 |
| } |
| ], |
| "training_suites": [ |
| "evals/devtools-v1", |
| "evals/documents-v1", |
| "evals/hard-v1", |
| "evals/v7/decision-v7" |
| ], |
| "training_partitions": [ |
| "evals/d1a/ja-jglue:train", |
| "evals/d1a/pr-labels:train", |
| "evals/d1a/pr-labels:train-blast", |
| "evals/d1a/pr-labels:train-ja", |
| "evals/d1a/routing:factory-train", |
| "evals/d1a/routing:generic-train" |
| ], |
| "same_corpus": [ |
| "calpr/rows.json: SAME CORPUS: evals/d1a/pr-labels:development is an eval partition of evals/d1a/pr-labels, whose train partitions ['evals/d1a/pr-labels:train', 'evals/d1a/pr-labels:train-blast', 'evals/d1a/pr-labels:train-ja'] the checkpoint trained on" |
| ], |
| "in_distribution": { |
| "allowed": true, |
| "problems": [ |
| "cal/rows.json: evals/v7/decision-v7 is training data of the checkpoint", |
| "cal/rows.json: it pools 12 training source(s) ['agnews', 'amazon', 'banking77', 'boolq', 'compositional', 'dbpedia14', 'imdb', 'legacy_policy'] ...", |
| "cal/rows.json: it reads the calibration partition of evals/v7/decision-v7, a training corpus", |
| "calpr/rows.json: SAME CORPUS: evals/d1a/pr-labels:development is an eval partition of evals/d1a/pr-labels, whose train partitions ['evals/d1a/pr-labels:train', 'evals/d1a/pr-labels:train-blast', 'evals/d1a/pr-labels:train-ja'] the checkpoint trained on", |
| "cannot list the sources of training source pr-labels:fresh-fit (no suite or partition of this checkout), training data of stage2" |
| ] |
| } |
| }, |
| "use_case_temperatures": {}, |
| "use_case_temperature_fits": { |
| "routing": { |
| "rows": [ |
| "rt-gen/rows.json", |
| "rt-hand/rows.json" |
| ], |
| "n": 315, |
| "method": "min micro mean NLL over a 121-point log grid 0.25..4", |
| "cross_validation": { |
| "raw": { |
| "n": 315, |
| "ece": 0.014988591883350024, |
| "brier": 0.02933234182471159, |
| "nll": 0.04260048719028577, |
| "confident_error_rate": 0.0031746031746031746, |
| "coverage_at_5pct_error": 1.0 |
| }, |
| "out_of_fold": { |
| "n": 315, |
| "ece": 0.012868974641462564, |
| "brier": 0.027818845682320778, |
| "nll": 0.04164313674943866, |
| "confident_error_rate": 0.0031746031746031746, |
| "coverage_at_5pct_error": 1.0 |
| }, |
| "ece_ci95": { |
| "raw": [ |
| 0.007028569881564169, |
| 0.03002489671354236 |
| ], |
| "out_of_fold": [ |
| 0.006437557142966775, |
| 0.02839721383944503 |
| ], |
| "delta": [ |
| -0.005527160442421324, |
| 0.002748606576340939 |
| ] |
| }, |
| "separated": false, |
| "temperatures": [ |
| 1.2599210498948732, |
| 1.1755479062836087, |
| 1.148698354997035, |
| 1.2030250360821164, |
| 1.2030250360821164 |
| ], |
| "folds": 5, |
| "seed": 0, |
| "samples": 1000, |
| "groups": 315, |
| "unit": "source-stratified (source, group); sibling questions and variants stay together" |
| }, |
| "interval": { |
| "level": 0.9, |
| "samples": 2000, |
| "seed": 0, |
| "lower": 0.9117224885582169, |
| "upper": 1.381912879967776, |
| "questions": 315 |
| }, |
| "fit_rows": [ |
| { |
| "rows": "rt-gen/rows.json", |
| "suite": "evals/d1a/routing", |
| "split": "generic-development", |
| "questions": 270 |
| }, |
| { |
| "rows": "rt-hand/rows.json", |
| "suite": "evals/d1a/routing", |
| "split": "handlabelled-45", |
| "questions": 45 |
| } |
| ], |
| "training_suites": [ |
| "evals/devtools-v1", |
| "evals/documents-v1", |
| "evals/hard-v1", |
| "evals/v7/decision-v7" |
| ], |
| "training_partitions": [ |
| "evals/d1a/ja-jglue:train", |
| "evals/d1a/pr-labels:train", |
| "evals/d1a/pr-labels:train-blast", |
| "evals/d1a/pr-labels:train-ja", |
| "evals/d1a/routing:factory-train", |
| "evals/d1a/routing:generic-train" |
| ], |
| "same_corpus": [ |
| "rt-gen/rows.json: SAME CORPUS: evals/d1a/routing:generic-development is an eval partition of evals/d1a/routing, whose train partitions ['evals/d1a/routing:factory-train', 'evals/d1a/routing:generic-train'] the checkpoint trained on", |
| "rt-hand/rows.json: SAME CORPUS: evals/d1a/routing:handlabelled-45 is an eval partition of evals/d1a/routing, whose train partitions ['evals/d1a/routing:factory-train', 'evals/d1a/routing:generic-train'] the checkpoint trained on" |
| ], |
| "in_distribution": { |
| "allowed": true, |
| "problems": [ |
| "rt-gen/rows.json: SAME CORPUS: evals/d1a/routing:generic-development is an eval partition of evals/d1a/routing, whose train partitions ['evals/d1a/routing:factory-train', 'evals/d1a/routing:generic-train'] the checkpoint trained on", |
| "rt-hand/rows.json: SAME CORPUS: evals/d1a/routing:handlabelled-45 is an eval partition of evals/d1a/routing, whose train partitions ['evals/d1a/routing:factory-train', 'evals/d1a/routing:generic-train'] the checkpoint trained on", |
| "cannot list the sources of training source pr-labels:fresh-fit (no suite or partition of this checkout), training data of stage2" |
| ] |
| } |
| } |
| } |
| } |
| } |
