Merge branch 'master' into autosp

Author: sfc-gh-truwase

csrc/aio/py_lib/deepspeed_cpu_op.cpp

@@ -45,7 +45,9 @@ void cpu_op_desc_t::finish()
             if (_buffer.is_xpu()) { _buffer.copy_(_cpu_buffer.to(torch::kXPU)); }
             if (_buffer.is_cpu()) { _buffer.copy_(_cpu_buffer); }
 #if defined(__ENABLE_CANN__)
-            if (torch_npu::utils::is_npu(_buffer)) {
+            // `DS_BUILD_OPS=1 install.sh` complains that ‘torch_npu’ has not
+            // been declared, so inline `torch_npu::utils::is_npu`.
+            if (_buffer.is_privateuseone()) {
                 auto device = at::Device("npu:0");
                 _buffer.copy_(_cpu_buffer.to(device));
             }

deepspeed/module_inject/containers/bloom.py

@@ -26,8 +26,7 @@ def __init__(self, **kwargs):
         major = int(vers[0])
         minor = int(vers[1])
         if major > 4 or (major == 4 and minor > 43):
-            import sys
-            sys.exit(
+            raise RuntimeError(
                 f"Transformers version {v_transformers} exceeds version 4.43.4! After transformers version 4.43.4, BLOOM inference with DeepSpeed is no longer supported."
             )
 

deepspeed/runtime/zero/stage_1_and_2.py

@@ -1572,7 +1572,7 @@ def reduce_ipg_grads(self, comm_dtype=None):
                     param = self.bit16_groups[group_idx][param_idx_in_group]
 
                     assert self.params_already_reduced[param_id] == False, \
-                        f"The parameter {param_id} has already been reduced. \
+                        f"The parameter {debug_param2name(param)} has already been reduced. \
                         Gradient computed twice for this partition. \
                         Multiple gradient reduction is currently not supported"
 

op_builder/npu/async_io.py

@@ -25,7 +25,8 @@ def sources(self):
             'csrc/aio/py_lib/deepspeed_py_aio.cpp', 'csrc/aio/py_lib/deepspeed_py_aio_handle.cpp',
             'csrc/aio/py_lib/deepspeed_aio_thread.cpp', 'csrc/aio/common/deepspeed_aio_utils.cpp',
             'csrc/aio/common/deepspeed_aio_common.cpp', 'csrc/aio/common/deepspeed_aio_types.cpp',
-            'csrc/aio/py_lib/deepspeed_pin_tensor.cpp'
+            'csrc/aio/py_lib/deepspeed_pin_tensor.cpp', 'csrc/aio/py_lib/deepspeed_py_io_handle.cpp',
+            'csrc/aio/py_lib/deepspeed_aio_op_desc.cpp', 'csrc/aio/py_lib/deepspeed_cpu_op.cpp'
         ]
 
     def include_paths(self):