dylanyunlong/project_6
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

fix: slot_ids dtype long→int32 in cache test

Browse Source
This commit is contained in:
claude
2026-08-14 11:20:07 +00:00
parent 1f51feee05
commit 49cd7def89
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
qwen3_6_scripts/verify_ast_chain.py
View File

@@ -160,7 +160,7 @@ def test_cache():
n_blocks = 8 n_blocks = 8
block_size = 16 block_size = 16
slot_ids = torch.tensor([0, 1, 16, 17], dtype=torch.long, device="cuda") slot_ids = torch.tensor([0, 1, 16, 17], dtype=torch.int32, device="cuda")
keys = torch.randn(n_tokens, n_kv_heads, head_dim, dtype=torch.float16, device="cuda") keys = torch.randn(n_tokens, n_kv_heads, head_dim, dtype=torch.float16, device="cuda")
values = torch.randn(n_tokens, n_kv_heads, head_dim, dtype=torch.float16, device="cuda") values = torch.randn(n_tokens, n_kv_heads, head_dim, dtype=torch.float16, device="cuda")
key_cache = torch.zeros(n_blocks, block_size, n_kv_heads, head_dim, key_cache = torch.zeros(n_blocks, block_size, n_kv_heads, head_dim,