• File size measurements show the distribution of size of files.
• Files are classified in four categories based on their size (lines of code): 1-100 (very small files), 101-200 (small files), 201-500 (medium size files), 501-1000 (long files), 1001+(very long files).
• It is a good practice to keep files small. Long files may become "bloaters", code that have increased to such gargantuan proportions that they are hard to work with.

• There are 1,119 files with 158,056 lines of code.
• 16 very long files (22,598 lines of code)
• 41 long files (27,777 lines of code)
• 164 medium size files (48,984 lines of codeclsfd_ftr_w_mp_ins)
• 204 small files (29,559 lines of code)
• 694 very small files (29,138 lines of code)

There are 430 files with lines longer than 120 characters. In total, there are 2300 long lines.

File	# lines	# units	# long lines
moe.bzl in maga_transformer/cpp/cutlass	261	-	240
mma_utils.cuh in maga_transformer/cpp/deep_gemm/include	809	-	77
GptModel.cc in maga_transformer/cpp/models	1224	21	71
ArmLayerNormOp.cc in maga_transformer/cpp/devices/arm_impl	1337	16	51
NormalBatchStreamProcessor.cc in maga_transformer/cpp/normal_engine	415	5	47
layernorm_kernels.cu in maga_transformer/cpp/kernels	652	-	35
deepseek_v2.py in maga_transformer/models	310	13	34
decoder_masked_multihead_attention_launch.h in maga_transformer/cpp/kernels/decoder_masked_multihead_attention	228	4	34
int8_gemm_template.h in maga_transformer/cpp/cutlass/cutlass_kernels/int8_gemm	406	13	33
rmsnormKernels.cu in maga_transformer/cpp/kernels	443	-	26
weight_module.py in maga_transformer/model_loader	430	49	23
per_block_fp8_quant_weight.py in maga_transformer/model_loader	234	21	23
FfnLayer.cc in maga_transformer/cpp/devices/base_impl	308	7	23
gpt_neox_weight.py in maga_transformer/models	106	3	22
whisper_weight.py in maga_transformer/models	64	3	22
conversation.py in maga_transformer/openai/renderers	1228	10	22
llava_vit.py in maga_transformer/models	750	60	20
fp8_gemm.cuh in maga_transformer/cpp/deep_gemm/include	387	-	20
gpt_init_model_parameters.py in maga_transformer/config	641	22	18
ffn_weight.py in maga_transformer/model_loader	230	22	17
ArmAttentionOp.cc in maga_transformer/cpp/devices/arm_impl	610	15	17
gpt_weight.py in maga_transformer/models	40	1	16
custom_renderer.py in maga_transformer/openai/renderers	851	46	16
CudaMlaAttentionOp.cc in maga_transformer/cpp/devices/cuda_impl	261	5	16
llama_weight.py in maga_transformer/models	281	13	15
frontend_worker.py in maga_transformer/server	250	9	15
CudaLayernorm.cc in maga_transformer/cpp/devices/cuda_impl	287	2	14
model_weight.py in maga_transformer/utils	1001	80	13
glm_v2_weight.py in maga_transformer/models	59	2	13
EmbeddingExecutor.cc in maga_transformer/cpp/embedding_engine	232	11	13
gpt_kernels.cu in maga_transformer/cpp/kernels	1633	-	13
gen.py in maga_transformer/cpp/cutlass	1104	29	13
MultimodalProcessor.cc in maga_transformer/cpp/multimodal_processor	164	5	13
CudaDevice.cc in maga_transformer/cpp/devices/cuda_impl	619	31	13
qwen_v2.py in maga_transformer/models	194	14	12
weights_quant.py in maga_transformer/tools/quant	316	16	12
smooth_quant_weight.py in maga_transformer/model_loader	342	20	12
moe_gemm_kernels_template.h in maga_transformer/cpp/cutlass/cutlass_kernels/moe_gemm	878	20	12
ArmGemmOptOp.cc in maga_transformer/cpp/devices/arm_impl	145	3	12
ArmGemmPacking.cc in maga_transformer/cpp/devices/arm_impl/gemm_opt	767	12	12
CudaLoraLinear.cc in maga_transformer/cpp/devices/cuda_impl	243	2	12
SpeculativeEngine.cc in maga_transformer/cpp/speculative_engine	576	24	12
ScoreBatchStreamProcessor.cc in maga_transformer/cpp/speculative_engine/score_executor	143	2	12
bge_m3_embedding_module.py in maga_transformer/models/downstream_modules/embedding	73	9	11
omni_quant_weight.py in maga_transformer/model_loader	207	5	11
per_tensor_int8_quant_weight.py in maga_transformer/model_loader	221	10	11
unfused_attention_kernels.cu in maga_transformer/cpp/kernels	2425	-	11
cutlass_preprocessors.cc in maga_transformer/cpp/cutlass/cutlass_kernels	538	17	11
FfnLayerTest.hpp in maga_transformer/cpp/devices/base_tests	506	20	11
qwen2_vl.py in maga_transformer/models/qwen2_vl	152	15	10