huggingface / optimum-neuron Unit Size

Intro

Unit size measurements show the distribution of size of units of code (methods, functions...).
Units are classified in four categories based on their size (lines of code): 1-20 (small units), 20-50 (medium size units), 51-100 (long units), 101+ (very long units).
You should aim at keeping units small (< 20 lines). Long units may become "bloaters", code that have increased to such gargantuan proportions that they are hard to work with.

Learn more...

Unit Size Overall

There are 1,089 units with 10,814 lines of code in units (36.3% of code).

4 very long units (724 lines of code)
15 long units (923 lines of code)
104 medium size units (2,969 lines of code)
165 small units (2,406 lines of code)
801 very small units (3,792 lines of code)

Legend:

101+

51-100

21-50

11-20

1-10

Unit Size per Extension

101+

51-100

21-50

11-20

1-10

Unit Size per Logical Component

primary logical decomposition

101+

51-100

21-50

11-20

1-10

Alternative Visuals

3D view of all units

Longest Units

Top 50 longest units

Unit	# lines	McCabe index	# params
def parse_args_neuronx() in optimum/commands/export/neuronx.py	270	1	1
def preshard_hook() in optimum/neuron/models/inference/backend/modules/attention/gqa.py	225	11	3
def __init__() in optimum/neuron/models/training/llama/modeling_llama.py	124	8	4
def parse_args_neuron() in optimum/commands/export/neuron.py	105	1	1
def main() in tools/auto_fill_neuron_cache.py	82	13	0
def create_accelerator_and_postprocess() in optimum/neuron/trainers.py	81	9	1
def forward() in optimum/exporters/neuron/model_wrappers.py	80	7	3
def _kernel_enabled_mlp() in optimum/neuron/models/inference/llama/modeling_llama.py	67	9	5
def training_function() in notebooks/text-classification/scripts/train.py	61	5	1
def main() in optimum/exporters/neuron/__main__.py	59	5	0
def main() in tools/create_examples_from_transformers.py	59	13	0
def __post_init__() in optimum/neuron/training_args.py	58	16	1
def parse_args() in notebooks/text-classification/scripts/train.py	55	2	0
def _get_model_param_count() in optimum/neuron/utils/training_utils.py	55	15	2
def forward() in optimum/neuron/models/training/loss_utils.py	55	8	6
def perform_prefill() in optimum/neuron/models/inference/backend/modules/attention/attention_base.py	55	7	7
def parse_args() in optimum/commands/neuron/cache.py	53	1	1
def neff_cache() in optimum/neuron/models/inference/backend/cache.py	52	6	1
def main() in benchmark/text-generation/performance/gen_barcharts.py	51	9	0
def generate() in optimum/neuron/utils/input_generators.py	50	10	5
def convert_mixtral_to_neuron_state_dict() in optimum/neuron/models/inference/mixtral/modeling_mixtral.py	49	5	3
def get_node_results() in benchmark/text-generation-inference/performance/tgi_live_metrics.py	47	11	1
def _maybe_log_save_evaluate() in optimum/neuron/trainers.py	47	12	8
def __init__() in optimum/neuron/models/inference/llama/modeling_llama.py	47	3	3
def get_relevant_parameter_names() in optimum/neuron/models/training/transformations_utils.py	46	14	2
def __init__() in optimum/neuron/models/training/llama/modeling_llama.py	45	2	3
def forward() in optimum/neuron/peft/tuners/lora/layer.py	43	11	4
def new_fwd() in optimum/neuron/generation/utils.py	42	14	2
def training_function() in notebooks/text-generation/scripts/sft_finetuning_qwen3.py	41	3	2
def __call__() in optimum/neuron/pipelines/diffusers/pipeline_controlnet_sd_xl.py	41	1	42
def run() in optimum/commands/env.py	40	4	1
def _save_xla() in optimum/neuron/trainers.py	40	13	2
def __init__() in optimum/neuron/models/inference/backend/config.py	40	2	0
def _load_models_to_neuron() in optimum/neuron/modeling_diffusion.py	37	19	3
def __init__() in optimum/neuron/peft/tuners/lora/layer.py	37	8	4
def cumsum() in optimum/neuron/models/inference/backend/modules/generation/sampling.py	37	8	3
def main() in benchmark/text-generation/performance/llama3.3-70b.py	36	4	0
def forward() in optimum/exporters/neuron/model_wrappers.py	36	7	2
def save_mutated_as_lora() in optimum/neuron/peft/peft_model.py	35	8	4
def __init__() in optimum/neuron/accelerate/state.py	35	9	3
def step() in optimum/neuron/accelerate/optimizer.py	35	12	2
def _neuron_config_init() in optimum/neuron/modeling_traced.py	34	6	2
def forward() in optimum/neuron/models/inference/backend/modules/decoder/decoder_wrapper.py	34	9	6
def _load_weights_from_path() in optimum/neuron/models/inference/backend/pretrained_model.py	33	5	2
def install_requirements() in optimum/neuron/utils/runner.py	32	7	3
def _xla_gather() in optimum/neuron/accelerate/utils/operations.py	32	9	2
def __init__() in optimum/neuron/modeling_diffusion.py	31	1	0
def create_optimizer() in optimum/neuron/trainers.py	31	13	1
def main_export() in optimum/exporters/neuron/__main__.py	30	1	0
def __call__() in optimum/neuron/pipelines/diffusers/pipeline_controlnet.py	30	1	0