xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-01-24 00:14:34 +08:00
Code Issues Packages Projects Releases Wiki Activity

[CPU][Bugfix] Fix _to_list in CPU model runner (#28824)

Browse Source 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
This commit is contained in:
Li, Jiang 2025-11-17 15:47:24 +08:00 committed by GitHub
parent 3380ed5e11
commit 577bb34fff
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 8 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
csrc/cpu/torch_bindings.cpp
View File

@ -100,6 +100,9 @@ void cpu_attention_with_kv_cache(
const torch::Tensor& scheduler_metadata,
const std::optional<torch::Tensor>& s_aux);
// Note: just for avoiding importing errors
void placeholder_op() { TORCH_CHECK(false, "Unimplemented"); }
TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
// vLLM custom ops
@ -275,6 +278,11 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
"sliding_window_left, SymInt sliding_window_right, Tensor block_table, "
"float softcap, Tensor sheduler_metadata, Tensor? s_aux) -> ()",
&cpu_attention_with_kv_cache);
// placeholders
ops.def("static_scaled_fp8_quant() -> ()", placeholder_op);
ops.def("dynamic_scaled_fp8_quant() -> ()", placeholder_op);
ops.def("dynamic_per_token_scaled_fp8_quant() -> ()", placeholder_op);
}
TORCH_LIBRARY_EXPAND(CONCAT(TORCH_EXTENSION_NAME, _utils), utils) {

3
vllm/v1/worker/cpu_model_runner.py
View File

@ -80,9 +80,6 @@ class CPUModelRunner(GPUModelRunner):
def _sync_device(self) -> None:
pass
def _to_list(self, sampled_token_ids: torch.Tensor) -> list[list[int]]:
return sampled_token_ids.tolist()
def get_dp_padding(self, num_tokens: int) -> tuple[int, torch.Tensor | None]:
# Note: For CPU backend, dp padding is not required for now.
return 0, None