vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2025-09-11 07:44:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix local_chat.py chunk_size not effect experts

Browse source
This commit is contained in:
qiyuxinlin 2025-05-23 02:35:01 +00:00
parent adc0906967
commit 71a5fc5770
2 changed files with 3 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

1
ktransformers/local_chat.py
View file

@ -70,6 +70,7 @@ def local_chat(
torch.set_grad_enabled(False) torch.set_grad_enabled(False)
Config().cpu_infer = cpu_infer Config().cpu_infer = cpu_infer
Config().chunk_size = chunk_size
if torch.xpu.is_available(): if torch.xpu.is_available():
use_cuda_graph = False use_cuda_graph = False

4
ktransformers/operators/experts.py
View file

@ -213,7 +213,7 @@ class KExpertsCPU(KExpertsBase):
self.config.num_experts_per_tok, self.config.num_experts_per_tok,
self.config.hidden_size, self.config.hidden_size,
self.config.moe_intermediate_size, self.config.moe_intermediate_size,
max(cuda_graphs), max(cuda_graphs) if isinstance(cuda_graphs, list) else Config().chunk_size,
gate_ptr, gate_ptr,
up_ptr, up_ptr,
down_ptr, down_ptr,
@ -231,7 +231,7 @@ class KExpertsCPU(KExpertsBase):
self.config.num_experts_per_tok, self.config.num_experts_per_tok,
self.config.hidden_size, self.config.hidden_size,
self.config.moe_intermediate_size, self.config.moe_intermediate_size,
max(cuda_graphs), max(cuda_graphs) if isinstance(cuda_graphs, list) else Config().chunk_size,
gate_ptr, gate_ptr,
up_ptr, up_ptr,
down_ptr, down_ptr,