iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

[CI/Build] Fix flaky failures in CPU CI image building (#55317)

Browse Source 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
This commit is contained in:
Li, Jiang
2026-09-04 20:27:02 +08:00
committed by GitHub
parent 8b6de0eb9a
commit 8340fe1bb9
4 changed files with 67 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
csrc/cpu/sgl-kernels/mla_cache.cpp
+3 -3
View File
@@ -59,9 +59,9 @@ void concat_and_cache_mla_kernel_impl(
// kv_c_normed : [num_tokens, kv_lora_rank]
// k_pe : [num_tokens, qk_rope_head_dim] or [num_tokens, 1,
// qk_rope_head_dim] kv_cache : [num_blocks, block_size, kv_lora_rank +
// qk_rope_head_dim] slot_mapping: [num_tokens] int32/int64, absolute physical
// row index
// qk_rope_head_dim]
// kv_cache : [num_blocks, block_size, kv_lora_rank + qk_rope_head_dim]
// slot_mapping: [num_tokens] int32/int64, absolute physical row index
// (block_id * block_size + block_offset); negative entries are
// skipped (padded tokens), matching `concat_and_cache_mla`.
void concat_and_cache_mla_cpu(const at::Tensor& kv_c_normed,