14 duplicated lines in:
  - benchmarks/experimental/experimental_async_approaches.py (63:78, 2%)
  - benchmarks/models/transformer_lm.py (28:45, 12%)

10 duplicated lines in:
  - benchmarks/experimental/experimental_async_approaches.py (98:110, 1%)
  - benchmarks/models/transformer_lm.py (178:192, 9%)

6 duplicated lines in:
  - benchmarks/experimental/experimental_async_approaches.py (87:94, 1%)
  - benchmarks/models/transformer_lm.py (167:174, 5%)