apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 16:48:03 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Bug Fix]Fix bug of append attention test case (#3202)
Some checks failed
Deploy GitHub Pages / deploy (push) Has been cancelled
Details

Browse Source
This commit is contained in:
RAM
2025-08-05 11:04:45 +08:00
committed by GitHub
parent e39159f3bd
commit c593e1a39c
1 changed files with 13 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
test/layers/test_append_attention.py
View File

@@ -352,6 +352,11 @@ class TestAppendGroupQueryAttnWithRope(unittest.TestCase):
self.max_dec_len_this_time = paddle.to_tensor([self.max_dec_len_this_time], "int32", place=paddle.CPUPlace())
self.seq_lens_this_time = self.seq_lens_encoder
self.decoder_batch_ids = paddle.full([self.batch_size], 0, dtype="int32")
self.decoder_tile_ids_per_batch = paddle.full([self.batch_size], 0, dtype="int32")
self.decoder_num_blocks_cpu = paddle.full([1], 0, dtype="int32").pin_memory()
self.max_len_tensor_cpu = paddle.full([8], 0, dtype="int32").cpu()
self.cache_shape = (
self.max_block_num,
self.kv_num_head,
@@ -414,16 +419,15 @@ class TestAppendGroupQueryAttnWithRope(unittest.TestCase):
kv_batch_ids,
kv_tile_ids_per_batch,
kv_num_blocks,
decoder_batch_ids,
decoder_tile_ids_per_batch,
decoder_num_blocks,
max_len_kv,
set_max_lengths,
) = get_block_shape_and_split_kv_block(
self.seq_lens_encoder,
self.seq_lens_decoder,
self.seq_lens_this_time,
self.cum_offset,
self.decoder_batch_ids,
self.decoder_tile_ids_per_batch,
self.decoder_num_blocks_cpu,
self.max_len_tensor_cpu,
64,
12,
(self.q_num_head + 2 * self.kv_num_head) // self.kv_num_head,
@@ -454,10 +458,10 @@ class TestAppendGroupQueryAttnWithRope(unittest.TestCase):
kv_batch_ids,
kv_tile_ids_per_batch,
kv_num_blocks,
decoder_batch_ids,
decoder_tile_ids_per_batch,
decoder_num_blocks,
set_max_lengths,
self.decoder_batch_ids,
self.decoder_tile_ids_per_batch,
self.decoder_num_blocks_cpu,
self.max_len_tensor_cpu,
max_len_kv,
self.rope_emb, # rope_emb
None, # attn_mask