apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Remove CUDA ERROR 9 of inputs of get_padding_offset kernel (#5440)

Browse Source 
Co-authored-by: K11OntheBoat <“ruianmaidanglao@163.com”>
This commit is contained in:
K11OntheBoat
2025-12-09 14:17:30 +08:00
committed by GitHub
parent 76649b45c1
commit 8d99bac532
8 changed files with 97 additions and 167 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/layers/test_attention_layer.py
View File

@@ -270,10 +270,10 @@ class TestAttentionPerformance(unittest.TestCase):
partial_rotary_factor=fd_config.model_config.partial_rotary_factor,
)
input_ids = paddle.zeros([batch_size, max_model_len], dtype="int64")
token_num = paddle.sum(seq_lens_this_time)
input_ids = paddle.zeros([batch_size, seq_len if mode == ForwardMode.EXTEND else 1], dtype="int64")
token_num = np.sum(seq_lens_this_time)
ids_remove_padding, batch_id_per_token, cu_seqlens_q, cu_seqlens_k = get_padding_offset(
input_ids, token_num, seq_lens_this_time
input_ids, seq_lens_this_time, token_num
)
forward_meta = ForwardMeta(