# Copyright 2020-2025 The HuggingFace Team. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. import unittest from trl.rewards import think_format_reward class ThinkFormatRewardTester(unittest.TestCase): def test_valid_format(self): completions = [ "This is my reasoning.This is my answer.", # Simple, one-line reasoning "\nThis is my reasoning.\n\nThis is my answer.", # Multiline reasoning "\nThis is\nmy reasoning.\n\nThis is my answer.", # Multiline reasoning "\nThis is my reasoning.\nThis is my answer.", # Reasoning including other tags "\nThis is my answer.", # Empty reasoning ] completions = [[{"content": completion}] for completion in completions] expected_rewards = [1.0, 1.0, 1.0, 1.0, 1.0] # All should be valid rewards = think_format_reward(completions) self.assertEqual(rewards, expected_rewards) def test_invalid_format(self): completions = [ "\nThis is my reasoning.\nThis is my answer.", # No closing "This is my reasoning.\nThis is my answer.", # No closing "This is my reasoning. This is my answer.", # No tags "This is my reasoning.\nThis is my answer.", # No tags "This is my reasoning.\nThis is my answer.", # No opening "This is my reasoning.This is my answer.", # No opening "Thisis my reasoning.\nThis is my answer.", # tag in the middle "This ismy reasoning.This is my answer.", # Nested tags "This is\nmy\nreasoning.\nThis is my answer.", # Multiline ] completions = [[{"content": completion}] for completion in completions] expected_rewards = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0] # All should be invalid rewards = think_format_reward(completions) self.assertEqual(rewards, expected_rewards) def test_mixed_format(self): completions = [ "This is my reasoning.This is my answer.", # Valid "\nThis is my reasoning.\n\nThis is my answer.", # Valid "This is my reasoning.\nThis is my answer.", # Invalid "This is my reasoning. This is my answer.", # Invalid ] completions = [[{"content": completion}] for completion in completions] expected_rewards = [1.0, 1.0, 0.0, 0.0] rewards = think_format_reward(completions) self.assertEqual(rewards, expected_rewards) if __name__ == "__main__": unittest.main()