You can deploy this model and start earning money today!

TAUR-dev/
rankalign-v6-gemma-4-31B-it-d0.15-e1-humaneval-v2.1correct-upper-all-tcn-sm0.1

Name: rankalign-v6-gemma-4-31B-it-d0.15-e1-humaneval-v2.1correct-upper-all-tcn-sm0.1
Availability: InStock

google/gemma-4-31B-it

finetune

endpoints_compatible

Views

No views yet

rankalign-v6-gemma-4-31B-it-d0.15-e1-humaneval-v2.1correct-upper-all-tcn-sm0.1

Fine-tuned checkpoint from the rankalign project.

Training Details

Field	Value
Base model	`google/gemma-4-31B-it`
Version	v6
Task	`humaneval-v2.1correct-upper-all`
Epoch	1
Delta	0.15
Typicality correction	neg
Length normalization	False
Preference loss weight	1
NLL validator weight	0
NLL generator weight	0
Validator log-odds	False
Force same-x	False
Semi-supervised ratio	0.1
Labeled-only ratio	None

Reproducibility

Original checkpoint name:

v6-google--gemma-4-31B-it-delta0.15-epoch1--humaneval-v2.1correct-upper-all--d2g--random--alpha1.0--tc-neg--full-completion--semi0.1

To evaluate:

bash

python scripts/eval_by_claude.py \
    --model TAUR-dev/rankalign-v6-gemma-4-31B-it-d0.15-e1-humaneval-v2.1correct-upper-all-tcn-sm0.1 \
    --task humaneval-v2.1correct-upper-all \
    --split_type random --gen-shots zero --disc-shots few --validator-log-odds --save-scores-csv \
    --neg-typicality