EvalConfig
DataConfig
Bases: ConfigBaseModel
Data config
Source code in utu/config/eval_config.py
10 11 12 13 14 15 16 17 18 19 20 |
|
dataset
instance-attribute
dataset: str
Built-in dataset name or custom dataset path
type
instance-attribute
type: Literal['single', 'mixed']
Whether the dataset contains only single benchmark data or multiple benchmarks
question_field
instance-attribute
question_field: str
Question field name in the dataset
gt_field
instance-attribute
gt_field: str
Ground truth field name in the dataset
EvalConfig
Bases: ConfigBaseModel
Evaluation config
Source code in utu/config/eval_config.py
23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 |
|
exp_id
class-attribute
instance-attribute
exp_id: str = 'default'
Experiment ID
db_url
class-attribute
instance-attribute
db_url: str = getenv('DB_URL', 'sqlite:///tesxt.db')
Database URL
concurrency
instance-attribute
concurrency: int
Rollout parallelism
judge_model
class-attribute
instance-attribute
judge_model: ModelConfigs = Field(
default_factory=ModelConfigs
)
Judge model config
judge_concurrency
instance-attribute
judge_concurrency: int
Judgement parallelism
eval_method
class-attribute
instance-attribute
eval_method: str = None
Evaluation method