openai / weak-to-strong - Conditional Complexity

Intro

Conditional complexity (also called cyclomatic complexity) is a term used to measure the complexity of software. The term refers to the number of possible paths through a program function. A higher value ofter means higher maintenance and testing costs (infosecinstitute.com).
Conditional complexity is calculated by counting all conditions in the program that can affect the execution path (e.g. if statement, loops, switches, and/or operators, try and catch blocks...).
Conditional complexity is measured at the unit level (methods, functions...).
Units are classified in four categories based on the measured McCabe index: 1-5 (simple units), 6-10 (medium complex units), 11-25 (complex units), 26+ (very complex units).

Learn more...

Conditional Complexity Overall

There are 58 units with 454 lines of code in units (26.2% of code).

0 very complex units (0 lines of code)
0 complex units (0 lines of code)
0 medium complex units (0 lines of code)
6 simple units (122 lines of code)
52 very simple units (332 lines of code)

Legend:

51+

26-50

11-25

6-10

1-5

Alternative Visuals

3D view of all units

Conditional Complexity per Extension

51+

26-50

11-25

6-10

1-5

Conditional Complexity per Logical Component

primary logical decomposition

51+

26-50

11-25

6-10

1-5

Most Complex Units

Top 50 most complex units

Unit	# lines	McCabe index	# params
def main() in sweep.py	25	8	3
def get_config_foldername() in train_simple.py	15	8	1
def eval_model_acc() in weak_to_strong/eval.py	33	7	3
def maybe_load_model() in weak_to_strong/train.py	15	6	1
def load_dataset() in weak_to_strong/datasets.py	20	6	3
def format_cosmosqa() in weak_to_strong/datasets.py	14	6	2
def clear_mem() in weak_to_strong/common.py	15	5	1
def get_model() in vision/run_weak_strong.py	13	4	1
def get_embeddings() in vision/run_weak_strong.py	21	4	2
def dumpkvs() in weak_to_strong/logger.py	6	3	1
def forward() in weak_to_strong/model.py	11	3	2
def main() in train_simple.py	23	2	25
def lr_schedule_fn() in weak_to_strong/train.py	5	2	1
def hf_loader() in weak_to_strong/datasets.py	4	2	2
def format_sciq() in weak_to_strong/datasets.py	8	2	2
def format_anthropic_hh() in weak_to_strong/datasets.py	4	2	2
def to_batch() in weak_to_strong/eval.py	3	2	2
def shutdown() in weak_to_strong/logger.py	3	2	1
def configure() in weak_to_strong/logger.py	5	2	1
def gradient_checkpointing_enable() in weak_to_strong/model.py	5	2	1
def main() in train_weak_to_strong.py	26	1	0
def train_model() in train_weak_to_strong.py	14	1	0
def train_model() in weak_to_strong/train.py	16	1	0
def train_and_save_model() in weak_to_strong/train.py	20	1	0
def register_dataset() in weak_to_strong/datasets.py	2	1	2
def tokenize_dataset() in weak_to_strong/datasets.py	4	1	0
def process_function() in weak_to_strong/datasets.py	5	1	1
def format_amazon_polarity() in weak_to_strong/datasets.py	2	1	2
def format_boolq() in weak_to_strong/datasets.py	4	1	2
def __call__() in weak_to_strong/loss.py	5	1	0
def __call__() in weak_to_strong/loss.py	2	1	0
def __init__() in weak_to_strong/loss.py	5	1	0
def __call__() in weak_to_strong/loss.py	5	1	0
def __init__() in weak_to_strong/loss.py	4	1	0
def __call__() in weak_to_strong/loss.py	5	1	0
def unpack() in weak_to_strong/eval.py	3	1	1
def get_tokenizer() in weak_to_strong/common.py	2	1	1
def append_to_jsonl() in weak_to_strong/logger.py	3	1	2
def __init__() in weak_to_strong/logger.py	3	1	0
def logkv() in weak_to_strong/logger.py	2	1	3
def logkvs() in weak_to_strong/logger.py	2	1	2
def is_configured() in weak_to_strong/logger.py	2	1	0
def get_current() in weak_to_strong/logger.py	3	1	0
def logkv() in weak_to_strong/logger.py	3	1	2
def logkvs() in weak_to_strong/logger.py	3	1	1
def dumpkvs() in weak_to_strong/logger.py	3	1	0
def shutdown() in weak_to_strong/logger.py	4	1	0
def __init__() in weak_to_strong/model.py	13	1	4
def from_pretrained() in weak_to_strong/model.py	2	1	3
def train_logreg() in vision/run_weak_strong.py	9	1	0