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

Method _do_init

codegeex/megatron/data/indexed_dataset.py:480–492  ·  view source on GitHub ↗
(self, path, skip_warmup)

Source from the content-addressed store, hash-verified

478 self._do_init(state)
479
480 def _do_init(self, path, skip_warmup):
481 self._path = path
482 self._index = self.Index(index_file_path(self._path), skip_warmup)
483
484 if not skip_warmup:
485 print_rank_0(" warming up data mmap file...")
486 _warmup_mmap_file(data_file_path(self._path))
487 print_rank_0(" creating numpy buffer of mmap...")
488 self._bin_buffer_mmap = np.memmap(
489 data_file_path(self._path), mode="r", order="C"
490 )
491 print_rank_0(" creating memory view of numpy buffer...")
492 self._bin_buffer = memoryview(self._bin_buffer_mmap)
493
494 def __del__(self):
495 self._bin_buffer_mmap._mmap.close()

Callers 2

__init__Method · 0.95
__setstate__Method · 0.95

Calls 4

print_rank_0Function · 0.90
index_file_pathFunction · 0.85
_warmup_mmap_fileFunction · 0.85
data_file_pathFunction · 0.85

Tested by

no test coverage detected