[Optimization] Use a separate driver when using Triton with Paddle (#6897)

---------

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

fastdeploy/model_executor/layers/batch_invariant_ops/batch_invariant_ops.py

@@ -6,6 +6,9 @@ from collections import namedtuple
 from collections.abc import Callable
 from typing import Any, Dict
 
+from fastdeploy.model_executor.ops.triton_ops.triton_utils import (
+    enable_compat_on_triton_kernel,
+)
 from fastdeploy.utils import get_logger
 
 logger = get_logger("worker_process", "worker_process.log")
@@ -37,6 +40,7 @@ def _matmul_launch_metadata(grid: Callable[..., Any], kernel: Any, args: Dict[st
     return ret
 
 
+@enable_compat_on_triton_kernel
 @triton.jit
 def _compute_pid(tile_id, num_pid_in_group, num_pid_m, GROUP_SIZE_M, NUM_SMS):
     group_id = tile_id // num_pid_in_group
@@ -47,6 +51,7 @@ def _compute_pid(tile_id, num_pid_in_group, num_pid_m, GROUP_SIZE_M, NUM_SMS):
     return pid_m, pid_n
 
 
+@enable_compat_on_triton_kernel
 @triton.jit(launch_metadata=_matmul_launch_metadata)
 def matmul_kernel_persistent(
     a_ptr,
@@ -226,6 +231,7 @@ def matmul_persistent(a: paddle.Tensor, b: paddle.Tensor, bias: paddle.Tensor |
     return c
 
 
+@enable_compat_on_triton_kernel
 @triton.jit
 def _log_softmax_kernel(
     input_ptr,
@@ -330,6 +336,7 @@ def log_softmax(input: paddle.Tensor, axis: int = -1) -> paddle.Tensor:
     return output.reshape(original_shape)
 
 
+@enable_compat_on_triton_kernel
 @triton.jit
 def mean_kernel(
     input_ptr,
@@ -475,6 +482,7 @@ def mean_dim(
 # We thank the SGLang authors and the Thinking Machines Lab for their contributions.
 
 
+@enable_compat_on_triton_kernel
 @triton.jit  # pragma: no cover
 def bmm_kernel_persistent(
     a_ptr,
@@ -724,6 +732,7 @@ def mean_batch_invariant(
 # ---------------------------------------------------------------------------
 
 
+@enable_compat_on_triton_kernel
 @triton.jit
 def _rms_norm_kernel(  # pragma: no cover
     input_ptr,