native top_p_sampling (#2901)

2025-10-01 14:52:33 +08:00 · 2025-07-22 14:09:59 +08:00
parent 0eedbdaee0
commit 2c6a9e887e
14 changed files with 93 additions and 7 deletions
--- a/fastdeploy/model_executor/layers/linear.py
+++ b/fastdeploy/model_executor/layers/linear.py
@@ -61,6 +61,7 @@ class LinearBase(nn.Layer):
            or current_platform.is_xpu()
            or current_platform.is_iluvatar()
            or current_platform.is_gcu()
+            or current_platform.is_dcu()
        ):
            self.forward = self.forward_cuda
        else: