|
| 1 | +# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved. |
| 2 | +# |
| 3 | +# Licensed under the Apache License, Version 2.0 (the "License"); |
| 4 | +# you may not use this file except in compliance with the License. |
| 5 | +# You may obtain a copy of the License at |
| 6 | +# |
| 7 | +# http://www.apache.org/licenses/LICENSE-2.0 |
| 8 | +# |
| 9 | +# Unless required by applicable law or agreed to in writing, software |
| 10 | +# distributed under the License is distributed on an "AS IS" BASIS, |
| 11 | +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 12 | +# See the License for the specific language governing permissions and |
| 13 | +# limitations under the License. |
| 14 | + |
| 15 | +import unittest |
| 16 | + |
| 17 | +import numpy as np |
| 18 | +import paddle |
| 19 | + |
| 20 | +from fastdeploy.model_executor.ops.gpu import get_padding_offset |
| 21 | + |
| 22 | + |
| 23 | +class TestGetPaddingOffset(unittest.TestCase): |
| 24 | + def test_get_padding_offset(self): |
| 25 | + max_len = 10 |
| 26 | + seq_lens = np.array([4, 3, 6], "int32").reshape(-1, 1) |
| 27 | + cum_offset = np.cumsum((max_len - seq_lens).flatten(), -1, "int32") |
| 28 | + token_num = np.sum(seq_lens) |
| 29 | + input_ids = np.array( |
| 30 | + [[8, 7, 8, 2, 0, 0, 0, 0, 0, 0], [4, 5, 5, 0, 0, 0, 0, 0, 0, 0], [7, 6, 1, 7, 2, 6, 0, 0, 0, 0]], "int64" |
| 31 | + ) |
| 32 | + ( |
| 33 | + x_remove_padding, |
| 34 | + batch_id_per_token, |
| 35 | + cu_seqlens_q, |
| 36 | + cu_seqlens_k, |
| 37 | + ) = get_padding_offset( |
| 38 | + paddle.to_tensor(input_ids), |
| 39 | + paddle.to_tensor(cum_offset), |
| 40 | + paddle.to_tensor(token_num), |
| 41 | + paddle.to_tensor(seq_lens), |
| 42 | + ) |
| 43 | + |
| 44 | + ref_x_remove_padding = np.array([8, 7, 8, 2, 4, 5, 5, 7, 6, 1, 7, 2, 6], "int64") |
| 45 | + ref_batch_id_per_token = np.array([0, 0, 0, 0, 1, 1, 1, 2, 2, 2, 2, 2, 2], "int32") |
| 46 | + ref_cu_seqlens_q = np.array([0, 4, 7, 13], "int32") |
| 47 | + ref_cu_seqlens_k = np.array([0, 4, 7, 13], "int32") |
| 48 | + |
| 49 | + np.testing.assert_allclose(x_remove_padding.numpy(), ref_x_remove_padding) |
| 50 | + np.testing.assert_allclose(batch_id_per_token.numpy(), ref_batch_id_per_token) |
| 51 | + np.testing.assert_allclose(cu_seqlens_q.numpy(), ref_cu_seqlens_q) |
| 52 | + np.testing.assert_allclose(cu_seqlens_k.numpy(), ref_cu_seqlens_k) |
| 53 | + |
| 54 | + |
| 55 | +if __name__ == "__main__": |
| 56 | + unittest.main() |
0 commit comments