Fix issues related to L1cpp tests

1,not find nvte_dgelu
2,fsdp_group is not none
3,CPUOffloadEnabled change to cpp_offload_v1
Signed-off-by: maxiao3 <maxiao3@sugon.com>

See merge request dcutoolkit/deeplearing/TransformerEngine!74

transformer_engine/common/__init__.py

@@ -392,10 +392,10 @@ if "NVTE_PROJECT_BUILDING" not in os.environ or bool(int(os.getenv("NVTE_RELEASE
         _CURAND_LIB_CTYPES = _load_curand()
         _CUBLAS_LIB_CTYPES = _load_nvidia_cuda_library("cublas")
         _CUDART_LIB_CTYPES = _load_nvidia_cuda_library("cuda_runtime")
-        _TE_LIB_CTYPES = _load_core_library()
 
         # Needed to find the correct headers for NVRTC kernels.
         if not os.getenv("NVTE_CUDA_INCLUDE_DIR") and _nvidia_cudart_include_dir():
             os.environ["NVTE_CUDA_INCLUDE_DIR"] = _nvidia_cudart_include_dir()
     except OSError:
         pass
+    _TE_LIB_CTYPES = _load_core_library()

transformer_engine/pytorch/module/batched_linear.py

@@ -605,7 +605,7 @@ class BatchedLinear(TransformerEngineBaseModule):
 
             weight_tensors_fp8 = [None] * int(self.num_gemms)
 
-            from ..cpu_offload import CPUOffloadEnabled
+            from ..cpu_offload_v1 import CPUOffloadEnabled
 
             if torch.is_grad_enabled():
                 linear_fn = _BatchLinear.apply

transformer_engine/pytorch/module/linear.py

@@ -117,7 +117,6 @@ class _Linear(torch.autograd.Function):
         ub_bulk_dgrad: bool,
         ub_bulk_wgrad: bool,
         ub_name: str,
-        fine_grained_activation_offloading: bool,
         fp8_output: bool,  # pylint: disable=unused-argument
         fsdp_group: Union[dist_group_type, None],
         module: torch.nn.Module,