vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2025-09-05 20:19:51 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

fix-gate-compile

Browse source
This commit is contained in:
Atream 2025-03-19 11:27:18 +08:00 • committed by GitHub
parent e788248364
commit 114995355b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 3 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
ktransformers/operators/gate.py
View file

@ -125,7 +125,7 @@ class KMoEGate(BaseInjectedModule, KMoEGateBase):
# adapted from https://github.com/vllm-project/vllm/blob/c77620d22d43daa7e0440e6267cbdd83f849ac64/vllm/model_executor/layers/fused_moe/fused_moe.py#L1071
# This is used by the Deepseek-V2 and Deepseek-V3 model
#@torch.compile(dynamic=True)
@torch.compile(dynamic=True)
def grouped_topk(hidden_states: torch.Tensor,
gating_output: torch.Tensor,
topk: int,
@ -225,9 +225,8 @@ class KMoEGateDeepSeekV3(BaseInjectedModule, KMoEGateBase):
hidden_states.type(torch.float32), self.weight.type(torch.float32), None
)
return grouped_topk(hidden_states, logits,
self.top_k, self.norm_topk_prob,
self.n_group, self.topk_group)
return grouped_topk(hidden_states, logits, self.top_k, self.norm_topk_prob,
self.n_group, self.topk_group, "sigmoid", self.e_score_correction_bias)
def load(self, w: dict | nn.Parameter | tuple | None = None, device: str|None = None):
if device is None: device = self.device