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

fix(CRITICAL): copy_blocks → vllm_copy_cache, swap_blocks → vllm_swap_blocks

Browse Source 
ixformer.functions exposes vllm_copy_cache and vllm_swap_blocks,
NOT copy_blocks/swap_blocks. Wrong function names crash engine
when prefix cache starts copying KV blocks (~9 min into eval).

Error was: AttributeError: module 'ixformer.functions' has no attribute 'copy_blocks'
at _custom_ops.py:1145 in copy_blocks
This commit is contained in:
project6-dev
2026-08-10 14:24:39 +00:00
parent 539d0fc6ff
commit 96a4afba43
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
qwen3_6_scripts/_custom_ops.py
View File

@@ -1142,12 +1142,12 @@ def reshape_and_cache_flashinfer(
def copy_blocks(key_caches: List[torch.Tensor],
value_caches: List[torch.Tensor],
block_mapping: torch.Tensor) -> None:
ixf_F.copy_blocks(key_caches, value_caches, block_mapping)
ixf_F.vllm_copy_cache(key_caches, value_caches, block_mapping)
def swap_blocks(src: torch.Tensor, dst: torch.Tensor,
block_mapping: torch.Tensor) -> None:
ixf_F.swap_blocks(src, dst, block_mapping)
ixf_F.vllm_swap_blocks(src, dst, block_mapping)
def convert_fp8(output: torch.Tensor,