EngineX/xc-llm-kunlun
3
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

[Bugsfix] Fix run failed (#198)

Browse Source 
Signed-off-by: xyDong0223 <dongxinyu03@baidu.com>
This commit is contained in:
Xinyu Dong
2026-02-13 14:07:10 +08:00
committed by GitHub
parent bf9369f733
commit 76ec220b43
1 changed files with 7 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
vllm_kunlun/vllm_utils_wrapper.py
View File

@@ -386,8 +386,8 @@ def silu_and_mul_quant_xpu(
pass pass
import kunlun_ops # noqa: E402
import torch # noqa: E402 import torch # noqa: E402
import xtorch_ops # noqa: E402
from torch.library import custom_op, impl # noqa: E402 from torch.library import custom_op, impl # noqa: E402
@@ -1402,7 +1402,7 @@ def awq_dequantize_cuda(
device=qweight.device, device=qweight.device,
) )
group_m = int(qweight.shape[0] / scales.shape[0]) group_m = int(qweight.shape[0] / scales.shape[0])
out = kunlun_ops.awq_dequantize( kunlun_ops.awq_dequantize(
qweight=qweight, qweight=qweight,
scales=scales, scales=scales,
zeros=zeros, zeros=zeros,
@@ -2034,8 +2034,7 @@ def I8_paged_mqa_logits(
out: torch.Tensor, out: torch.Tensor,
use_xfa_boost: Optional[bool] = False, use_xfa_boost: Optional[bool] = False,
) -> None: ) -> None:
kunlun_ops.sparse_prefill_fwd_opt( kunlun_ops.I8_paged_mqa_logits(
.I8_paged_mqa_logits(
q=q, q=q,
fused_kv_cache=fused_kv_cache, fused_kv_cache=fused_kv_cache,
weights=weights, weights=weights,
@@ -2351,10 +2350,7 @@ def fast_topkv2(
score: torch.Tensor, lengths: torch.Tensor, topk: Optional[int] = 2048 score: torch.Tensor, lengths: torch.Tensor, topk: Optional[int] = 2048
) -> torch.Tensor: ) -> torch.Tensor:
assert topk == 2048, "fast_topkv2 only supports topk = 2048 by now" assert topk == 2048, "fast_topkv2 only supports topk = 2048 by now"
topk_indices = kunlun_ops.fast_topkv2( topk_indices = kunlun_ops.fast_topkv2(score=score, lengths=lengths, topk=topk)
score=score,
lengths=lengths,
topk=topk)
return topk_indices return topk_indices
@@ -2363,10 +2359,7 @@ def fast_topkv2_cuda(
score: torch.Tensor, lengths: torch.Tensor, topk: Optional[int] = 2048 score: torch.Tensor, lengths: torch.Tensor, topk: Optional[int] = 2048
) -> torch.Tensor: ) -> torch.Tensor:
assert topk == 2048, "fast_topkv2 only supports topk = 2048 by now" assert topk == 2048, "fast_topkv2 only supports topk = 2048 by now"
topk_indices = kunlun_ops.fast_topkv2( topk_indices = kunlun_ops.fast_topkv2(score=score, lengths=lengths, topk=topk)
score=score,
lengths=lengths,
topk=topk)
return topk_indices return topk_indices
@@ -2805,7 +2798,7 @@ def lora_matmul_inplace(
alpha: float = 1.0, alpha: float = 1.0,
beta: float = 1.0, beta: float = 1.0,
) -> None: ) -> None:
xtorch_ops.matmul( kunlun_ops.matmul(
x=x.contiguous(), x=x.contiguous(),
w=w.contiguous(), w=w.contiguous(),
out=output_tensor, out=output_tensor,
@@ -2826,7 +2819,7 @@ def lora_matmul_inplace_cuda(
alpha: float = 1.0, alpha: float = 1.0,
beta: float = 1.0, beta: float = 1.0,
) -> None: ) -> None:
xtorch_ops.matmul( kunlun_ops.matmul(
x=x.contiguous(), x=x.contiguous(),
w=w.contiguous(), w=w.contiguous(),
out=output_tensor, out=output_tensor,