• There are 71 files with 11,276 lines of code.
• 1 very long files (1,319 lines of code)
• 4 long files (2,534 lines of code)
• 13 medium size files (4,266 lines of code)
• 11 small files (1,611 lines of code)
• 42 very small files (1,546 lines of code)

File	# lines	≈ tokens to read	# changes (1y)	≈ lines edited per change	lines read per line changed
base_evaluator.py in deepspec/eval	643	~3k–10k	2	~2	~300
modeling.py in deepspec/modeling/dspark/gemma4	568	~3k–10k	2	~1	~500
base_trainer.py in deepspec/trainer	364	~2k–7k	3	~3	~100
modeling.py in deepspec/modeling/dspark/qwen3	494	~3k–7k	2	~1	~500
prepare_target_cache.py in scripts/data	365	~2k–7k	2	~2	~200
distributed.py in deepspec/utils	106	~700–1.5k	3	~3	~30
optim.py in deepspec/utils	124	~700–2k	2	~2	~70
dspark_gemma4_12b.py in config/dspark	61	~300–1k	3	~3	~20
dspark_qwen3_14b.py in config/dspark	61	~300–1k	3	~3	~20
dflash_gemma4_12b.py in config/dflash	59	~300–1k	3	~3	~20
dflash_qwen3_14b.py in config/dflash	59	~300–1k	3	~3	~20
dspark_qwen3_4b.py in config/dspark	53	~300–1k	3	~3	~15
dspark_qwen3_8b.py in config/dspark	53	~300–1k	3	~3	~15
dflash_qwen3_4b.py in config/dflash	51	~300–1k	3	~3	~15
dflash_qwen3_8b.py in config/dflash	51	~300–1k	3	~3	~15
jsonl_dataset.py in deepspec/data	117	~700–2k	1	~5	~20
train.py in root	38	under 1k	2	~2	~20
eagle3_gemma4_12b.py in config/eagle3	53	~300–1k	1	~5	~10
eagle3_qwen3_14b.py in config/eagle3	53	~300–1k	1	~5	~10
eagle3_qwen3_4b.py in config/eagle3	45	under 1k	1	~7	~5

Tokens are lines × 7 to 14 (analysis.fileReadsForChanges). Lines edited per change are the file's lines added and deleted divided by its commits over the whole history. Agents read about 2,000 lines at a time: when a 17,000-line file was split into modules, the tokens an agent read for the same change fell by 83% (The Economic Benefit of Refactoring, 2026).

There are 3 files with lines longer than 120 characters. In total, there are 1370 long lines.

File	# lines	# units	# long lines
jsonl gsm8k.jsonl in eval_datasets	1319	-	1319
jsonl mbpp.jsonl in eval_datasets	257	-	50
eval.py in root	59	2	1

• 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.