Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 15 additions & 2 deletions paddleformers/datasets/template/template.py
Original file line number Diff line number Diff line change
Expand Up @@ -135,8 +135,19 @@ def _encode(
"""
system = system or self.default_system
encoded_messages = []
elements = []
last_mask = True
for i, message in enumerate(messages):
elements = []
if message["role"] == Role.USER or message["role"] == Role.OBSERVATION:
mask = True
elif message["role"] == Role.ASSISTANT or message["role"] == Role.FUNCTION:
mask = False
else:
raise NotImplementedError("Unexpected role: {}".format(message["role"]))

if mask != last_mask:
encoded_messages.append(self._convert_elements_to_ids(tokenizer, elements))
elements = []

if i == 0:
elements += self.format_prefix.apply()
Expand All @@ -159,7 +170,9 @@ def _encode(
else:
raise NotImplementedError("Unexpected role: {}".format(message["role"]))

encoded_messages.append(self._convert_elements_to_ids(tokenizer, elements))
last_mask = mask

encoded_messages.append(self._convert_elements_to_ids(tokenizer, elements))

return encoded_messages

Expand Down