MCPcopy Create free account
hub / github.com/zai-org/CodeGeeX / make_builder

Function make_builder

codegeex/megatron/data/indexed_dataset.py:55–61  ·  view source on GitHub ↗
(out_file, impl, vocab_size=None)

Source from the content-addressed store, hash-verified

53
54
55def make_builder(out_file, impl, vocab_size=None):
56 if impl == "mmap":
57 return MMapIndexedDatasetBuilder(
58 out_file, dtype=__best_fitting_dtype(vocab_size)
59 )
60 else:
61 return IndexedDatasetBuilder(out_file)
62
63
64def make_dataset(path, impl, skip_warmup=False):

Callers

nothing calls this directly

Calls 3

__best_fitting_dtypeFunction · 0.85

Tested by

no test coverage detected