Well, that's impossible. That's how bulk_create is defined:
def bulk_create(self, objs, batch_size=None):
for parent in self.model._meta.get_parent_list():
if parent._meta.concrete_model is not self.model._meta.concrete_model:
raise ValueError("Can't bulk create a multi-table inherited model")
if not objs:
return objs
self._for_write = True
connection = connections[self.db]
fields = self.model._meta.concrete_fields
objs = list(objs)
self._populate_pk_values(objs)
with transaction.atomic(using=self.db, savepoint=False):
if (connection.features.can_combine_inserts_with_and_without_auto_increment_pk
and self.model._meta.has_auto_field):
self._batched_insert(objs, fields, batch_size)
else:
objs_with_pk, objs_without_pk = partition(lambda o: o.pk is None, objs)
if objs_with_pk:
self._batched_insert(objs_with_pk, fields, batch_size)
if objs_without_pk:
fields = [f for f in fields if not isinstance(f, AutoField)]
self._batched_insert(objs_without_pk, fields, batch_size)
return objs
And the _batched_insert:
def _batched_insert(self, objs, fields, batch_size):
"""
A little helper method for bulk_insert to insert the bulk one batch
at a time. Inserts recursively a batch from the front of the bulk and
then _batched_insert() the remaining objects again.
"""
if not objs:
return
ops = connections[self.db].ops
batch_size = (batch_size or max(ops.bulk_batch_size(fields, objs), 1))
for batch in [objs[i:i + batch_size]
for i in range(0, len(objs), batch_size)]:
self.model._base_manager._insert(batch, fields=fields,
using=self.db)
So, as you can see bulk_create is basically a for loop inside an transaction.atomic.
One more thing. It is also impossible to save only some entries inside the transaction block. It either is executed fully or not executed at all.